Roxochelys wanderleyi, , MCT, 1722

Gaffney, Eugene S., Meylan, Peter A., Wood, Roger C., Simons, Elwyn & De Almeida Campos, Diogenes, 2011, Evolution Of The Side-Necked Turtles: The Family Podocnemididae, Bulletin of the American Museum of Natural History 2011 (350), pp. 1-237 : 196-224

wanderleyi . The surface of the shell of Podocnemis is typically rather smooth.

Nuchal contacts and shape: The nuchal bone in Bauruemys elegans , as in nearly all podocnemidids, contacts the first peripheral anterolaterally, the first costal posterolaterally, and the first neural posteromedially. The only exception to these contacts occurs in those forms in which the first costals meet on the midline between the nuchal and first neural. The nuchal bone in Bauruemys is slightly wider than long in all specimens available to us (fig. 94). It is about 15% of carapace length. It narrows anteriorly more than in Roxochelys but less than in Lapparentemys . The widest point occurs threequarters of the way along its length so the nuchal appears more hexagonal in Bauruemys than in species in which this point is more posterior ( Cambaremys , Portezueloemys , Podocnemis ). The nuchal in Roxochelys is also wider than long (fig. 94) and does not narrow markedly along the anterior margin. The anterior margin is the widest among the taxa of interest, about 67% of maximum nuchal width. In Lapparentemys the anterior margin of the nuchal bone is very narrow and the anterolateral sides converge more noticeably. The anterior margin is about 42% of maximum nuchal width.

In all of the South American taxa included here, as in all known pelomedusoides, the cervical scale is absent. The first pleural scales do not reach the nuchal so this element is covered by the first vertebral scale posteriorly and the paired first marginal scales anteriorly. The first pleural contact with the nuchal appears to be absent among podocnemidids. Among bothremydids, it is absent from all known nuchal bones except for those of two genera within the Bothremydini, Chedighaii and Araiochleys ( Gaffney et al., 2006).

Much of the variation in the shape of the nuchal bone reflects differences in the relative lengths of the contact of the first costal and first peripheral. In some forms the contacts of the first costal are shorter, and the suture is more transverse (most Podocnemis ), and the overall nuchal shape is more trapezoidal. In others this contact is longer and at a lower angle relative to the midline making the nuchal more hexagonal ( Bauruemys ). Forms with longer nuchal bones generally have longer contacts between the nuchal and first peripherals. Variation in nuchal shape is not well suited to treatment as a discretely variable character; however, figure 94 suggests that four of the taxa of interest have nuchals that are nearly always wider than long and two have nuchals that are nearly always longer than wide.

There are two distinct nuchal morphologies among the unnamed material from Peirópolis. The more common one, presumed to be Peirópolis A, is a longer nuchal with the widest point more posterior, as in Cambaremys and Podocnemis . The other, treated here as Peirópolis B, is shorter and wider anteriorly, more similar to Bauruemys elegans and Roxochelys . In Peirópolis A, the nuchal—first peripheral suture is markedly sinuous in the specimen used in the reconstruction (figs. 91, 95). This suture is straight in Cambaremys . In Peirópolis A there is also a slight midline notch and the anterior margin is slightly scalloped. In Peirópolis B, as in Bauruemys elegans and Lapparentemys , the anterior margin is straight.

Neural series: The neural series in Bauruemys is typically made up of six elements (fig. 88). In most specimens we have seen, the second neural is four sided and the seventh and eighth costals meet on the midline between the last neural and a single suprapygal. In all six of the South American podocnemidids in this study the neural series contacts the nuchal anteriorly. Contact of the neural series to the suprapygal does not occur in podocnemidids, but some have longer neural series than that seen in typical Bauruemys . One of seven complete neural series in Bauruemys includes a small seventh neural that partially separates the seventh costals (MN 7017-V; Romano, 2008), but it is asymmetrical and appears to be anomalous. In Roxochelys wanderleyi there is a seventh neural present in one specimen (MCT 1722- R) but six in the other (Mezzalira, DGM uncataloged, figured in Romano et al., 2009: figs. 2B, 3C); the carapace of the type specimen is incomplete. Similarly in Lapparentemys , some have six neurals (WUS 2160) and some have seven (RM 20.5155, MHNC 6904, MNHNP VIL 1). Cambaremys has been reconstructed with seven neurals and Peirópolis A with six. Reconstruction of the former seems certain; the latter is more tentative. Too much of the carapace of Peirópolis A remains unknown to reconstruct the neural series with certainty.

Bauruemys appears to be the only taxon in this study in which all specimens have a four-sided second neural. In other taxa a four-sided first neural is the dominant condition. However, among the three neural series that we consider to represent Peirópolis A, the four-sided neural is number one in two and number two in the other. In Lapparentemys the four-sided first neural is present in the type and RM 20.5155, but in WUS 2160 the contact between the first and second neurals is asymmetrical with the first neural in contact with the second right costal. Thus, among the taxa of interest here, the neurals offer no concrete characters that help to determine relationships.

The neural series is complete to the suprapygal and excludes midline contact of posterior costals in Platychelys , Euraxemys , Cearachelys , some species of Pelusios and some specimens of Araripemys barretoi . The degree of variability of this character within certain genera makes it of questionable phylogenetic value.

Peripherals: There is variation in the degree of guttering at the pleuro-marginal sulcus on the bridge peripheral bones. Distinct guttering is present along the bridge in B. elegans . Among the disarticulated material from Peirópolis there are bridge peripherals of two distinct types; some are guttered like those of B. elegans , but others are completely smooth and without gutters. Guttering of the bridge peripherals is present on peripherals six and seven on the left bridge of the large white shell that is the best specimen of Peirópolis A. Thus, we consider the bridge peripherals of Peirópolis A to be guttered, those of Peirópolis B to be unguttered. In Cambaremys , the pleuro-marginal sulcus is described as being superimposed on the costo-peripheral sutures ( França and Langer, 2005), which would suggest that guttering is absent. Shallow guttering is also present in the bridge peripherals of Lapparentemys in WUS 2160 to the extent seen in B. elegans (specimen 1969-3). In Lapparentemys these costo-marginal sulci are clear but not as deeply incised as the lateral sulci of the vertebral scales. Bridge peripherals were not preserved with the type of R. wanderleyi , but in the two referred specimens, guttering is absent on the bridge peripherals.

Variation is present in the peripheral contacts with the bridge peripherals. In B. elegans , contact of the third peripheral with the axillary buttress is clear in the type and 1969-1. Contact of the eighth peripheral to the inguinal buttress is clear in 1969-4, showing that peripherals 3–8 contribute to the bridge. These contacts can also be seen in two specimens of Lapparentemys and one complete and one reconstructed specimen of ‘‘Peirópolis 321,’’ as well as the Mezzalira specimen of R. wanderleyi . Cambaremys differs from these forms in having the axillary buttress in contact with peripheral 2 ( França and Langer, 2005).

Peirópolis A appears to possess an uncommon set of contacts of the posterior peripherals. At least one point of evidence that the reconstructed carapace and good plastron specimen are both referable to Peirópolis A comes from an unusual arrangement of the posterior peripherals and costals. In most podocnemidids each posterior peripherals contacts two adjacent costals. In both specimens of Peirópolis A, costal 8 is very wide and contacts peripherals 9, 10, and 11 (fig. 91). Costal 7 is correspondingly narrow, so peripheral 9 contacts costals 6, 7, and 8. This is clear in the associated set of posterior elements used in the reconstruction of Peirópolis A and also on the interior surface of the carapace of the specimen with the complete plastron, DGM MCT 1499-R (fig. 91). The same arrangement has also been observed in the three specimens of Podocnemis , the extinct ‘‘ P. ’’ negrii (Caravalho et al., 2002) and in two recent species P. expansa (PCHP 4711) and P. unifilis (P.A.M. unnumbered).

Pygal region: Most Pelomedusoides have a single suprapygal, which contacts the pygal posteriorly, the 11th peripherals posteriolaterally, and the eighth costals anteriorly. There is some variation in the contacts of the eighth costals and adjacent peripherals in Peirópolis A. This is treated above under peripherals.

SCALES OF THE CARAPACE

Vertebrals: The first vertebral scale in Bauruemys , like those of nearly all pelomedusoides, is very wide and thus excludes contact of the first pleural scale with the nuchal bone. The portions of the first vertebral scale that are preserved in Euraxemys suggest that this scale was very wide and prevented contact of the first pleural scales to the nuchal bone, however, this remains uncertain. This contact appears to be absent among all podocnemidids (although it may be present in juvenile Peltocephalus [P. Pritchard collection]). Among bothremydids, it is absent from all known nuchal bones, except for those of two genera within the Bothremydini for which this region is available, Chedighaii and Araiochleys ( Gaffney et al., 2006). The derived condition is also seen in some species of Pelusios .

The shape of vertebral scales varies among these taxa. In Bauruemys vertebral scales 2–4 are hexagonal in shape with lateral apexes extending between pairs of pleural scales. Most other podocnemidids are similar, being hexagonal to some degree. However, in most Lapparentemys specimens the vertebral scutes have nearly parallel lateral sides (including the type). Only in the referred specimen, MHNC 6904, are they more hexagonal suggesting that this specimen may in fact belong to a different taxon. Recognition of parallel, deep, lateral sulci of the vertebral scales as a character of Lapparentemys supports referral of AMNH 14444 to this taxon. The piece of carapace collected with this important skull shows a nearly straight, deep furrow, parallel to the midline that is formed by the vertebral-pleural sulcus.

Pleuromarginal sulci: There is some variation in the location of the pleuro-marginal sulcus among podocnemidids. In Bauruemys it is generally located about one-third of the distance across the peripheral bones. In Lapparentemys and also in Peirópolis A these sulci lie near the middle of the peripherals. In Cambaremys and Podocnemis this sulcus is nearly coincident with the costo-peripheral suture particularly at the bridge. On the most anterior and posterior peripherals it extends slightly more distally onto the peripherals.

BONES OF CARAPACE IN INTERNAL VIEW

Axillary buttress: The axillary buttress in the six taxa of interest shows useful variation in the shape, medial extent and position relative to the suture of the first and second costal bone. In some forms there is contribution of the second costal to the axillary buttress.

In Cambaremys and Peirópolis A, the suture for the axillary process of the hyoplastron is nearly uniform in width (fig. 93A; França, 2004: fig. 17; França and Langer, 2005: fig. 5). The suture itself cannot be seen in any of the specimens we consider to represent Lapparentemys , but in RM 20.5155 the right hyoplastral process is articulated in place in the carapace and appears to be of equal thickness throughout its contact with the overlying costal and thus seems to represent the same morphology as in Cambaremys and Peirópolis A. In Peirópolis B, the suture for the dorsal process of the hyoplastron is wide in the middle part of the first costal but becomes significantly narrowed to a single ridge laterally (fig. 93B). This lateral narrowing of the suture can also be seen in R. wanderleyi (MCT 1722-R and the Mezzalira specimen). Bauruemys elegans is more difficult to assess, but in the right side of the type, the suture for the hyoplastron on the first costal appears to be reduced to a ridge laterally. Thus, B. elegans , R. wanderleyi , and Peirópolis B appear to share an alternate morphology of the buttress suture.

In the type of Roxochelys wanderleyi ( Price, 1953: fig. 3), the medial end of the axillary buttress can be seen to extend medially very close to the midline. It reaches medial to the level of the middle of the first peripheral. A very similar morphology is present in the two specimens we refer to R. wanderleyi and in Peirópolis B. In B. elegans (holotype and 4348V), Lapparentemys (RM 20.5155), Cambaremys , and in Peirópolis A, the hyoplastral contribution to the axillary buttress terminates more laterally. It reaches medially to about the middle of the first costal and to the level of the middle of the second peripheral. Medial to the hyoplastral process in all of these taxa, the buttress is composed entirely of the first costal. There is a strong continuous ridge between the hyoplastron buttress and the rib heads that extend to the first and second vertebrae.

Lapparent de Broin and Werner (1998) have noted that in some pelomedusoides the second rib head and axillary buttress are very posteriorly located on the first costal. However, in B. elegans, Peirópolis B , and R. wanderleyi (best seen in MCT 1722-R), the axillary buttress is supported by thickening of the second costal bone. In these taxa the second costal is thickened, sometimes massively, in the middle of its anterior margin (fig. 93F, G).

In these forms, with the axillary buttress posteriorly located on the first costal the second costal has a slightly curved anterior margin (in ventral view) to accommodate the thickened and posteriorly expanded first costal. Furthermore, the second costal is thickened along its anterior edge, at the point where it is adjacent to the axillary buttress. This can be seen in B. elegans (MN 4349-V) and an unnumbered fragment in the DNPM. All three of the specimens that we refer to R. wanderleyi show this morphology, as do several first costals of Peirópolis B. This participation of the second costal in the axillary buttress is most developed in R. wanderleyi, MCT 1722 -R, in which a short process of the second costal extends ventrally posterior to the thickening of the first costal where the latter articulates with the dorsalmost portion of the hyoplastron. In Peirópolis A, Cambaremys , and Lapparentemys , this contribution to the axillary buttress by the second costal is not present (fig. 93E).

Inguinal buttress: In all the taxa described here, the inguinal buttress is formed by a dorsal process of the hypoplastron, which contacts the fifth costal along a suture that is located in the middle of the element. This suture varies in width and medial extent. In Peirópolis A, and Cambaremys , this sutural area is wide, 40%–50% of the costal width and limited to the most lateral one-quarter of the costal (fig. 93C). In B. elegans (type) and in Lapparentemys (RM 20.5155), the inguinal buttress is also very laterally placed. In Peirópolis B, it is narrower and extends further medially extending medially for about one-third of the costal (fig. 93D). R. wanderleyi (best seen in Mezzalira specimen, figured in Romano et al., 2009: figs. 2B, 3C) is similar to Peirópolis B.

Iliac suture: There is a difference in the sutural attachment of the ilium to the overlying costals seven and eight among the Peirópolis and other species, with two iliac attachment morphologies present (figs. 96, 97). In one, the iliac scar forms a rough oval on costals seven and eight, with the seventh costal scar area being an anterolaterally protruding convexity (figs. 96A, 97A). In the other, the anterior margin of the iliac scar on the seventh costal is an anterolaterally facing concavity. In both the area and shape of the eighth costal, attachment is roughly similar; it is the shape of the seventh costal sutural surface that differs. Naturally, this difference can be seen in the ilium itself (fig. 96) as well as the carapace (fig. 97). In Peirópolis A the iliac scar is well preserved in two sets of associated coastal bones, seven and eight, one from the left side and one from the right. The scar in these costals protrudes anteriorly (fig. 97D). The medial limit of the scar is defined by a thickened area of bone, mostly on costal eight but extending slightly onto costal seven. This raised block of bone is the site where the rib heads of the sacral vertebrae arise to extend medially to meet their respective vertebral centra. These are thoracic vertebrae 9, 10, and 11, as in a specimen of Podocnemis erythrocephala (UF 57921).

In Peirópolis B the iliac scar extends onto the seventh costal both proximally and distally (fig. 98B) suggesting an anteriorly concave ilium (fig. 97B) as in Podocnemis . It is not convex anteriorly, as in Peirópolis A (figs. 96A, 97D). The difference in these scars is a result of the shape of the dorsal sutural surface for the ilium. In Peirópolis A it is convex anteriorly, in Peirópolis B; it must be concave anteriorly, as it is in Podocnemis expansa (fig. 96B). In the latter, the suture for the ilium crosses from the eighth costal onto the seventh at two sites, one medially and one laterally.

An anteriorly concave scar, like that of Peirópolis B can also be seen in Cambaremys ( França, 2004: fig. 19) and in Lapparentemys (RM 20.5155). In the latter specimen, the dorsal half of the ilium is preserved in articulation with the carapace and can be seen to have an anteriorly concave cross section dorsally. Both the anteriorly concave anterior margin of the dorsal suture of the ilium and the double sutural contact to the seventh costal are on both sides of the Mezzalira specimen of R. wanderleyi . In MCT 1722-R both the medial and lateral contacts to the seventh costal are apparent, but the anteriorly concave shape of the suture has been lost due to crushing of the ilium.

The condition of this character is hard to establish in the material of B. elegans . Nearly all specimens available to us have the plastron articulated to the carapace and the area of the iliac suture has not been prepared. Only the type is fully prepared in this region, but it is somewhat difficult to interpret. The distal end of the ilium on the left side is preserved in place against the carapace and covers over the sutural area. On the right side, although the ilium is also present, both the anterior and posterior sutures for the seventh costal can be seen for much of the length of this element, but the lateralmost part is missing. The centerline of the seventh costal appears to be thickened, but there is no clear sutural area for the ilium either medially or laterally and the ilium is positioned entirely in contact with the eighth costal. Thus, there is no evidence of an anteriorly concave dorsal suture in the ilium of B. elegans .

Peripherals: In all but one of the taxa under study, the posterior peripherals are like those of most turtles; there is no internal guttering along the proximal scute sulcus where the marginal scales meet the skin. Peirópolis A appears to possess a unique condition of the posterior peripherals. In this form there is an internal gutter on the ventral surface of peripherals 10, 11, and the pygal, that is deepest on peripheral 11 (fig. 92). This is not seen in any of the other fossil podocnemidids studied and appears to be an autapomorphy for Peirópolis A.

PLASTRON

Bones of plastron: In the taxa of interest here, the number and arrangement of the plastron elements is relatively uniform as it is among most members of the Pelomedusoides ( Araripemys and Pelusios being notable exceptions). Paired epiplastra and paired hyoplastra enclose a midline entoplastron. Laterally placed mesoplastra are present in all. The paired hyo- and hypoplastra make up the majority of the plastron, and paired xiphiplastra are present as in all turtles. The only variation among these elements is in the relative size of some elements, the scales that cover them (see below), the development of buttresses and the contact of the buttresses to the overlying carapace (described above).

In the Peirópolis material in the DNPM there are two epiplastral morphologies. Most epiplastra in the Peirópolis collections, including two pairs that can be associated with Peirópolis A, have a significant rounded lip along the anterior edge. There is a distinct groove just inside the anterior margin that marks the scale overlap by the intergular and gular scales in this form. In a single additional Peirópolis epiplastron, this rolled lip is absent and instead the epiplastron has an angular margin as is seen in Bauruemys elegans . This epiplastron we assign to Peirópolis B. There are additional differences between these epiplastral morphologies in the suture with the hyoplastron. In Peirópolis A, the posterior margin at the epiplastron-hyoplastron suture is straight but has a small medial notch for an anterior projection of hyoplastron. In Peirópolis B, instead of a notch there is a significant posteromedial projection extending posteriorly well beyond the lateral part of the suture. This latter morphology is also seen in the type and other specimens of B. elegans , and in all three specimens of R. wanderleyi . This area is not visible in the type of Lapparentemys , in WUS 2160, and is too smashed in RM 20.5155. In MNHN VIL3, there is a posterior projection, but it is very weak, and this specimen shows the distinctively thickened and rounded anterior margin seen in Peirópolis A. In the well-preserved shells from Tiupampa ( Broin, 1991), there are also two epiplastral morphologies. The epiplastron is not known in Cambaremys .

Rathke’s gland pores: Much of this material is too poorly preserved to establish the presence or absence of Rathke’s gland pores. However, in Peirópolis A, DGM 321 (a specimen with an entire plastron), there is a prominent inguinal pore between the medial edge of peripheral eight and the lateral part of the inguinal buttress. The axillary region of this specimen is also very well preserved and it is clear that there are no axillary Rathke’s gland pores, as are present in Podocnemis .

Scales of the plastron: Although the shell of podocnemidids in general has a highly conserved morphology, one aspect that is variable is the contacts of the scales that cover the anterior plastral lobe as far posterior as the mesoplastra. For these scales we use the traditional terminology of Zangerl (1969) rather than that proposed by Hutchison and Bramble (1981), which differs only in calling the intergular of Zangerl the gular, and the gular of Zangerl the extragular. Entire anterior plastral lobes are not known for Peirópolis B or Cambaremys .

In Bauruemys, Peirópolis A , Roxochelys wanderleyi , and Lapparentemys , an unpaired intergular covers the medial limits of both epiplastra and extends onto the anterior portion of the entoplastron. These contacts are consistent, but there is variation in the width of this scale among the taxa of interest. In the type of R. wanderleyi ( Price, 1953: fig. 4) and the two referred specimens, this scale is wider than in the other taxa in which this feature is known. In all three specimens, this scale is wider than the adjacent gular. In Bauruemys, Peirópolis A , and Lapparentemys , as in Podocnemis , the intergular width is equal to or smaller than the width of the adjacent gulars. For our purposes, this character is a useful autapomorphy for R. wanderleyi but does not appear to be phylogenetically informative.

In the four taxa for which we have plastra, the gulars are relatively small and usually restricted to the epiplastron. In two forms ( Lapparentemys and some Bauruemys elegans ) it reaches the entoplastron, but in others (Peirópolis A and Roxochelys ) it does not. Bauruemys shows intraspecific variation with some individuals having gular-entoplastron contact and others not. One specimen, DNPM uncataloged 1969-1, has this contact on the right side but not on the left. In most B. elegans , this contact is present.

The humeral scales in Bauruemys elegans cover most of the epiplastron, the anterior part of the entoplastron, and a very small anterolateral part of the hyoplastron. In most B. elegans the humero-pectoral sulcus extends anterolaterally from the midline across the entoplastron and onto the epiplastron and then near the edge of the plastron turns more laterally to cross the anteriormost part of the hyoplastron (fig. 88). The only exception to this is in DNPM uncataloged 1969- 4 in which this suture is more sinuous and crosses onto the epiplastron in two places. The humeral scales meet on the midline in all four of the taxa for which we have plastra, but the length of the contact varies. In B. elegans and Lapparentemys this midline contact is very short, about J or less of the length of the entoplastron. In Roxochelys wanderleyi and Peirópolis A it is longer, extending K or more of the length of the entoplastron. In both of the latter forms this results from more posteriorly place pectoral scales.

The pectoral scale in B. elegans covers the anterior part of the hyoplastron and extends anteriorly to cover the posterior half of the entoplastron and a small portion of the epiplastron. Posteriorly, it does not contact the mesoplastron itself but does reach the suture for this bone in some specimens. Laterally it contacts marginal scales four and five. There are no inframarginal scales present. In two of three Lapparentemys plastra for which this area is known, contacts of the pectoral scale are identical to Bauruemys . In the third (WUS 2160) the pectoral crosses the anterolateral corner of the mesoplastron. Contacts of the pectoral scales in the Mezzalira specimen of R. wanderleyi (the only one with complete bridges) are like those of Bauruemys in having no mesoplastron contact, but like those of Peirópolis A (both the complete plastron and DGM 321) in not reaching the epiplastra anteriorly. In the complete plastron of Peirópolis A, the pectoral scales cross the anterior part of the mesoplastron clearly on both sides. Thus, mesoplastral contacts of the pectoral scales are similar in Bauruemys and Lapparentemys , and they differ from the condition seen in Peirópolis A in which they are more posteriorly located.

The presence of the pectoral-abdominal sulcus crossing the mesoplastron appears to be the primitive condition for pelomedusoides. It is clearly seen in Platychelys , Euraxemys , Pelomedusa , Cearachelys , ‘‘ Galianemys , ’’ Chedighaii , and Rosasia . In other taxa this sulcus may lie anterior to the mesoplastron entirely on the hyoplastron as in all podocnemidids, and among the Bothremydidae in Kurmademys, Polysternon , and Araiochelys ; or it may be coincident with the mesoplastron-hyoplastron suture as in all Pelusios .

In Araripemys the anterior edge of the pectoral scale crosses the anterior lobe of the plastron well posterior to the entoplastron and the epiplastra. This is also the case in Pelomedusa and Platychelys . In Euraxemys , Pelusios , Cearachelys , ‘‘ Galianemys , ’’ and Rosasia , this sulcus crosses the midline of the plastron at or near the suture of the entoplastron with the hyoplastra. In podocnemidids and all known representatives of the bothremydid tribes Taphrosphyini and Bothremydini (other than Rosasia ), the pectoral sulcus crosses a significant part of the entoplastron.

The abdominal scale in Bauruemys elegans is the largest scale in the plastron covering the posterior half of the hyoplastron, the anterior half of the hypoplastron and most of the mesoplastron. It meets marginal scales six and seven laterally, often in a sinuous suture as is also seen in the type of Lapparentemys . There is no evidence of an inguinal scale. In Peirópolis A the abdominal scale contacts marginal scales six to eight laterally rather than just marginals six and seven. Furthermore, the posterior margin of the scale is not transverse with respect to the midline but rather angles anteriorly such that it reaches the midline just posterior to the hyohypoplastral suture (fig. 91B). Contacts of the abdominal scale in Lapparentemys are the same as in Peirópolis A, but the abdominofemoral sulcus is placed further posterior and the midline sulcus more sinuous.

The femoral scales in Bauruemys cover the posterior half of the hypoplastron and the anterior half of the xiphiplastron. Anal scales cover the posterior half of the xiphiplastron in all taxa of interest (Peirópolis B is unknown). The only notable variation in these scales among the taxa of interest is the sinuous midline suture seen in the type of Lapparentemys vilavilensis .

BONES OF THE PLASTRON IN INTERNAL VIEW

If our assumptions are correct and Peirópolis B has straight margins to the anal notch, and that xiphiplastra with a rounded anal notch represent Peirópolis A, then we have material that can be used to describe sutural areas for the pubis and ischium in Peirópolis A. The pubic suture is located midway between the midline and lateral edge. It is oblong, about twice as long as wide, and is oriented at an angle of about 30 ° from the midline. The ischial suture is more transverse and covers much of the posteriormost part of the xiphiplastron, although it is well offset from the margin of the bone. The sutural area extends laterally from the midline about two-thirds of the way across the xiphiplastron. The anterior border of the sutural area is straight, the posterior border curves posteriorly, so that the sutural area is about twice as wide laterally as it is medially.

In the Mezzalira specimen of R. wanderleyi the internal surface of the xiphiplastron has the distal end of the pubis and ischium still in place. The pubic suture is oval and not as elongate as in Peirópolis A. The ischial suture is narrower than in Peirópolis A. It extends from the midline laterally at a slight angle to transverse. It does not expand laterally as much in this specimen as in Peirópolis A. MCT 1722-R is similar, but the sutures are slightly more robust.

In the type of B. elegans the dorsal surface of the xiphiplastron is fully prepared. The pubis is articulated on the left side but not on the right. The sutural area is like that of Peirópolis A, twice as long as wide but at an angle of about 45 °. Both ischia are in place and expand both laterally and medially, although the suture itself may be more expanded laterally. This morphology of the ischium differs significantly from that in Podocnemis expansa . In the latter, the ischia narrow markedly towards the midline and do not make sutural contact with the xiphiplastra, so that the ischial sutural area on the xiphiplastron does not approach the midline.

In Cambaremys , França (2004: fig. 26) has indicated a pubic suture similar to those described above. However, the ischial suture is unusual; it appears to be more L-shaped with a long lateral portion and a narrow section crossing to the midline. In Lapparentemys the dorsal surface of the xiphiplastron is visible only in RM 20.5155. The surface of the bone is badly fractured, but it appears that the pubic sutures are larger (longer) than in the other South American forms and oriented at a higher angle to the midline. The ischial suture is similar to Peirópolis A (DGM 321) in that it expands from the midline laterally into a triangular area that is more than twice as wide laterally as medially. Although there is some variation in the sutures for the pubis and ischium, they do not appear to be distinctive enough to be used as scored characters.

GIRDLES AND LIMBS

We can associate disarticulated nonshell postcrania from Peirópolis only on the basis of size. Thus, the larger scapula, possibly representing Peirópolis A, has a dorsal process without a dorsal keel. Peirópolis B has a scapula with a keel present on the dorsal process. The type of B. elegans includes both scapulas (left one is complete), left femur, and left humerus; as in Podocnemis expansa , the femur is longer than the humerus. In the scapula there is no evidence of expansion or ridges on the dorsal process. However, the acromion process is somewhat flattened as is it in Peirópolis A. In the femur of the type the head extends dorsal to the trochanters to the same degree as in P. expansa . In the humerus, the head is very slightly lower than in P. expansa .

CERVICAL VERTEBRAE

The most complete specimen of Peirópolis A, DGM 321 has an associated atlas neural arch and centrum and an associated plaster bed for a cervical series. A complete cervical series found in the Peirópolis material is the correct size for DGM 321, but doesn’t seem to fit the plaster bed associated in the collection with DGM 321. However, it has the same preservation as the atlas with DGM 321 and consists of cervicals 2–8. In this cervical series number two is biconvex, 3–8 are procoelus. All are generally very similar to Podocnemis expansa to which they are compared here. Comparisons are also made to Erymnochelys madagascariensis (DGM 279 RR). Cervical 2 has a saddle-shaped condyle. The long ventral keel is longer than that of P. expansa . Transverse processes are long in both the fossil and P. expansa . Cervical 3 has a distinctly saddle shaped condyle as in Podocnemis . In Erymnochelys the condyle is very wide and dorsoventrally flattened. Transverse processes are like those in Podocnemis , strong and laterally directed. In Erymnochelys the processes project slightly more ventrally. In Podocnemis the postzygapophyses join to make a single, continuous articular surface. The fossil is more like Erymnochelys in this regard. Only the right postzygapophysis is preserved, but it is a rounded structure, separate from that of the other side. The keel in the fossil is much deeper than in either recent cervical.

Cervical 4 is saddle-shaped posteriorly, much like P. expansa . Transverse processes extend laterally as in P. expansa , not slightly ventrally as in Erymnochelys . Postzygapophyses are united in P. expansa and the fossil, but not in Erymnochelys . The combined postzygapophyseal surface is constricted medially in the fossil but not in P. expansa . Dorsally, the lateral processes of the fossil are connected to prezygapophyses by a broad area of bone making a somewhat winglike structure. The keel is deepest in the fossil. Cervical 5 is saddle shaped like Podocnemis . Transverse processes flare slightly upward and are connected to prezygapophyses by a broad expanse of bone. Postzygapophyses in the fossil are barely connected to one another while in P. expansa they are strongly connected as in cervicals 3 and 4. Cervical 6 also has a saddle-shaped condyle and broad plates of bone from transverse processes to prezygapophyses. Erymnochelys differs from the fossil and P. expansa in having the neural spine extend significantly above the postzygapophyses and in having a laterally expand- ed, convex condyle. Postzygapophyses are not connected in any of the three species. Cervical 7 looks most alike in these three species. The condyle is more balllike in Podocnemis and Erymnochelys . Postzygapophyses are separate in all three and the neural spine extends dorsal to them. Cervical 8 is like 7 in having a rounded condyle and tall neural arch. The transverse process of Podocnemis extends slightly more ventrally than in the fossil.

There are not enough available cervicals for an extensive comparative study among the Podocnemididae at present. These cervicals were found in the same locality as the type specimens of Pricemys and Peiropemys , but they are most consistent with the larger size of Pricemys and Peirópolis A.

PHYLOGENETIC ANALYSIS PREVIOUS WORK

The early recognition of the enlarged carotid opening ( Rütimeyer, 1873; Seibenrock, 1897) in ‘‘ Podocnemis ’’ (sensu lato) might be considered the starting point for recognizing a family level taxon as a monophyletic group, although clearly the literature grouped the living species of ‘‘ Podocnemis ’’ earlier in the 19th century ( Pritchard and Trebbau, 1984). The early recognition of three genera in the group ( Erymnochelys Baur, 1888 ; Peltocephalus Dumeril and Bibron, 1835 ; as well as the original Podocnemis Wagler, 1830 ) effectively ended with Boulenger’s (1889) influential catalog that synonymized the three genera. Later, some authors, including Williams (1954c), Frair et al. (1978), and Gaffney (1979), resurrected these genera to reflect the degree of diversity in the group, and this has become current practice.

Dacqué (1912) and later Zangerl (1948) proposed the first overt phylogenetic hypothesis for what we would now call the Podocnemididae . Although not using these generic names, this was essentially resolved as ( Erymnochelys ( Podocnemis , Peltocephalus )), the idea that there was an ‘‘African’’ lineage and a separate ‘‘South American’’ lineage. Although this resolution remained popular, Williams (1954c) questioned this hypothesis and used fossils to argue for what he called the ‘‘ Dacquemys-Erymnochelys-Peltocephalus series,’’ what could be interpreted as ( Podocnemis ( Erymnochelys , Peltocephalus )). This is also the hypothesis favored by França and Langer (2006), Meylan et al. (2009), Cadena et al. (2010) and by us. In arguing for his hypothesis, Williams (1954c: 6) states: ‘‘In the living species Peltocephalus dumeriliana … the skull has definite, strong similarity to that of Erymnochelys or of Dacquemys and thus also to the Moghara skull [here named Mogharemys ].’’ Later (ibid.: 7) ‘‘I differ with Dacque´, Zangerl and others in that I separate from the South American group the species dumeriliana … and regard the few resemblances of the latter species to the other South American forms—similarity in gular pattern and in cervical articulations—as convergent only.’’

Nonetheless, it was the Dacqué-Zangerl hypothesis that remained popular and was adopted by Gaffney (1988) and Gaffney and Meylan (1988). The first formalization of a podocnemidid phylogenetic hypothesis by naming the subfamilies Erymnochelyinae and Podocnemidinae was Smith and James (1958). They examined cloacal bursae in turtles and made the following conclusions (1958: 95): ‘‘African pelomedusids lack bursae; South American and Madagascar representatives possess them. This distinction, coupled with the presence of saddle cervical joints in the South American group, a deep temporal notch in the African group, and their probable course of isolation, is regarded as significant at the subfamily level; the South American subfamily is designated PODOC- NEMINAE (new), the African subfamily the PELOMEDUSINAE of Williams, the Madagascar subfamily the ERYMNOCHELY- DINAE (new).’’ This is the same resolution proposed by Broin (1988 [1989]), and it is clearly the ( Erymnochelys ( Podocnemis , Peltocephalus )) hypothesis. These two subfamilies of the Podocnemididae were elaborated and diagnosed by Broin (1988 [1989]): the subfamily Podocnemidinae and the subfamily Erymnochelinae (also see discussion in Systematics section). Although used in a number of additional papers ( Broin, 1991; Lapparent de Broin, 2000a, 2001, 2003a, 2003b), the most complete expression of the hypothesis is Lapparent de Broin (2000b) as follows: ‘‘Subfamily Podocnemidinae’’ (sensu Broin, 1991)— Podocnemis , Peltocephalus , Bauruemys , aff. Roxochelys vilavilensis [here named the new genus Lapparentemys ], Stupendemys . The character that seems to be dominant in Lapparent de Broin (2000b, and other papers) for this taxon is the saddle-shaped cervicals (see also Character Descriptions for a different interpretation), but, as in the case of Smith and James (1958), biogeography is also used as a dominant phylogenetic character. There are a series of shell characters as well, but the distributions are not consistent. Apparently, the fact that Bauruemys lacks the cervical character was not yet known and the author used various shell features and geography to place this species in this group. In our analysis, this subfamily is simply a paraphyletic group of South American podocnemidids.

‘‘Subfamily Erymnochelinae’’ (sensu Broin, 1991)— Erymnochelys , Neochelys , Stereogenys , Shweboemys , Dacquemys , Carteremys . This group is essentially the present authors’ magnatribe Erymnochelydand. The reflection of the cladogram in the classification requires its change in category, otherwise we would have kept the original name; we agree with most of its content and principle characters. The primary character used for this group ( Lapparent de Broin, 2000b: 70) is the ‘‘Much eroded roof of the enlarged carotid canal, the prootic and quadrate being so much eroded that the floor of the canalis cavernosus is broken and this canal is anteriorly confluent with the part of ‘enlarged canal’ leading to the sulcus cavernosus (not known in Dacquemys , homoplastic but with a less eroded roof in the podocnemidine Peltocephalus ). The interesting recognition by Lapparent de Broin (2000b) that the character does actually occur in Peltocephalus , a member of the other subfamily, does not escape the author, but the geographic consideration is too great. There are now a few cervicals showing that members of Lapparent de Broin’s Erymnochelyinae did have saddle-shaped centra.

The advent of nonmorphological characters has produced results inconsistent with either of the two morphological Podocnemididae hypotheses developed over the past century. Frair et al. (1978) produced a resolution of: ( Erymnochelys ( Peltocephalus ( P. expansa , P. vogli , P. lewyana , P. unifilis , P. erythrocephala ( P. sextuberculata , )))) based on a serological analysis. The Frair (1980) serologic study emphasized chelids and only concluded that Peltocephalus and Erymnochelys were distinct from Podocnemis . The karyologic study of Rhodin (1978) also concluded that Peltocephalus and Erymnochelys were distinct from Podocnemis . When just considering the Recent genera, none of the molecular results reproduce the Dacqué (1912), Zangerl (1948), Gaffney (1988), Gaffney and Meylan (1988), and Lapparent de Broin (2000b) resolution of ( Erymnochelys ( Podocnemis , Peltocephalus )), or the Williams (1954c) and França and Langer (2006) hypothesis of ( Podocnemis ( Peltocephalus Erymnochelys )); rather they present the hypothesis advanced by Frair et al. (1978), the ( Peltocephalus ( Podocnemis Erymnochelys )) arrangement.

Molecular analyses of the Podocnemididae first only resolved generic level taxa ( Seddon et al., 1997; Georges et al., 1998; Noonan, 2000; Noonan and Chippendale, 2006) and presented the following resolution: ( Peltocephalus ( Podocnemis Erymnochelys )). A later molecular study ( Vargas-Ramírez et al., 2008) was the first to include all the Recent Podocnemididae species, and supported the earlier resolution of the genera. It is interesting that the molecular studies have agreed on an alternative that has not been expressed by any of the earlier morphologic analyses. So all three possible outcomes for resolving the three living genera are supported by some study.

It should be noted that the topology of the species resolution in the Vargas-Ramírez et al. (2008: fig. 3) molecular study has almost nothing in common with our resolution of species based on morphology. Although we feel that we have relatively strong support for the union of Peltocephalus and Erymnochelys , we do not have similar optimism about our resolution of the species within Podocnemis . The characters used are frequently subject to homoplasy in other turtle taxa, but, osteologically speaking, these Recent species are very similar to each other. Some are not even readily identifiable from skeletal parts alone, hence the low confidence in our resolution of the Podocnemis species.

CHARACTER DESCRIPTIONS

1. NA, nasal: (0) present; (1) absent.

Discussion: See Gaffney et al. (2006).

2. FR, orbital position: (0) facing laterally, anterolaterally; (1) facing dorsolaterally; (2) facing dorsally.

Discussion: The difficulty in using characters like this one is distinguishing discrete characters that are actually gradational, therefore making them even more subjective than usual. The problem is discussed in Gaffney et al. (2006: 577). See also Meylan et al. (2009: 10). For the restricted group of the Podocnemididae we felt that this was possible because there is not that much individual or specific variation among the living forms and a number of the fossils are definitive in their morphology. State 2 only occurs in Bauruemys . State 1 is primitive for the family.

3. FR, interorbital groove: (0) absent; (1) present.

Discussion: The narrow groove in the narrow interorbital bar formed by the dorsal surface of the prefrontal and frontal in all the living species of Podocnemis and the extinct Podocnemis bassleri ( Williams, 1956) is a good synapomorphy for this genus. Williams (1954a) used this character for Podocnemis .

4. FR, prefrontal and frontal in lateral view: (0) flat or slightly convex; (1) strongly convex dorsally.

Discussion: Bairdemys (except B. winklerae ) differs from the other tribe Stereogenyini in having a pronounced bulge in the skull roof between the orbits resulting in a markedly attenuated preorbital region. The other members of the tribe Stereogenyini have flat or less pronounced profiles. This character is approached outside the tribe Stereogenyini by Podocnemis ( Gaffney, 1979: fig. 134), and Bairdemys is still more extreme. See Gaffney et al. (2008) for further discussion. Although this character is rare in pleurodires, a convex interorbital profile occurs in sea turtles and batagurids. The simple morphology of the character makes homoplasy testing difficult.

5. PAR, quadratojugal-parietal contact: (0) absent; (1) short contact; (2) long contact.

Discussion: Gaffney et al. (2006: 578) discusses this character. See also Fuente (2003) and França and Langer (2006: 371).

6. PAR, parietal-pterygoid contact in septum orbitotemporale: (0) absent; (1) present and wider; (2) present and narrower.

Discussion: The septum orbitotemporale and its associated sulcus palatinopterygoideus are described and figured in Gaffney et al. (2006: 118–126, figs. 23–25). The wider parietal-pterygoid contact unites the tribe Stereogenyini with Neochelys .

7. PAR, temporal emargination: (0) moderate to absent; (1) extreme, as in Pelusios ; (2) shallow, cheek emargination extending posterodorsally to or above quadrate; (3) emargination absent due to expanded parietal/ supraoccipital.

Discussion: Many authors have used temporal emargination as a character. It is a gradational feature, hard to make objective, and highly subject to homoplasy. França and Langer (2006: 371) use criteria that are as good as any, but here we have taken a more extreme view. We distinguish only the relatively narrow postorbital roofing present in Pelomedusidae and Araripemys as one state and all others that are more roofed as a single state. Certainly such conditions as found in Bauruemys and Bairdemys sanchezi are more emarginate than Cordichelys and Dacquemys , but we have not tried to tease out the various conditions.

State 2 characterizes chelids that retain parietal-squamosal contact, except in Chelodina (contact lost due to extensive cheek emargination, not temporal emargination). State 3 seeks to unite the skull roofing morphology seen in Dacquemys (Gaffney et al., 2002) with an as yet undescribed specimen from the Lake Turkana Miocene, UCMP 42008 (see Systematics section). The character is a completely covered temporal roof caused by the large posterior part of the parietal with some lateral expansion of the supraoccipital.

8. PAR, interparietal scale: (0) absent; (1) equilateral triangle; (2) elongate triangle; (3) parallel sided; (4) broad posteriorly.

Discussion: Williams (1954a: 284) used the head scalation of podocnemidids and we have attempted to use these features. See also Meylan et al. (2009: 11). There is a great deal of homoplasy in the distribution of the states, however. State 1 appears to be primitive for the family Podocnemididae and state 3 appears to be a subtribe Stereogenyina synapomorphy. Turkanemys could be considered trapezoidal ( Wood, 2003), but it seems to us as very similar to the equilateral triangle condition.

9. JU, jugal-quadrate contact: (0) absent; (1) present.

Discussion: This is the classic character that unites Peltocephalus and Erymnochelys ( Baur, 1890; Williams, 1954c; Siebenrock, 1897, 1902) in contrast to Podocnemis among the living fauna. But it is still a nearly unique character when viewed from the context of the fossil record. Only the heavily roofed UCMP 42008 also has this character, and there is some morphologic reason to think that they are not homologous. The cheek of UCMP 42008 has a number of bone contacts that differ from Erymnochelys and Peltocephalus , and its sister taxon, Dacquemys , has no jugal-quadrate contact. So this does appear to be a synapomorphy for Peltocephalus + Erymnochelys , despite the fact that it is homoplastic in other turtle groups ( Gaffney, 1979).

This character may occur in an undescribed specimen, KNM-RU 18401 ( Witmer, 1990), from the Miocene of Rusinga Island, Kenya (see discussion above under Turkanemys ).

10. JU, jugal-parietal contact: (0) absent; (1) present.

Discussion: The jugal-parietal contact unites the recent species of Podocnemis (and P. bassleri ) due to the very small postorbital bone characteristic of this genus ( Ruckes, 1937; also in Gaffney, 1979). There is no homoplasy known.

11. JU, Cheek emargination: (0) slight; (1) reaches level of orbit; (2) reaches above level of orbit; (3) reaches above quadrate.

Discussion: Cheek emargination, like temporal emargination is basically a gradational character that needs to be broken up into states that are as objective as possible. Our interpretation of states is described and discussed in Meylan et al. (2009: 11). State 1 is primitive for podocnemidids in our analysis with state 2 diagnostic for the Peiropemys + Lapparentemys + Pricemys grouping.

Cheek emargination in the subtribe Stereogenyina is determinable only in Bairdemys , Stereogenys , and Cordichelys . In Bairdemys the emargination is most extensive and very similar in size and shape to Podocnemis . Stereogenys has a slight, nearly absent emargination with Cordichelys appearing to be intermediate, although it could be more extensive. Skull roof emargination is variable in turtles, and other podocnemidids, Peltocephalus and Erymnochelys , have closed cheeks, although the bones involved ( Gaffney, 1979) differ from the subtribe Stereogenyina .

12. SQ, ventral vertical flange: (0) absent; (1) present.

Discussion: This character seems to be an attachment area for the M. depressor mandibulae (see Gaffney et al., 2006: 582, for discussion in bothremydids). Small squamosal flanges occur variably in Podocnemis expansa , Peiropemys , Lapparentemys , and Bauruemys , but for this character we distinguish the distinct and deep flange that characterizes the species of Bairdemys (see also Gaffney et al., 2008).

13. PO, size: (0) equal to orbit; (1) smaller than orbit.

Discussion: The unusually small postorbital in Podocnemis has been described before ( Ruckes, 1937; also in Gaffney, 1979), but we have seen specimens of P. unifilis that have a more normal sized postorbital, a presumed reversal in our analysis. The character does not occur elsewhere.

14. PM, premaxillae reach apertura narium interna: (0) no; (1) yes.

Discussion: State 1 may include two states, depending on interpretation, as always. The subtribe Stereogenyina have a well-developed secondary palate with a midline cleft. The position of the apertura narium interna is somewhat arbitrary, as it could be considered to lie at the anterior end of the cleft or at the posterior end of the secondary palate, its position in other turtles with secondary palates, such as cheloniids. In this case, we have interpreted the midline cleft as constituting the apertura narium interna. We base this on the purely morphologic condition that the premaxillae have a free posterior edge in the subtribe Stereogenyina as they do in Neochelys , Peltocephalus , and Erymnochelys . This interpretation results in a grouping of the subtribe Stereogenyina with Neochelys , Peltocephalus , and Erymnochelys .

However, it is possible that the cleft was covered with soft tissue, not preserved, and the apertura narium interna would be more posterior. The more posterior position is also more likely than the thin cleft as a pathway for incoming air, if one were to use a more physiologic interpretation. If one were to consider the subtribe Stereogenyina condition as an alternate state, then it would provide another synapomorphy for the subtribe Stereogenyina and remove one uniting the three previously named genera. Take your pick.

15. PM, pinched snout: (0) absent; (1) concave outline near premaxilla-maxilla contact, snout not elongated; (2) concave outline posterior to premaxilla-maxilla contact, snout elongated.

Discussion: This character is relevant to the subtribe Stereogenyina . It was used by Gaffney and Wood (2002), criticized by Sánchez-Villagra and Winkler (2006), and defended (or further confused) by Gaffney et al. (2008). See the latter publication for further discussion and possibly more precise description. As presently interpreted, the pinched snout in all its magnificence occurs in the infratribe Stereogenyita (not known in Brontochelys ) and in Peltocephalus . Bairdemys winkleri has a unique elongated snout with a broadly concave outline that we do not interpret as the same state as in the infratribe Stereogenyita . However, this is a morphologically simple character, whose homology is difficult to test for, and it is found widely homoplastic throughout turtles. Other podocnemidids usually have straight snouts, so the pinched condition is presumably derived.

16. MX, medial expansion of triturating surface: (0) absent; (1) present, forming median maxillary ridge; (2) secondary palate with midline cleft.

Discussion: The palates of Bauruemys , Peiropemys , Lapparentemys , and Pricemys , as well as the outgroups to the Podocnemididae , Hamadachelys, Portezuelemys , and Brasilemys , all have the anterior part of the palate with a wide concavity on the midline, formed by the premaxillae and anterior part of the maxilla. This concavity is related to the relatively narrow anterior (and sometimes posterior as well) part of the triturating surface, formed mostly by the maxilla. In the remaining Podocnemididae , with the exclusion of the subtribe Stereogenyina , the anterior part of the maxilla is medially expanded, usually in the form of a variably developed ridge, the median maxillary ridge. The median maxillary ridge extends ‘‘the length of the maxilla along the middle of the triturating surface and then extends onto the premaxilla anterior and ventral to a well-defined foramen praepalatinum’’ ( Meylan et al., 2009: 13). The medial expansion of the maxilla constricts the midline concavity, reducing it to a short trough.

In the present analysis the medial expansion of the maxilla is an important character that defines the group consisting of Podocnemis + infrafamily Erymnochelyodda with a secondary modification in the subtribe Stereogenyina . The palate in the subtribe Stereogenyina is highly modified, but the presence of the medial expansion in the maxilla can be considered as the anterior part of the secondary palate. The medial expansion of the maxilla is relatively variable within cryptodires. Thus, the consistency of this character within the Podocnemididae is interesting and somewhat unexpected.

Character state 2, the complete secondary palate formed by maxilla and palatine, is a very distinctive feature of the subtribe Stereogenyina and does not occur outside that group. A number of turtles have evolved secondary palates; some like Osteopygis ( Gaffney, 1979) are more extensive than in the subtribe Stereogenyina . But all other chelonian secondary palates have the bones meeting in the midline, usually with contributions from the vomer, often with a vomerine pillar dorsally, separating the two choanal passages ( Gaffney, 1979). The subtribe Stereogenyina are unique among turtles in having a secondary palate with a median cleft and no contribution from the vomer, which is absent in the group. The posterior extent of the secondary palate varies in the subtribe Stereogenyina , but the width of the median cleft is relatively constant. There is no way to determine whether or not the cleft was filled in life with cartilage or soft tissue or whether it allowed the passage of incoming air. The morphologic consistency and the unique distribution of the median cleft support the homology of this character among the subtribe Stereogenyina .

17. MX, secondary palate long: (0) no, relatively short (palate length/skull length less than 0.6); (1) yes, relatively long (palate length/skull length more than 0.7).

Discussion: In Stereogenys the ratio of palate length to skull length varies among four specimens from 0.71 to 0.73 and in Shweboemys the ratio is 0.68. These are relatively long palates. The shorter palates represented by the other taxa range from 0.47 ( Brontochelys ) to 0.58 ( Cordichelys ). We interpret the shorter palates as primitive only because the outgroups lack secondary palates. The larger secondary palate of Stereogenys and Shweboemys is made up of a relatively larger palatine bone when compared with the other subtribe Stereogenyina .

18. MX, triturating surface convexity: (0) absent or shallow; (1) deep.

Discussion: Two species of Bairdemys , B. venezuelensis and B. hartsteini , differ from the other subtribe Stereogenyina in the large size of the palatal swellings. Latentemys and Cordichelys also have these swellings, but they are not as large or pronounced as in these species of Bairdemys .

19. MX, labial ridge: (0) high and narrow; (1) low and thick.

Discussion: Stereogenys , Shweboemys , Lemurchelys , and Brontochelys have relatively low and thick labial ridges in contrast to the other subtribe Stereogenyina .

20. MX, accessory ridges: (0) absent; (1) one or two.

Discussion: See Gaffney et al. (2006: 587) and Meylan et al. (2009: 13) for discussion and description. There is some homoplasy in this character in our analysis. It occurs in Podocnemis + infrafamily Erymnochelyodda , with the exception of Neochelys , and a reversal for the subtribe Stereogenyina . The absence of any ridges in the subtribe Stereogenyina is a synapomorphy of that group.

Williams (1956: 4) describes the ridges in Podocnemis expansa and Podocnemis bassleri : ‘‘(1) a short anterior ridge beginning on the premaxilla and extending a short distance onto the maxilla; (2) a ridge parallel to this beginning at the premaxillary suture and extending posteriorly almost to the end of the triturating surface; (3) a broad roughened area parachoanal in position, converging anteriorly towards the second ridge, not parallel to it. This third ridge or roughened area is very low and broad in the fossil, as it is occasionally in P. expansa . In the fossil there is no evidence of the faint ridge on the internal surface of the tomium which is regularly present in P. expansa .’’

21. MX, meet broadly on midline: (0) no; (1) yes.

Discussion: Descriptions of this character can be found in Gaffney et al. (2002: 10) and Meylan et al. (2009: 13). In the present analysis, the midline meeting of the maxillae is completely homoplastic, occurring independently four times, forming no groups.

22. VO, vomer: (0) present; (1) absent.

Discussion: The presence/absence of the vomer in the Podocnemididae requires multiple losses in our analysis. The vomer has been lost three times within the Pelomedusoides, once in the Pelomedusidae plus Araripemys (the magnafamily Pelomedusera of Gaffney et al., 2006), once within Podocnemis , and once within the Infrafamily Erymnochelyodda , as the vomer is absent in all of that clade except Caninemys ( Meylan et al., 2009: 13) . The very thin, barely recognizable, vomer in Podocnemis vogli could be interpreted as a neomorph as its morphology has very little in common with the vomer in the outgroups. It is identified on the basis of its position rather than its morphology.

23. PAL, medial edges of palatal cleft: (0) absent; (1) medial edges parallel; (2) medial edges curved.

Discussion: In Stereogenys , Lemurchelys , and Shweboemys the medial edges of the palatal cleft are parallel to each other. In the other genera of the subtribe Stereogenyina they are curved, convex toward the midline. In the absence of an outgroup with a secondary palate, it is impossible to determine which (if either) condition is primitive or derived. The parallel-edged condition is found in the same taxa that have the longest secondary palate and it is possible that these are functionally associated in some way. In any case, the sister taxon relations of Stereogenys , Lemurchelys , and Shweboemys are supported by other characters.

24. PAL, palatine extent in triturating surface: (0) narrow or absent; (1) moderate, but much less than extent of maxilla; (2) large, equal to or slightly less than extent of maxilla.

Discussion: This character is somewhat gradational, and we have separated out the most extreme condition, that of the secondary palate in the subtribe Stereogenyina , as a second state. State 1 is a synapomorphy for the Podocnemididae .

25. PAL, dorsal process of palatine contacts parietal in septum orbitotemporale: (0) no; (1) yes.

Discussion: The septum orbitotemporale and its associated sulcus palatinopterygoideus are described and figured in Gaffney et al. (2006: 118–126, figs. 23–25). Shweboemys has a large process of the palatine that rises dorsally in the septum orbitotemporale to form much of the septum. It contacts the parietal posterodorsally and is interposed between the jugal and the pterygoid laterally and the postorbital and the pterygoid medially.

In Stereogenys this palatine process can be seen in BMNH R.3191. In this specimen the palatine-parietal and palatine-pterygoid sutures show the posterior and dorsal extent of the palatine, but the anterior and anterodorsal contacts with the jugal and postorbital are obscured. None of the other Stereogenys specimens are well enough preserved to determine sutures in this area. In other subtribe Stereogenyina and other podocnemidids the palatine forms only the ventral edge of the fossa temporalis and the jugal and pterygoid meet to form the septum orbitotemporale.

26. PAL, dorsal process of palatine contacts frontal in septum orbitotemporale: (0) no; (1) yes.

Discussion: The dorsal palatine process is in the anterior margin of the septum orbitotemporale. The frontal sends a process ventrally to meet the dorsal process of the palatine about halfway up the height of the wall. The septum orbitotemporale and its associated sulcus palatinopterygoideus are described and figured in Gaffney et al. (2006: 118–126, figs. 23–25). This character is a synapomorphy for the tribe Stereogenyini , which is the subtribe Stereogenyina plus Mogharemys .

27. PAL, fossa orbitalis posterior pocket in septum orbitotemporale: (0) absent; (1) present.

Discussion: The subtribe Stereogenyina are characterized by a relatively thick septum orbitotemporale separating the fossa orbitalis from the fossa temporalis. The anterior surface of the septum orbitotemporale is an anteriorly facing concavity, presumably containing eyeball attachments or orbital glands. None of the other Pelomedusoides have this concavity. The concavity occurs in all the subtribe Stereogenyina that have this area exposed. However, Shweboemys and Bairdemys winkleri have matrix filling the two specimens representing this taxon, so the distribution of this character is not definite.

28. PAL, basisphenoid-palatine contact separates pterygoids: (0) no; (1) yes.

Discussion: In Shweboemys the basisphenoid separates the unusually small pterygoids to barely contact the palatines just on the midline. Some of both pterygoids and basisphenoid have been eroded off the surface, but the original positions of the sutures seem to have had a basisphenoid-palatine contact. Stereogenys also has basisphenoid-palatine contact. BMNH R.3191 is the only Stereogenys specimen clearly showing the basicranial sutures. In this skull the elongate basisphenoid completely separates the pterygoids, to a greater extent than in Shweboemys , and broadly contacts the palatines, partially separating them, also in contrast to Shweboemys . The basisphenoid-palatine contact does not occur in other pleurodires.

29. QU, antrum postoticum: (0) large; (1) smaller; (2) smallest and slitlike.

Discussion: In most pleurodires (and cryptodires) the antrum postoticum is a cone-shaped cavity formed by the squamosal and communicating with the cavum tympani of the quadrate. In Bairdemys (unknown in B. winkleri ), Lemurchelys , and Stereogenys , however, the antrum is collapsed to a small slit and contains almost no space. This is in contrast to the relatively open antrum seen in Latentemys and Cordichelys . Unfortunately, the region is missing in Shweboemys and Brontochelys . Caninemys also has a very small antrum, not quite the slit seen in the subtribe Stereogenyina , but we have coded it as state 2. In our analysis, state 2 appears three times independently.

30. QU, fossa precolumellaris: (0) very small to absent; (1) present but shallow; (2) deep and well defined.

Discussion: See discussion in Gaffney et al. (2006: 600). In our analysis, we have used three states to express this character. We do not order them, however, it is hard to see, transformationally speaking, how state 0 (small or absent) could reach state 2 (deep) without going through state 1 (shallow). The rather diverse distribution of this messy character has inclined us to make as few assumptions as possible. Consistent with earlier analyses (e.g., Fuente and Iturralde-Vincent, 2001; Fuente, 2003; Gaffney et al., 2006; Meylan et al., 2009) our analysis shows that the deep and well-defined state is primitive for the Eupleurodira at least, although we have scored the absent condition as primitive because it is absent in Proganochelys and other non-casichelydian turtles.

31. QU, eustachian tube separated by bone from fenestra postotica: (0) no; (1) yes.

Discussion: The fenestra postotica ( Gaffney, 1979: figs. 85–102) of turtles is very variable in the degree to which structures traversing it are delimited by bone. In Bairdemys and Latentemys there is a bony wall subdividing the fenestra postotica that is not found in any other turtles. The wall appears to separate the Eustachian tube from the lateral head vein (vena capitis lateralis).

32. QU, incisura columellae auris: (0) no posterior bony restrictions; (1) eustachian tube separated from stapes by bone or narrow fissure; (2) eustachian tube and stapes enclosed or nearly enclosed by bone.

Discussion: State 2 is a synapomorphy for the family Podocnemididae plus Hamadachelys , but is unfortunately unknown for Portezuelemys. Further discussion in Gaffney et al. (2006) and Meylan et al. (2009).

33. QU, quadrate-basioccipital contact: (0) absent; (1) present.

Discussion: See Gaffney et al. (2006) for discussion. This character (among others) unites the superfamily Podocnemidoidea , essentially the family Bothremydidae plus epifamily Podocnemidinura.

34. QU, medial quadrate process reaches braincase: (0) absent; (1) present.

Discussion: See Gaffney et al. (2006) for discussion.

35. PT, cavum pterygoidei: (0) absent; (1) partial; 2) complete.

Discussion: Rütimeyer (1873: 58–62) seems to have been the first to describe the enlarged carotid opening in Podocnemis , and we have been unable to find an earlier statement. Seibenrock (1897) described it in detail for Erymnochelys madagascariensis . Siebenrock’s description (1897: 301, 302; given here in translation from the German courtesy of B. Werscheck) is worth repeating at this point:

In Podocnemis [ Erymnochelys madagascariensis is the species Seibenrock, 1897, figured in pl. 6,

fig. 38, and is presumably the one he describes below] the canalis caroticus internus is unusually widened, and Rütimeyer very fittingly compared it to a bony funnel. Rütimeyer states that from an osteological point of view the reason for this shape cannot be explained, but goes on to say

‘that it could have something to do with the blood supply and/or discharge to and from the brain and the eyes,’ and that it was probably for the blood vessels of the carotids and jugulars. The latter is quite right, but even for this the canal does not have to be that wide. If you look at the canal closely you can see four holes that indicate either the start or the termination of four canals. At the medial wall, behind the otosphenoideum [5 prootic], we find the foramen for the lower branch of the nervus facialis, which unites with the nervus vidianus, which is moving forward in the canalis caroticus internus. Slightly more towards the front of the basisphenoideum we find the foramen, which leads the carotis interna towards the base of the brain and which terminates in the fossa hypophyseos in the cranial cavity. At the roof of the second canal we find the foramen for the anterior branch of the carotis externa, which is led through the recessus cavi tympani into the cranial cavity and from there to the eye. Finally the nervo vidiano comes through the floor of the pterygoideum through the foramen pro ramo nervo vidiano, enters into a canal and terminates at the upper surface of the pterygoideum behind the orbital cavity and medial to the foramen palatinum posterius. The canalis caroticus externus is the only canal that starts in the recessus cavi tympani, and above, it is between the otosphenoideum and the quadrate, together with the carotico-temporale, and they terminate into the wide canal. The canalis cavernosus, which normally terminates here in the turtles, is missing. Therefore, we can assume that in Podocnemis the vena jugularis interna and the nervus facialis move through the widened canalis caroticus internus. Judging from the size of the foramina for the carotids, one can see that they are not abnormally large, therefore, only the venous system seems to be much more developed than normally. However, the question now arises of why is it only in Podocnemis that the venous system is so much larger than the arterial system. Therefore, the width of the canal must have other reasons as well, but these can probably only be determined in a recently killed animal.

The ‘‘enlarged carotid canal’’ has been and is used by many previous authors as a character. The term ‘‘cavum pterygoidei’’ is a more formalized name for the ‘‘pterygoideus muscle chamber’’ or ‘‘enlarged carotid channel’’ of Gaffney (1979: fig. 86) and others. First formalized by Gaffney and Wood (2002: 20) as the cavum pterygoideus, it was changed for no reason to cavum pterygoidei in Gaffney et al. (2006: 603). State 2 is hypothesized as diagnostic for the family Podocnemididae , as it has been for over 100 years ( Rütimeyer, 1873; Siebenrock, 1897, 1902).

The cavum pterygoidei is a relatively large opening from the palate into the braincase located at the posterior end of the pterygoid, containing a subdivision of the pterygoideus muscle ( Schumacher, 1954, 1955a, 1955b, 1973). It is differentiated from the fossa pterygoidea found in bothremydids ( Gaffney et al., 2006) by having at least a partial covering ventrally and an anteromedial opening into the braincase. The two sister groups of the Podocnemididae , Brasilemys ( Lapparent de Broin, 2000b) and Hamadachelys (figs. 1–6; Tong and Buffetaut, 1996), have a cavum pterygoidei that is hidden anteromedially by the underlapping basisphenoid medially and the pterygoid laterally. In these taxa the cavum is not as deep as in all other Podocnemididae , but the cavum pterygoidei is interpreted here as homologous in Hamadachelys , Brasilemys , and Podocnemididae . The partial condition in Brasilemys , Hamadachelys , and Portezuelemys, is interpreted as additive to the complete state.

The cavum pterygoidei (for contents see Siebenrock, 1897; Albrecht, 1976; Gaffney, 1979; Schumacher, 1954, 1955a, 1955b, 1956, 1973), in its completely developed state, is formed by four bones: the basisphenoid anteriorly and medially, the pterygoid ventrally and laterally, the prootic posterodorsally, and the quadrate posterodorsolaterally. There are five foramina in the cavum pterygoidei in Pricemys and Peiropemys . Posteromedially along the length of the basisphenoid is the foramen nervi abducentis (1). More anteriorly, also within the basisphenoid, near the anterior limit of the cavum pterygoidei is the foramen posterius canalis carotici interni (2); just lateral to this is the foramen caroticum laterale (3). The pterygoid-basisphenoid suture is lateral to the foramen caroticum laterale, and lateral to that contact and within the pterygoid is the foramen nervi vidiani (4). More posteriorly in the cavum, in its dorsal surface, the prootic is exposed. The foramen nervi facialis (5) lies in the center of the prootic. The foramen nervi vidiani is frequently hard to find as it is usually very small.

These foramina can be probed in the braincase of Pricemys , which has had nearly all the matrix removed in the cavum cranii and both cava pterygoidei. Bauruemys MCT 1753 -R (figs. 11, 12) has the cavum pterygoidei exposed on both sides. Here the foramen posterius canalis carotici interni appears to be combined with the foramen caroticum laterale as one oval opening, possibly damaged by preparation. So this specimen has four foramina comparable in position to those in Pricemys and Peiropemys . Unfortunately, none of the Roxochelys specimens are as well preserved or as matrix-free as the Pricemys and Peiropemys , but the areas preserved agree with the Pricemys and Peiropemys morphology. The skull figured in Broin (1991: pl. 2, figs. 1–7, ‘‘Tiupampa’’), which is Lapparentemys vilavilensis and not the shell genus, Roxochelys , has nearly all of the cavum preserved on one side or the other, and it has the Pricemys and Peiropemys condition.

This character is used by Gaffney and Meylan (1988), Meylan (1996), Lapparent de Broin and Werner (1998), Lapparent de Broin (2000b), Fuente and Iturralde-Vinent (2001), Fuente (2003), and Gaffney et al. (2006), among others, and has been used since Rutimeyer (1873), Siebenrock (1897, 1902), and later Williams (1954a).

36. PT, anterior opening of cavum pterygoidei: (0) absent (cavum pterygoidei absent); (1) small opening; (2) moderate opening; (3) large opening, foramen cavernosum opens in roof of cavum pterygoidei.

Discussion: The anterior opening of the cavum pterygoidei has what we have identified as three states. In the most primitive condition the foramen caroticum laterale, foramen anterius canalis carotici interni, and foramen nervi abducentis are small foramina, but the cavum pterygoidei anterior wall is otherwise a solid wall of bone formed by the basisphenoid, pterygoid. and prootic. The canalis cavernosus does not communicate with the cavum in any way. In the morphologically intermediate condition, both the foramen posterius canalis carotici interni and the foramen anterius canalis carotici interni are much larger; the more posterior foramen nervi abducentis is very large and its anterior opening is part of the large foramen anterius canalis carotici interni. The wall formed by the pterygoid ventrally and the prootic dorsally is eroded away to some extent in this intermediate condition, making the canalis/sulcus cavernosus open to the more ventral cavum pterygoidei.

In the most extreme condition, state 3, this prootic/pterygoid wall is absent and there is no contact between them in the anterior region of the cavum pterygoidei. The canalis cavernosus is more posteriorly placed and is an opening in the prootic-quadrate suture in the dorsal surface of the cavum pterygoidei near the foramen nervi facialis. In this condition there is no demarcation between the cavum pterygoidei proper and the sulcus cavernosus; presumably the lateral head vein runs anteriorly with the other structures exiting the cavum pterygoidei. We interpret the large opening state as distinct from the character ‘‘foramen cavernosum opens in roof of cavum pterygoidei.’’ However, these characters may not be independent. Nonetheless, we have run the analysis with the latter character removed and the result is the same.

Described by Lapparent de Broin (2000b), the opening of the canalis cavernosus/foramen cavernosum into the roof of the cavum pterygoidei is a character that differentiates the living Podocnemis , which lack it, from Peltocphalus and Erymnochelys , which have it (see discussion above under Sysytematics for Erymnochelys ). However, Lapparent de Broin (2000b) prefers an alternate phylogenetic hypothesis, uniting Podocnemis and Peltocphalus, by arguing that the occurrence of this character in Peltocephalus is homoplastic. She states that there are some differences in position of the foramen cavernosum in Peltocephalus and Erymnochelys , and that Peltocephalus agrees with Podocnemis in geography and the possession of saddleshaped cervical centra, absent in Erymnochelys . In addition, Broin (1991: 513) explains that in Erymnochelys and Neochelys the ‘‘inner surface of the canalis caroticus is also much enlarged but differently’’ from the condition in Peltocephalus and Podocnemis and gives a description. This statement agrees with our observations as well, and the entry of the carotid artery into the basisphenoid is shaped differently in these taxa. However, this does not affect the homology of the foramen cavernosum character, and that part of the morphology is not related to the carotid as far as we can determine. Our analysis comes to different conclusions (as discussed elsewhere) from those of Broin (1991) and Lapparent de Broin (2000b), because we interpret this character is a synapomorphy of our infrafamily Erymnochelyodda , with Peltocephalus and Erymnochelys as sister taxa within that clade.

37. PT, pterygoid flange around cavum pterygoidei: (0) absent or very small; (1) partial; (2) complete.

Discussion: Figured by França and Langer (2006) and described in Gaffney et al. (2006: 603), we recognize two states, but do not order them.

38. PT, processus trochlearis pterygoidei: (0) absent; (1) oblique; (2) right angle.

Discussion: The processus trochlearis pterygoidei (figs. 23, 24, 70) is an important pleurodire synapomorphy described in Schumacher (1954, 1955a, 1955b, 1956) and Gaffney (1975b, 1979). This character is used in Gaffney and Meylan (1988), Rougier et al. (1995), Lapparent de Broin and Werner (1998), Lapparent de Broin (2000b), Fuente and Iturralde-Vinent (2001), and Gaffney et al. (2006). We have recognized two states as have a number of the above authors, and order them on the basis of state 1 occurring in the outgroup taxa and state 2 occurring only in the family Podocnemididae .

39. SO, roof exposure: (0) absent or slight; (1) present, small; (2) present, very large.

Discussion: Discussed by Meylan et al. (2009: 15), this character possibly requires three independent acquisitions with state 2 possibly being a fourth. The states could be ordered because the large extent of the supraoccipital requires the smaller state as an evolutionary predecessor.

40. SO, horizontal plate along ventral edge of crista supraoccipitalis: (0) absent; (1) present.

Discussion: This is a variable character recognized as more than a swelling along the ventral edge of the supraoccipital, but a distinct horizontal flange or plate. It is a synapomorphy for the subfamily Podocnemidinae and lost in Erymnochelys and Peltocephalus .

41. EX, occipital condyle: (0) basioccipital plus exoccipitals; (1) exoccipitals only.

Discussion: This character is described and discussed in Gaffney et al. (2006: 607). The absence of the basioccipital in the condylus occipitalis is a synapomorphy for the family Pelomedusidae but occurs elsewhere only in bothremydids and the podocnemidid, Peiropemys .

42. EX, foramen jugulare posterius: (0) closed partially; (1) closed completely.

Discussion: An extensive description and discussion is in Gaffney et al. (2006: 607); also see Gaffney and Wood (2002). In the present analysis, this character helps define the subfamily Podocnemidinae , which is all podocnemidids except Bauruemys .

43. EX, foramina nervi hypoglossi: (0) separated on occipital surface; (1) combined and recessed below occipital surface.

Discussion: In other podocnemidids the two pairs of foramina nervi hypoglossi open onto the posterior surface of the skull in clearly separated foramina. These foramina are relatively small, distinctly smaller than the foramen jugulare posterius that is just lateral to them. This condition is common throughout turtles ( Gaffney, 1979) and is interpreted as the primitive condition. In the tribe Stereogenyini , however, the two foramina nervi hypoglossi are very close together and separated by a relatively thin bar of bone. Both foramina are sunk below the surface to exit as a single foramen. This single foramen is about the same size as the foramen jugulare posterius in the tribe Stereogenyini . Although the foramen jugulare posterius varies in size throughout turtles, it is distinctly smaller in the tribe Stereogenyini than in other pelomedusoides such as Podocnemis and Pelusios . The occipital aspect of the tribe Stereogenyini , then, has two equally sized, paired foramina just lateral to the condylus occipitalis (fig. 65F), but one of these is the combined foramina nervi hypoglossi and the other is the foramen jugulare posterius.

This character is absent in Bairdemys sanchezi and Lemurchelys , which require reversals. It is present in Mogharemys and diagnostic of the entire tribe Stereogenyini .

44. BO, basioccipital short in ventral view, precondylar fossa very small: (0) no; (1) yes.

Discussion: In Shweboemys the basioccipital is a small curved bone, very short anteroposteriorly, with a small amount of ventral exposure. In Stereogenys the basioccipital is even more displaced posteriorly and has only a slight amount of ventral exposure. This results in a very small precondylar fossa in these two taxa. The other members of the tribe Stereogenyini have larger basioccipitals, similar or identical to other podocnemidids. The character is also discussed in Meylan et al. (2009: 15)

45. BO, basioccipital tubera width: (0) closer to median; (1) farther from median.

Discussion: This feature is described and discussed in Meylan et al. (2009: 15). In the present analysis, it is a synapomorphy for the subfamily Podocnemidinae , occurring outside this clade only in the bothremydids.

46. BO, horizontal occipital shelf: (0) absent; (1) present.

Discussion: In the primitive condition, such as in Euraxemys and chelids (but not pelomedusids), the occipital surface is relatively flat and vertical, with a variably expressed tuberculum basioccipitale. In Podocnemis , Bauruemys , Hamadachelys , and the tribe Peiropemydini , there is a horizontal shelf that extends posteriorly from the base of the occiput formed almost entirely by the basioccipital and partially by the exoccipital. The character is a synapomorphy for Podocnemididae plus Hamadachelys and is lost in the infrafamily Erymnochelyodda .

47. PR, ventral exposure: (0) completely exposed; (1) partially or completely covered.

Discussion: See Gaffney et al. (2006).

48. PR, pterygoid covers prootic: (0) no; (1) partially or completely.

Discussion: This character is described and discussed in Gaffney et al. (2006: 610).

49. PR, foramen posterius canalis carotici interni: (0) in prootic; (1) in basisphenoid within cavum pterygoidei; (2) variably in pterygoid, quadrate, basisphenoid; (3) pterygoid and basisphenoid.

Discussion: In all the podocnemidids examined by us the foramen posterius canalis carotici interni lies inside the cavum pterygoidei, in the medial wall of the basisphenoid; possibly in some cases the prootic enters the foramen. The larger cavum pterygoidei of Erymnochelys and Peltocephalus still has a formed foramen posterius canalis carotici interni in the basisphenoid despite the erosion of much of the bone defining the foramen posterius canalis carotici interni anteriorly.

50. OP, processus interfenestralis: (0) exposed; (1) covered.

Discussion: This character is described and discussed in Gaffney et al. (2006: 611).

51. OP, fenestra postotica: (0) open; (1) partially or completely closed.

Discussion: This character is described and discussed in Gaffney et al. (2006: 612).

52. BS, foramen nervi abducentis: (0) small; (1) moderate to large.

Discussion: This character is discussed in Meylan et al. (2009: 16). A relatively large foramen nervi abducentis helps define the genus Podocnemis . It also occurs in Pricemys , but the difficulty in seeing it in most fossils precludes its wider use.

53. BS, basisphenoid-quadrate contact: (0) present; (1) absent.

Discussion: This character is described and discussed in Gaffney et al. (2006: 613).

54. DEN, symphyseal contact: (0) fused; (1) sutured.

Discussion: This character is described and discussed in Gaffney et al. (2006: 616).

55. ART, processus retroarticularis: (0) long and posterior; (1) short or absent; (2) long and posteroventral.

Discussion: This character is described and discussed in Gaffney et al. (2006: 617; see also Gaffney et al., 1998; Gaffney and Forster, 2003). The long, posteroventrally directed processus retroarticularis is a feature of the epifamily Podocnemidinura.

56. ART, chorda tympani enclosed in processus retroarticularis: (0) no; (1) yes.

Discussion: The absence of lower jaws for many fossil taxa precludes the more general use of this character. However, it is present in all infrafamily Erymnochelyodda for which the lower jaw is known, but also occurs in Brasilemys and Euraxemys .

57. SP, splenial: (0) present; (1) absent.

Discussion: The splenial is absent in all of the taxa included (where area is preserved) except chelids.

58. Vertebrae, cervical centra saddle shaped: (0) absent, procoelous; (1) completely heterocoelic as in Podocnemis ; (2) wide as in Erymnochelys .

Discussion: The ‘‘saddle-shaped’’ or heterocoelic central articulations are described and discussed in Gaffney et al. (2006: 618) with references to the earlier works. This character is a synapomorphy for the redefined, subfamily Podocnemidinae , being absent as a reversal, or third state, only in Erymnochelys . In our analysis, Bauruemys is the only taxon in the family Podocnemididae with the primitive cervical condition (see below for Erymnochelys ). The absence of heterocoelic central articulations (‘‘saddle-shaped’’) in Bauruemys rests on Kischlat’s (1994) description and on our observations of the same material, MN V 4487, two cervicals of unknown position, but with simple, oval procoelous articular surfaces. However, other very similar cervicals are also seen among the uncataloged Bauruemys material in the DGM (specifically DGM uncataloged, collected 1969, Campos and Silva, carapace and plastron disarticulated with postcranial material in carapace part, including an eighth cervical vertebra).

Kischlat (1994: 348) described the two cervicals (MN V 4487) of unknown position of Bauruemys elegans as having simple, oval procoelous articular surfaces, that is, lacking saddle-shaped articulations. We have also examined these vertebrae and a number of similar cervicals in the uncataloged collections of the DGM, and we agree with the primitive condition of these articulations. The articulations are clearly similar to the outgroup cervicals, such as in Euraxemys and pelomedusids, in being a relatively simple, hemispherical, concave-convex articulation, not the more complex surface seen in Podocnemis and Peltocephalus .

Cervicals of Lapparentemys vilavilensis cervicals are known from a number of specimens and described by Broin (1991: 513) as ‘‘with saddle-shaped centra (like Podocnemis and Peltocephalus ).’’ We have examined these and agree that they are the same state as in the Recent Podocnemis and Peltocephalus .

The classically termed ‘‘saddle-shape’’ condition is worth reexamining, particularly in view of the discovery that the Miocene Turkanemys has a condition similar to that in Erymnochelys . Cervicals four to seven of Erymnochelys ( Vaillant, 1881; Tronc and Vuillemin, 1974) and Turkanemys , although not heterocoelic or saddle shaped with the curved central articulation extending posterolaterally as seen in Podocnemis and Peltocephalus , are also significantly different from the presumed primitive condition as seen in the Podocnemininae (as defined here) outgroups, such as Bauruemys and Euraxemys . The cervical articulation surfaces in Erymnochelys and Turkanemys are not simple oval articulations, but are wider than high and become V-shaped with dorsolateral extensions in the more posterior cervicals in the series. It is possible to interpret this as morphologically intermediate between the primitive condition and the ‘‘saddle-shaped’’ condition, but we do not make that assumption. Rather, we code this as a third state, unordered, but not the same state as the primitive condition, from which it differs considerably. The MPCs do not differ if the Erymnochelys and Turkanemys condition is coded as primitive, whether two or three states are used. Also the MPC is the same if the states are ordered.

59. Vertebrae, second cervical biconvex: (0) no; (1) yes.

Discussion: This character is described and discussed in Gaffney et al. (2006: 618).

60. Vertebrae, cervical zygapophyses: (0) none fused; (1) some fused.

Discussion: This character is described and discussed in Gaffney et al. (2006: 618).

61. Carapace, cervical scale: (0) present; (1) absent.

Discussion: This character is described and discussed in Gaffney et al. (2006: 621).

62. Carapace, nuchal bone width: (0) two or more times length; (1) width greater than length, but less than two times; (2) width equals length; (3) very reduced nuchal ( Araripemys ).

Discussion: This character is described and discussed in Gaffney et al. (2006: 621).

63. Carapace, neural series extent: (0) to suprapygal; (1) to costals eight; (2) to costals seven; (3) to costals six; (4) neurals discontinuous or absent.

Discussion: This character is described and discussed in Gaffney et al. (2006: 622). As many of the genera analyzed here lack shells, the distribution of this character lacks specificity. However, it and the closely related character, neural number, are consistent with the Erymnochelys + Peltocephalus + tribe Stereogenyini clade, as well as a monophyletic Podocnemis .

64. Carapace, neural number: (0) eight or more; (1) seven; (2) six; (3) five or fewer.

Discussion: This character is described and discussed in Gaffney et al. (2006: 622). See also neural series extent character above.

65. Carapace, keeled neurals: (0) none; (1) some.

Discussion: Two species of Podocnemis , P. sextuberculata and P. unifilis , have this character, but no other characters support this grouping and it is not present in the shortest tree.

66. Carapace, four-sided neural position: (0) first neural; (1) second neural; (2) third neural; (3) neurals absent or discontinuous; (4) four-sided neural absent.

Discussion: This character is described and discussed in Gaffney et al. (2006: 623). All Podocnemididae have the first neural four sided, except Bairdemys venezuelensis (neurals absent) and Bauruemys (second neural).

67. Carapace, costal two anterior edge thickened near buttress: (0) no; (1) yes.

Discussion: Apparently present only in Bauruemys, Peirópolis B , and the shell taxon Roxochelys (see discussion in Systematics section).

68. Carapace, axillary buttress: (0) reaches peripheral three; (1) reaches peripheral two.

Discussion: This character is described and discussed in Gaffney et al. (2006: 626). In the present analysis it forms a group containing all Podocnemis except P. vogli , and occurs independently in Cambaremys .

69. Plastron, axillary musk duct (1): (0) in buttress; (1) absent in buttress.

Discussion: Unknown for all fossils and scored only for the living podocnemidids, it forms no groups.

70. Plastron, axillary musk duct (2): (0) not in bridge; (1) one opening in bridge; (2) three openings in bridge; (3) four openings in bridge.

Discussion: Unknown for all fossils and scored only for the living podocnemidids, it resolves some groups within Podocnemis . State 2 occurs in P. sextuberculata , P. erythrocephala , and P. unifilis .

71. Plastron, mesoplastra: (0) small and lateral; (1) absent.

Discussion: This character is described and discussed in Gaffney et al. (2006: 628).

72. Plastron, pectoral scales contact mesoplastra: (0) yes; (1) no.

Discussion: This character is described and discussed in Gaffney et al. (2006: 637). As currently known for this analysis, it differentiates Lapparentemys from all other epifamily Podocnemidinura in which the shell is known.

73. Plastron, pectoral scales contact entoplastron: (0) no; (1) yes.

Discussion: This character is described and discussed in Gaffney et al. (2006: 630). It occurs in all epifamily Podocnemidinura in which the shell is known.

74. Plastron, pectoral scales contact epiplastra: (0) no; (1) yes.

Discussion: This character is described and discussed in Gaffney et al. (2006: 636). It occurs in all members of the family Podocnemididae . It is absent in Portezuelemys and unknown in Hamadachelys and Brasilemys .

CONCLUSIONS

The phylogenetic analysis (fig. 98) was made with PAUP* version 4.0b10 ( Swofford, 2002) using the parsimony algorithm. All characters were run unweighted and unordered. Characters were entered and cladograms examined using MacClade version 4.05 ( Maddison and Maddison, 2000). The character matrix (appendix 1) used as the basis of this analysis has 37 taxa and 74 characters. Five of these characters are parsimony uninformative. Characters 1 (presence of nasals), 57 (presence of splenial), 59 (second cervical biconvex), and 61 (cervical scale absent) are variable but absent in only one taxon, the Chelidae , the most extreme outgroup in this analysis. Character 34 (medial process of quadrate present) occurs in all taxa. These characters were included as they were used in the matrix of Gaffney et al. (2006). The PAUP* analysis of appendix 1, using both heuristic and stepwise addition (random) searches, results in 9 equally parsimonious trees of 173 steps (fig. 98).

Although we have run all multistate characters as unordered, some arguments could be made for ordering at least some of them. These are discussed under the relevant character listed above. The possibility exists that some multistate characters should be ordered even though we have no evidence for this.

The outgroups chosen are based on the work of Gaffney et al. (2006). The Chelidae , Pelomedusidae , Araripemys , Euraxemydidae , and Bothremydidae are dealt with in greater detail in that paper.

The characters used here have been developed over the course of the pleurodire work by E.S.G. and P.A.M. An early version of our matrix was published in Meylan (1996) and a later version in Meylan et al. (2009). We have incorporated characters from the literature as well, particularly Lapparent de Broin (2000b), Fuente (2003), França and Langer (2006), and Romano and Azevedo (2006). Cadena et al. (2010) became available too late to be considered in our analysis, but the matrix used in that paper is similar to ours.

The family Podocnemididae consists of 20 genera and 30 species diagnosable by cranial characters (an additional shell taxon is present) of which three genera and eight species persist into the Recent fauna. The family extends from the Late Cretaceous to the Recent and occurs in North and South America, Europe, and Africa. The family Podocnemididae is reconfirmed as monophyletic, using the unique structure, a cavum pterygoidei that is formed by the basisphenoid, pterygoid, prootic, and quadrate, underlain by pterygoid and basisphenoid, among other characters.

A phylogenetic analysis analyzes 33 taxa in the Podocnemididae (37 taxa including outgroups). The resulting nine equally parsimonious cladograms are the basis for a new classification of the family. Much of our basal resolution agrees with that of França and Langer (2006), which can be modified and restated as follows: ( Bauruemys ( Lapparentemys ( Podocnemis ( Peltocephalus , Erymnochelys )))). Cadena et al. (2010) could also be restated as ( Bauruemys Lapparentemys ( Podocnemis ( Peltocephalus , Erymnochelys ( Neochelys ( Dacquemys ( Bairdemys , Shweboemys , Stereogenys ))))), and agreeing in its main resolution with ours as well, particulary in the resolution of the living podocnemidids as: ( Podocnemis ( Peltocephalus , Erymnochelys )).

PODOCNEMIDIDAE

Within the family Podocnemididae , the sister taxon to all other podocnemidids is Bauruemys elegans ( Suárez, 1969a) . All other podocnemidids, the redefined subfamily Podocnemidinae Cope, 1868 , are united by a long quadratojugal-parietal contact, a completely enclosed foramen jugulare posterius, wide basioccipital tubera, horizontal plate present on ventral edge of crista supraoccipitalis (except in Peltocephalus and Erymnochelys ), and cervical centra saddle shaped (except in Erymnochelys and Turkanemys , which nonetheless lack the primitive condition).

BASAL PODOCNEMIDIDAE

A basal group of Cretaceous-Paleocene podocnemidids that are the sister group to all remaining podocnemidids are Peiropemys mezzalirai , n. gen. et sp., Lapparentemys vilavilensis (Broin) , n. gen., and Pricemys caiera , n. gen. et sp. The remaining podocnemidids are the infrafamily Podocnemidodda Cope, 1868 , new rank, that are characterized by the possession of a cheek emargination that does not reach above the level of the orbit, the medial expansion of triturating surface with a median maxillary ridge present, and the presence of accessory ridges. This group contains the living podocnemidids and a series of extinct forms, including the marine broad-jawed taxa. Meylan (1996) recognized ‘‘ Roxochelys ’’ (5 Lapparentemys ) as the sister group to all other Podocnemididae that were in his analysis, which excluded Bauruemys .

INFRAFAMILY PODOCNEMIDODDA

Within the Podocnemidodda , the genus Podocnemis is the sister group to all the remaining taxa, i.e., the magnatribe Erymnochelydand. The infrafamily Podocnemidodda is characterized by the medial expansion of the triturating surface with the median ridge, and the presence of one or two accessory ridges on the triturating surface.

The magnatribe Podocnemidand consists only of the genus Podocnemis . Our resolution of the Podocnemis species is: ( P. vogli ( P. lewyana ( P. unifilis ( P. erythrocephala ( P. sextuberculata , P. expansa ))))). We do not consider this part of the analysis to be strongly supported as the resolution collapses into a multichotomy in one step more than the shortest resolution. Osteologically speaking, these Recent species are very similar to each other and have very conservative morphologies, even within a group that has changed little since the Late Cretaceous.

MAGNATRIBE ERYMNOCHELYDAND

The magnatribe Erymnochelydand, with all taxa included, is diagnosed only by the reduced cheek emargination, which is reversed in Bairdemys and Cordichelys . Caninemys is the sister taxon to all remaining magnatribe Erymnochelydand, which is united by the absence of a vomer (present in Caninemys but also absent in most, possibly all, Podocnemis ) and the lateral, anterolateral orbital position, both characters homoplastic in other pleurodires. However, this is primarily the effect of missing characters in these taxa and, if Caninemys , Dacquemys , and UCMP 42008 are exclud- ed, collapsing these nodes, the number and consistency of characters uniting the remaining clade increases: 2, orbits facing laterally, anterolaterally; 11, cheek emargination slight; 30, fossa precolumellaris very small to absent; 39, supraoccipital roof exposure absent or slight; and 56, chorda tympani enclosed in processus retroarticularis in lower jaw.

The condition of the cavum pterygoidei is not preserved in Caninemys , Dacquemys, UCMP 42008, or Turkanemys , so the distribution of this character is unknown for a significant number of Erymnochelydand taxa, but it is known to be present above Turkanemys .

Within the magnatribe Erymnochelydand the resolution of taxa can be seen in figure 98. The group above Caninemys and Dacquemys + UCMP 42008 is united by the premaxillae entering the apertura narium interna in ventral view. The group above Caninemys , Dacquemys + UCMP 42008, and Turkanemys is united by the absence of the horizontal occipital shelf.

Cervical vertebrae 4–7 of both Erymnochelys and Turkanemys are very similar to each other in being wider than high and differ from other Podocnemidinae (as defined here) in lacking the well-developed heterocoely or saddle-shaped centra. This condition could be interpreted as an intermediate between the Podocnemis fully heterocoelous centra that wrap around posterolaterally and the condition seen in the basal podocnemidid, Bauruemys . Nonetheless, our analysis still places Turkanemys outside Erymnochelys + Peltocephalus requiring that the wide articular condition be acquired twice, despite the proximity of these taxa in the cladogram. Again, this is only a few steps from a group containing Erymnochelys , Peltocephalus , and Turkanemys . We feel that although our MPC resolves these three taxa, in view of the missing data for Turkanemys it would be more realistic to conclude that the three are an unresolved trichotomy at present.

In contrast to some previous analyses, among the living taxa, our results show Podocnemis as the sister taxon to Erymnochelys + Peltocephalus . As constituted here, the Erymnochelydand is united only by the small to absent cheek emargination, a character subject to frequent homoplasy in turtles. However, some of the fossil taxa (for example, Turkanemys , Caninemys , Dacquemys ) are not known for some important characters and, if only the living species are used in the analysis, Erymnochelys + Peltocephalus are united by a number of characters: 2, orbits facing laterally, anterolaterally; 9, jugal-quadrate contact present; 11, cheek emargination slight; 14, premaxillae reach apertura narium interna (also in some Podocnemis ); 36, anterior opening of the cavum pterygoidei large and including foramen cavernosum in roof; 39, supraoccipital roof exposure slight or absent; 46, horizontal occipital shelf absent; 56, chorda tympani enclosed in processus retroarticularis; 63, neural series extends to costals six; and 70, axillary musk duct not in bridge.

When all taxa are included, Peltocephalus and Erymnochelys are united by the jugal-quadrate contact and the loss of supraoccipital exposure on the skull roof. Although the jugal-quadrate contact and associated loss of cheek emargination do appear in cryptodires, it is a unique character within pleurodires. The supraoccipital character is probably subject to more individual variation than we have seen and may not be useful at this level of refinement. However, it should be noted that despite arguments for homoplasy of characters in common ( Lapparent de Broin, 2000b, 2003b), there are similarities between Peltocephalus and Erymnochelys that are hard to ignore. Even the shells of the two taxa are similar. The carapace of both usually has seven neurals with costals seven and eight meeting on the midline (with some individual variation). The plastron has a pectoral-abdominal sulcus very close to and just barely contacting the mesoplastron, the pectoral scale extends onto the posterior half of the entoplastron, and the very small intergular scale just touches or barely extends onto the entoplastron. The main shell difference has to do with the gular scales, which are large and meet on the midline in Erymnochelys and are small and are separat- ed in Peltocephalus .

The cervicals of Erymnochelys ( Vaillant, 1881; Tronc and Vuillemin, 1974) and Turkanemys , although not heterocoelic or saddle shaped as in Podocnemis , are also not the same as the presumed primitive condition as seen in outgroups such as Bauruemys and Euraxemys . It is possible to interpret this as morphologically intermediate between the primitive condition and the ‘‘saddle-shaped’’ condition, but we do not make that assumption. Rather, we code this as a third state, run unordered, not the same state as the primitive condition, from which it differs considerably. The MPCs do not differ if the Erymnochelys and Turkanemys condition is coded as primitive, whether two or three states are used. Also the MPC is the same if the states are ordered.

As described above, Peltocephalus + Erymnochelys are united with remaining magnatribe Erymnochelydand by the premaxillae reaching the apertura narium interna and the absence of the horizontal occipital shelf. The group above Peltocephalus + Erymnochelys consists of an unresolved trichotomy of Neochelys , Papoulemys , and the tribe Stereogenyini . These are united by the parietal-pterygoid contact in the septum orbitotemporale.

While the relative position of Neochelys may rest on one character, its resolution within the magnatribe Erymnochelydand, like that of Peltocephalus + Erymnochelys , is well supported by the same characters that unite Peltocephalus + Erymnochelys with the tribe Stereogenyini . The resolution of Caninemys within the Erymnochelydand is not strongly supported; in only one step it becomes a multichotomy with Podocnemis and the infrafamily Peiropemydodda . Neochelys-Papoulemys and Dacquemys , however, are strongly supported as part of the magnatribe Erymnochelydand, as proposed earlier ( Broin, 1991; Lapparent de Broin, 2000b, 2001, 2003a, 2003b).

TRIBE STEREOGENYINI

The tribe Stereogenyini has a dorsal process of the palatine that reaches the frontal in the septum orbitotemporale, the fossa precolumellaris is absent, and both foramina nervi hypoglossi are combined and recessed in a short canal that opens on the occipital surface. Within the tribe Stereogenyini , Mogharemys is the sister taxon to the well-defined subtribe Stereogenyina . Until the description of Bairdemys ( Gaffney and Wood, 2002) , the subtribe Stereogenyina consisted only of Stereogenys and Shweboemys and has been recognized as monophyletic by the common possession of a well-developed secondary palate with a midline cleft. Stereogenys and Shweboemys have been referred to as the ‘‘groupe Schweboemys [sic]’’ Broin (1988), the ‘‘B9 taxon’’ ( Gaffney and Meylan, 1988), the ‘‘ Schweboemys [sic] group’’ ( Lapparent de Broin, 2000b), and the ‘‘ Shweboemys Group’’ ( Gaffney and Wood, 2002). The subtribe Stereogenyina is united by these characters: a secondary palate unique among turtles in being formed by maxillae and palatines that are separated on the midline by a narrow cleft, a palate characterized by a variably developed anteri- or convexity and posterior concavity, the palatine making up half or more of the secondary palate, and a fossa orbitalis with an extensive posterior pocket behind the orbital rim enclosed by the septum orbitotemporale.

Two groups are recognized within the subtribe Stereogenyina : the infratribe Bairdemydita , which has a cheek emargination that reaches the level of the orbit (only known for one genus), a labial ridge that is high and narrow, and the angle of the front of the skull is 90 ° or less. The infratribe Stereogenyita has a labial ridge that is low and blunt, a pinched snout (not known for Brontochelys ), a cheek emargination that does not reach the level of the orbit, and the angle of the front of the skull exceeds 90 °.

Within the infratribe Bairdemydita the genera resolve as ( Cordichelys ( Latentemys , Bairdemys )) with one character, the separation of the eustachian tube by bone from the fenestra postotica, found only in Latentemys and Bairdemys . The four species of Bairdemys are united by a vertical flange on the ventral surface of the squamosal. Three species, Bairdemys hartsteini , B. venezuelensis , and B. sanchezi are united by the presence of a strongly convex prefrontal. Two of these, B. hartsteini , and B. venezuelensis are united by the strong anterior convexity on the palate. The resulting resolution of the group is: ( Cordichelys ( Latentemys ( B. winkleri ( B. sanchezi ( B. hartsteini , B. venezuelensis ))))).

Within the infratribe Stereogenyita , Lemurchelys , Shweboemys , and Stereogenys are united by the medial edges of the midline palatal cleft being parallel rather than curved. Shweboemys and Stereogenys are united by a number of characters and are the best supported of all the groups in the subtribe Stereogenyina . They have a very short basioccipital, a palatine-basisphenoid contact that separates the pterygoids, a dorsal process of the palatine that reaches the parietal, and a relatively long secondary palate. The resolution of the infratribe is ( Brontochelys ( Lemurchelys ( Shweboemys , Stereogenys ))).