Dortoka, Lapparent de Broin and Murelaga, 1996

GAFFNEY, EUGENE S, TONG, HAIYAN & MEYLAN, PETER A, 2006, EVOLUTION OF THE SIDE-NECKED TURTLES: THE FAMILIES BOTHREMYDIDAE, EURAXEMYDIDAE, AND ARARIPEMYDIDAE, Bulletin of the American Museum of Natural History 300 (300), pp. 1-698 : 571-655

Dortoka

This shell taxon was described by Lapparent de Broin and Murelaga (1999).

Chelidae

Monophyly of the Chelidae is well corroborated. Within-group relationships based on morphology ( Gaffney, 1977b; Gaffney and Meylan, 1988) and molecules ( Shaffer et al., 1997; Seddon et al., 1997; Georges et al., 1998; Krenz et al., 2005; Near et al., 2005) differ significantly, and the molecular phylogenies seem to be well supported. The molecular analyses tend to recognize separate Australian and South American clades, but these relationships become less certain when all chelid genera are represented in the dataset ( Georges et al., 1998). We have not adopted a within-group cladogram for chelids, and we have therefore coded some characters as variable. Note that recent discoveries of Cretaceous chelids (Fuente et al., 2001; Lapparent de Broin and Fuente, 2001; Fuente, 2003; Bona and Fuente, 2005) support the morphology-based analyses.

CHARACTER DESCRIPTIONS

The characters chosen here include ones not only relevant for bothremydids but for all pleurodires, outside podocnemidids, and chelids. Although characters primarily resolving Proganochelys and all other turtles are included in the analysis, they are only referenced and briefly described, having been discussed elsewhere. The same format of headings used in Gaffney (1996) is used here. The history of use of particular characters in the literature is primarily taken from papers with actual datasets (i.e., Meylan, 1996; Tong et al., 1998; Lapparent de Broin and Murelaga, 1999; Fuente and Iturralde-Vinent, 2001) which explicitly show taxon distribution, although reference is also made to literature with more extensive character descriptions (i.e., Antunes and Broin, 1988; Lapparent de Broin and Werner, 1998; Lapparent de Broin, 2000a), even though they lack character matrices.

These descriptions are also shown in the Character List (appendix 2) and their distributions are shown in the Data Matrix (appendix 3). In the Character List, the senior author has attempted to organize the characters by bone in the same order as in the Cranial Morphology sections. In order to keep track of which bone is involved, in the Character List each character is preceded by the abbreviation of the bone followed by the character name. However, AMNH editorial policy requires that when used as text these abbreviations must be spelled out. Therefore instead of ‘‘NA, nasal bones’’ the discussion below identifies the character as the more cumbersome ‘‘Nasal, nasal bones.’’ And so on.

1. Nasal, nasal bones: present ( Emydura ) 5 0; absent ( Pelusios ) 5 1.

Morphology: When present (fig. 6A), nasals make up the anteriormost part of the skull roof anterior to the orbits. Skull figures showing the morphology of nasal bones in the turtles in which they occur are in Gaffney (1979a). The absent condition is seen in figures 6B and 21A. See discussion of this character in cryptodires in Gaffney (1996).

Primitive condition: Paired nasals are present throughout amniotes and are present in Proganochelys , most chelids, and primitive cryptodires.

Homoplasy: There is good evidence that nasals have been lost independently within the Pleurodira and Cryptodira . Within the Pleurodira they are absent in all Pelomedusoides for which this portion of the skull is known. Among chelids they are absent or fused to the frontals only in the genus Chelus ( Gaffney, 1979a: fig. 144).

Discussion: The absence of nasal bones is a synapomorphy for Pelomedusoides. This character is used in Antunes and Broin (1988), Gaffney and Meylan (1988), Gaffney et al. (1991), Meylan (1996), and Lapparent de Broin and Werner (1998).

2. Lacrimal, lacrimal bone: present ( Proganochelys ) 5 0; absent ( Emydura ) 5 1.

See Gaffney (1990) and Gaffney et al. (1991) (also used in Gaffney and Meylan, 1988; and Rougier et al., 1995).

3. Lacrimal, lacrimal foramen: present ( Proganochelys ) 5 0; absent ( Pelusios ) 5 1.

See Gaffney (1990), Gaffney et al. (1991), and Gaffney and Kitching (1995) (also used in Gaffney and Meylan, 1988; and Rougier et al., 1995).

4. Prefrontal, prefrontals meet on midline in dorsal view: no ( Proganochelys ) 5 0; yes ( Pelusios ) 5 1.

Morphology: When nasals are present, nasal-frontal contact typically excludes midline contact of the prefrontals (fig. 6A). However, prefrontals are not always in contact when nasals are absent (e.g., Chelus, Gaffney, 1979a : fig. 144), and on occasion prefrontals may be in midline contact although nasals are present (e.g., Hydromedusa, Gaffney, 1979a : fig. 146). State 1 is seen in figure 6B–J.

Primitive condition: The nasal-frontal contact to the exclusion of prefrontal midline contact in Proganochelys , most chelids, and primitive cryptodires suggests that prefrontals meeting on the midline represents a derived condition.

Homoplasy: Except for Hydromedusa ( Gaffney, 1979a, 1979c), none is known within pleurodires.

Discussion: Prefrontals meeting on the midline is a synapomorphy for the Pelomedusoides. This character is used in Gaffney and Meylan (1988) and Gaffney et al. (1991).

5. Prefrontal, preorbital skull broad: narrow ( Galianemys ) 5 0; very broad ( Bothremys ) 5 1.

Morphology: The snout or preorbital region of turtles generally forms a wide angle of about 70–90 °. However, in some species the snout is greatly widened anteriorly, giving the skull a very broad outline in dorsal view (figs. 7, 136, 146). Many bones make up this region; we have arbitrarily chosen to treat this character under the prefrontal.

Primitive condition: The snout of Proganochelys forms an angle of about 80 ° and those of primitive cryptodires and chelids are similarly uninflated, and this is presumed to be the primitive condition.

Homoplasy: Araiochelys lies within the Bothremydini and it is much narrower than other Bothremydini (fig. 7E) and is a reversal from the very broad condition seen in other Bothremydini .

Discussion: A wide snout, in excess of 80 °, is a synapomorphy for the Bothremydini .

6. Prefrontal, anterior margin: straight, broadly convex margin in dorsal view ( Pelomedusa ) 5 0; narrow midline process, at least partially dividing nares ( Bothremys ) 5 1.

Morphology: The dorsal margin of the apertura narium externa forms an anterior process on the midline, partially dividing the apertura (fig. 7G, H). In some bothremydids, such as Araiochelys (fig. 125D), the process almost completely divides the apertura.

Primitive condition: Although the anterior margin is broadly convex in pelomedusids, it is relatively straight in chelids and euraxemydids.

Homoplasy: This character has a CI of 0.33 for this dataset and has originated or been lost multiple times elsewhere within the Pleurodira .

Discussion: The protrusion is present in Araiochelys and Bothremys within the Bothremydini . It is present only in Labrostochelys and Ummulisani , among the Taphrosphyini , suggesting that it appeared twice within the family. It helps to define the group Araiochelys , Bothremys , and Chedighaii within the Bothremydina . There is some ambiguity in identifying the degrees of protrusion. Comparing Bothremys cooki and some Galianemys specimens shows only a slight difference. As coded, it supports the ( Araiochelys , Bothremys , Chedighaii ) group.

7. Prefrontal, prefrontal-palatine contact: widely separated ( Galianemys ) 5 0; closely spaced or in contact ( Bothremys ) 5 1.

Morphology: Cryptodires generally have a large ventral process of the prefrontal that contacts the palatine posteriorly ( Gaffney, 1979a: 73). However, in most pleurodires the descending process of the prefrontal is smaller and usually not in contact with the vomer or palatine. State 1 is seen through the orbit in Labrostochelys (fig. 193). In the taxa in which the bones are close but not in contact (e.g. Bothremys cooki , fig. 130A), there is no overlap in morphology with those taxa that have these bones widely separated.

Primitive condition: Gaffney (1990: 37) reported a wide prefrontal-palatine contact in Proganochelys , so this contact appears to be primitive for turtles. However, the contact is absent in chelids other than an occasional Emydura ( Gaffney,1979a: 73) and it is absent in pelomedusids, except for an occasional Pelusios ( Gaffney, 1979a: 73) , Araripemys , euraxemydids, Hamadachelys , and members of the Podocnemididae . This distribution suggests that the absence of this contact is primitive for the Bothremydidae .

Homoplasy: Among the Pleurodira this contact occurs as a variant in some individuals of Pelusios and Emydura ( Gaffney, 1979a: 73) . These appearances occur independently of those within the Bothremydidae . The character occurs independently in the Bothremydini ( Araiochelys , Bothremys , and Chedighaii ) and in the Taphrosphyini ( Labrostochelys ).

Discussion: In spite of homoplasy, this character helps resolve relationships within Bothremydini .

8. Prefrontal, prefrontal-vomer contact: absent ( Proganochelys ) 5 0; present ( Pelusios ) 5 1.

This character is used by Gaffney and Meylan (1988), Gaffney (1990), Gaffney et al. (1991), and Rougier et al. (1995).

9. Prefrontal, fissura ethmoidalis: very wide ( Proganochelys ) 5 0; narrower ( Pelusios ) 5 1.

See Gaffney (1990).

10. Prefrontal, prefrontal-parietal contact: absent ( Galianemys ) 5 0; present ( Phosphatochelys ) 5 1.

Morphology: In most turtles the skull roof is formed by paired prefrontals, frontals, and parietals, with the frontals normally intervening between the other pairs. Prefrontal-parietal contact occurs when the frontals are reduced (fig. 196) or absent (fig. 204).

Primitive condition: In the skull roof of Proganochelys and other outgroups this contact is absent ( Gaffney, 1990: figs. 16, 114), as is the case in all cryptodires, chelids, pelomedusids, and podocnemidids. Absence of this contact is clearly primitive.

Homoplasy: We are not aware of any homoplasy in this character. When the frontals are greatly reduced in cryptodires (i.e., Platysternon ), the postorbitals intervene between the prefrontals and parietals ( Gaffney, 1979a: 221).

Discussion: This unique feature is a synapomorphy for the two genera Ummulisani and Phosphatochelys .

11. Frontal, orbits facing upward: orbits facing more laterally ( Galianemys ) 5 0; orbits facing more dorsally ( Bothremys ) 5 1.

Morphology: In most turtles the skull roof forms a horizontal table, with the maxillae forming nearly vertical walls and the orbits open laterally in a vertically oriented face. In Bothremys and Chedighaii (figs. 7, 127) there is very little change in slope from the roof and the face of the skull so that the orbits face dorsally and only slightly laterally.

Primitive condition: The orbits of Proganochelys are quite vertically oriented ( Gaffney, 1990: figs. 16, 114). In primitive cryptodires, chelids, pelomedusids, and podocnemidids, orientation is variable but generally more vertical than horizontal. Thus, the more vertical orientation is considered primitive.

As coded here, the primitive condition includes a wide variety of orbital orientations. However, efforts to specify states, such as nearly vertical, laterally facing orbits as found in Phosphatochelys and Azabbaremys , were abandoned due to the difficulty of distinguishing gradational conditions in other Taphrosphyini and Cearachelyini .

Homoplasy: Within the Bothremydidae no forms other than Bothremys and Chedighaii have dorsally facing orbital openings. However, Araripemys has the closest approach to this condition, and within chelids the tribe Chelini of Gaffney (1977b) has dorsally facing orbits.

Discussion: The horizontal orbits of Bothremys and its friends are presumably related to the great anterior expansion and flattened shape of the skull. This character is a synapomorphy for Bothremys + Chedighaii .

12. Frontal, foramen interorbitale: high ( Galianemys ) 5 0; low ( Bothremys ) 5 1.

Morphology: The foramen interorbitale is the opening medial to the orbit defined by the sulcus olfactorius dorsally, the descending process of the parietal posteriorly, and the vomer and dorsal surface of the palatal elements ventrally (fig. 21). In most turtles this opening is round or oval in shape ( Gaffney, 1979a). In the Bothremydini the foramen interorbitale is low and reduced in height by a high palate and low sulcus olfactorius (fig. 144).

Primitive condition: The foramen interorbitale is quite tall and rounded in Proganochelys ( Gaffney, 1990) . It is generally broadly rounded in cryptodires and all pleurodires other than Chelus and members of the Bothremydini .

Homoplasy: Independent acquisition of this character has occurred in Chelus ; it has no homoplasy within Pelomedusoides.

Discussion: Although the low foramen interorbitale is diagnostic of the Bothremydini , the character is indeterminable in some Bothremydini ( Polysternon , Zolhafah , Bothremys arabicus , and Chedighaii barberi ). Examination of the endocast (YPM PU 12951) of a presumed C. barberi shows that it was low.

13. Parietal, quadratojugal-parietal contact: absent ( Proganochelys ) 5 0; present, quadratojugal large and anterior to quadrate ( Euraxemys ) 5 1; present, quadratojugal small and dorsal to quadrate ( Phosphatochelys ) 5 2.

Morphology: In the skull roof of turtles, the parietal is normally separated from the jugal and quadratojugal by an elongate postorbital that reaches the temporal margin of the skull. In some forms, contact of the quadratojugal and parietal prevents posterior exposure of the postorbital on the temporal margin. This can occur due to a very large quadratojugal that reaches the cheek emargination and extends posteromedially over the quadrate and meets the parietal medially (state 1, figs. 42, 47), or by a laterally enlarged parietal that meets a small quadratojugal that lies dorsal to the quadrate and does not reach the cheek emargination (state 2, figs. 178, 183, 185). We treat these two morphologies as independent states of the same character.

Primitive condition: It is clear that the primitive condition for the turtle skull roof is to have the postorbital separating the parietal and quadratojugal. The postorbital either reaches the squamosal ( Proganochelys , Kayentachelys , Pleurosternon , and other primitive cryptodires; Gaffney, 1990) or the temporal margin of the skull (Polycryptodira, Pelomedusidae , Araripemys ). A parietal-quadratojugal contact is clearly a derived condition.

Homoplasy: The presence of a quadratojugal-parietal contact with a large quadratojugal (state 1) occurs in Euraxemydidae and Podocnemididae + Hamadachelys . The most parsimonious cladogram shows these as independent acquisitions.

Discussion: State 1 supports Euraxemydidae and, independently, Podocnemididae + Hamadachelys . State 2, quadratojugal-parietal contact with a large parietal and small quadratojugal, occurs in Taphrosphys , Ummulisani , Labrostochelys , and Phosphatochelys and is without homoplasy in the subgroup of Taphrosphyini that it defines.

This character is used in Gaffney and Meylan (1988) and Lapparent de Broin (2000a).

14. Parietal, temporal emargination: absent, slight ( Proganochelys ) 5 0; extreme ( Kurmademys ) 5 1; intermediate ( Galianemys ) 5 2.

Morphology: The fossa temporalis superi- or of turtles contains the adductor mandibulae and is primitively roofed over by the skull roof elements, particularly the parietal, squamosal, postorbital, and quadratojugal. Emargination of this skull roof both posteriorly (temporal emargination) and laterally (cheek emargination) is a common theme in turtle evolution ( Gaffney, 1979a: 83–86).

In Pelomedusoides the temporal emargination varies a great deal, and dividing this variation into a suite of characters that have a good chance of being homologous is difficult due to the simple morphology involved. We have chosen to identify three conditions in order to obtain some information from this morphologic area. The most extreme emargination, state 1, is relatively easy to recognize. In state 1 only a narrow bridge of bone is present between the orbit and the temporal margin, as found in pelomedusids, Araripemys (fig. 26), and the tribe Kurmademydini (figs. 5, 54, 65). Although it is somewhat subjective, the degree of emargination seen in the tribe Cearachelyini (fig. 81), state 2, is defined as intermediate between state 1 and the remaining bothremydids. In state 2 the otic chamber is fully exposed and the bone (postorbital) between the temporal margin and orbit is at least twice as long as in state 1. The principal ambiguity is differentiating state 0 from state 2 in some Bothremydini (such as Bothremys maghrebiana and Chedighaii , which have most but not all of the otic chamber exposed), and the degree is partially dependent on how the skull is oriented.

Primitive condition: Proganochelys ( Gaffney, 1990) has a complete skull roof with no emargination. In the Chelidae ( Gaffney, 1979c) temporal emargination is slight or absent except in those cases where cheek emargination appears to have joined with temporal emargination. In other outgroups to the Bothremydidae , temporal emargination is extreme ( Pelomedusidae , Araripemys ), moderate ( Hamadachelys ), or variable ( Podocnemididae ). The extensive skull roof of the Bothremydini and the Taphrosphyini is interpreted as a reversal.

Homoplasy: This is a character in which there is a great deal of homoplasy within cryptodires, both in the form of independent acquisition of states and as reversals. The presence of extensive skull roofs in primitive cryptodires such as Kayentachelys , Pleurosternon, Glyptops , and most baenids ( Gaffney, 1979b) also suggests that well-developed emargination has occurred independently in pleurodires and cryptodires. Similarly, the presence of well-roofed forms in the Meiolaniidae, Chelonioidea , Chelydridae , Podocnemididae , and Bothremydidae suggests that secondary redevelopment of the skull roof has occurred on multiple occasions.

The emargination states as scored here show homoplasy in state 1, but none in state 2. The extreme temporal emargination of state 1 has arisen three times in the MPC. The conditions in Kurmademys and Sankuchemys are very similar, but the skulls of Araripemys and pelomedusids are not, and the emargination is not identical.

State 2, intermediate or moderate emargination, is not as subjective as it sounds and is recognizable within the Pelomedusoides. The degree of emargination is very similar in Cearachelys and Galianemys , supporting the monophyly of the Cearachelyini .

Discussion: The principal difficulty with this character is the primitive condition, state 0, a well-roofed skull. The morphology of Proganochelys and the Taphrosphyini , for example, is not even similar and is questionably homologous on morphological grounds alone. Yet the well-roofed skulls of the Taphrosphyini and the Bothremydini are similar and probably homologous considering that the two immediate outgroups, the tribes Cearachelyini and Kurmademydini , have distinctly greater degrees of temporal emargination.

This character is used in many papers (see Gaffney and Meylan, 1988; Lapparent de Broin and Werner, 1998; Lapparent de Broin and Murelaga, 1999; Lapparent de Broin, 2000a).

15. Parietal, parietal-squamosal contact: present ( Proganochelys ) 5 0; absent ( Pelusios ) 5 1.

Morphology: This condition (state 1) is related to the degree of posterior skull roof (temporal) emargination, and it is seen in most Pelomedusoides (figs. 6–8).

Primitive condition: The contact is present (state 0) in Proganochelys , Kayentachelys and other primitive cryptodires, and chelids and is considered the primitive condition.

Homoplasy: Skull roof emargination is homoplastic in this analysis (see character 14), but the widespread loss of the squamosal-parietal contact in Pelomedusoides suggests that at least the development of temporal emargination to the point of parietal-squamosal separation is synapomorphic for the group. Dirqadim is the only reversal.

Discussion: It is possible to put this character as a state in character 14, temporal emargination. However, that character seeks to identify the most emarginate extremes, while this one deals with the other end of the spectrum, the most minimal emargination.

This character is used in Gaffney et al. (1991).

16. Parietal, parietal contacts pterygoid at base of processus trochlearis pterygoidei: absent ( Pelusios ) 5 0; ventral process of parietal reaches pterygoid on lateral side of sulcus palatinopterygoideus ( Bothremys ) 5 1.

Morphology: A ventral process of the parietal contacts the pterygoid lateral to the sulcus palatinopterygoideus near the base of the processus trochlearis pterygoidei (fig. 62).

Primitive condition: In Proganochelys and in all of the pleurodiran outgroups to the Bothremydidae , parietal-pterygoid contact is restricted to the processus inferior parietalis.

Homoplasy: This contact is not known to occur outside of the Bothremydidae . However, within the Bothremydidae it has arisen three times in the MPC. There is no indication of different morphologies for the separate occurrences.

Discussion: Unfortunately this character requires a well-preserved and well-prepared skull for its determination, and the inadequate preservation among the Taphrosphyini in particular limits its use. As known, however, it supports Bothremys + Chedighaii .

17. Parietal, sulcus palatinopterygoideus: absent ( Proganochelys ) 5 0; present and high ( Galianemys ) 5 1; present and low due to thick parietal and postorbital roof ( Bothremys ) 5 2.

Morphology: The sulcus palatinopterygoideus (figs. 23, 25, 62, 78) is the space found only in pleurodires between the fossa orbitalis and the fossa temporalis, with its medial wall formed by the processus inferior parietalis plus crista pterygoidea and its lateral wall formed by the processus trochlearis pterygoidei. We adopt the term ‘‘sulcus palatinopterygoideus’’ following Antunes and Broin (1988: figs. 3, 4, 8). However, they also used a synonym ‘‘gouttière ptérgoidienne’’. In Lapparent de Broin and Werner (1998) they used ‘‘pterygoid sulcus’’ as well as ‘‘sulcus palatinopterygoideus’’. Schumacher (1954: fig. 25; 1955a: figs. 1, 3) used the broader term ‘‘orbitale Pterygoideusrinne’’ or ‘‘orbitaler Pterygoideuskanalrinne’’ for this space in pleurodires. However, he also used the term for the dorsal surface of the pterygoid in cryptodires ( Schumacher, 1955a: fig. 6), so it is clearly not the restricted usage we employ here. The sulcus palatinopterygoideus is defined laterally by the processus trochlearis pterygoidei and is therefore found only in taxa with this structure, that is, pleurodires.

The two conditions of the sulcus recognized here are the more common: the relatively high sulcus (state 1, fig. 78) and the more restricted, relatively low condition (state 2, figs. 130, 279), formed by a thicker and lower parietal, with some participation of postorbital, as seen in Bothremys and Chedighaii hutchisoni .

Primitive condition: For turtles, the absence of this structure is primitive. Within the Pleurodira , the higher condition of the sulcus palatinopterygoideus is present in chelids, pelomedusids, and euraxemydids and can be considered primitive.

Homoplasy: There is no homoplasy in the analysis for the presence of the sulcus palatinopterygoideus, which is a pleurodiran synapomorphy (fig. 297). However, a low sulcus palatinopterygoideus also occurs in some Podocnemididae as well as within the Bothremydini (fig. 297). The conditions do not appear to be distinguishable morphologically.

Discussion: This character is not the same as simply having a low skull. Araripemys and Labrostochelys have very low skulls, but they have a relatively high sulci palatinopterygoideus. The degree of dorsal restriction of the sulcus is not clear. Rosasia could be interpreted as having a lower sulcus than that in Cearachelyini but higher than that in Bothremys . We have chosen to recognize only the more clearly identifiable extreme condition. Several key taxa, Araiochelys and Zolhafah , among Bothremydini , are not well enough preserved to code this character and that limits its usefulness.

18. Parietal, enters orbital margin: no ( Galianemys ) 5 0; yes ( Phosphatochelys ) 5 1.

Morphology: The dorsal plate of the parietal enters the posteromedial margin of the orbit in Rhothonemys , Ummulisani , and Phosphatochelys (figs. 196, 204, 208). In ventral view, the parietal forms part of the fossa nasalis roof.

Primitive condition: In all outgroups and in Euraxemydidae and Cearachelyini , the parietal does not enter the orbital margin.

Homoplasy: No homoplasy seen in the MPC.

Discussion: This character occurs only in three taxa in Pelomedusoides: Rhothonemys , Ummulisani , and Phosphatochelys . The shape of the parietal itself is very similar in the three taxa, but some of the surrounding elements differ. In Phosphatochelys and Ummulisani the frontal is very small (or absent) and does not enter the orbital margin due to a prefrontal-parietal contact. In Rhothonemys the frontal is larger, widely enters the orbital margin, and there is no prefrontal-parietal contact. However, the parietals of all three taxa also agree in being relatively farther forward in the skull roof than in any other Pelomedusoides, so that they form a significant part of the fossa orbitalis roof, also in contrast to other Pelomedusoides.

19. Supratemporal, supratemporal bone: present ( Proganochelys ) 5 0; absent ( Pelusios ) 5 1.

See Gaffney (1990), Gaffney et al. (1991), and Rougier et al. (1995).

20. Jugal, jugal retracted from orbital margin: jugal enters orbit ( Pelusios ) 5 0; jugal slightly retracted ( Cearachelys ) 5 1; jugal widely retracted ( Galianemys ) 5 2.

Morphology: In Cearachelys (figs. 71E, 78) the postorbital has a long ventral process extending along the orbital margin that nearly meets the maxilla ( Cearachelys, Postorbital ). There is some variation in degree and preservation, but this condition (state 1) occurs in all three Cearachelys skulls. In the two species of Galianemys (figs. 81, 89), the postorbital and maxilla widely meet, completely separating the jugal from the orbital margin (state 2).

Primitive condition: Substantial jugal exposure in the orbital margin is clearly the primitive condition, as seen in Proganochelys ( Gaffney, 1990: fig. 16) and nearly all pleurodires.

Homoplasy: This feature appears to have occurred independently in several lineages of cryptodires, but there is no indication of homoplasy within the Pleurodira (fig. 298).

Discussion: This character is run ordered in the preferred analysis, with state 1, partial retraction, the primitive state relative to state 2, complete retraction. The additive nature of this character could also be expressed as two states only: no retraction, and at least partial jugal retraction present. When the analysis is run with the three states ordered, the character unites the Cearachelyini (with the partial condition) and the genus Galianemys (with full retraction). When the character is run unordered, the same MPC results (fig. 298).

21. Jugal, jugal narrow dorsoventrally: absent ( Bothremys ) 5 0; present ( Taphrosphys , Labrostochelys ) 5 1.

Morphology: In Taphrosphys and Labrostochelys (figs. 178, 185, 190) the lateral exposure of the jugal is long horizontally and short vertically. Although there is a contribution to the orbital margin, the height of the bone is significantly less than in other Taphrosphyini and in turtles in general.

Primitive condition: The jugal in Proganochelys , most chelids, pelomedusids, Araripemys , euraxemydids, and podocnemidids has a large exposure on the cheek.

Homoplasy: Reduced jugal exposure on the cheek also occurs in some chelids with well-developed cheek emargination, such as Hydromedusa and Platemys , and in some testudinoids ( Cuora , Melanochelys , Rhinoclemmys , Terrapene , Gopherus , Kinixys , and Psammobates ). Within the Pelomedusoides, there is no evidence of homoplasy. In the MPC, the character occurs independently in Labrostochelys and Taphrosphys .

Discussion: The character occurs only in members of the Taphrosphyini . It unites two species of Taphrosphys , T. ippolitoi and T. congolensis ; it is not known in T. sulcatus .

22. Jugal, jugal-quadrate contact: absent ( Bothremys ) 5 0; present ( Azabbaremys ) 5 1.

Morphology: In the Taphrosphyini the cheek is relatively short with a dorsally retracted quadratojugal, allowing the jugal to contact the large quadrate (figs. 178, 185, 196).

Primitive condition: Although the sutures in this area are not definite in Proganochelys , contacts observed in most Casichelydia, including primitive cryptodires and pleurodires, show that separation of these elements by the quadratojugal is the primitive condition for Pleurodira .

Homoplasy: Although there is no homoplasy in the Bothremydidae , this feature occurs within the Podocnemididae in Peltocephalus and Erymnochelys . It occurs with a large jugal crossing the area where cheek emargination normally occurs; the quadratojugal is not reduced in size. In the Taphrosphyini with this contact, the quadratojugal is small and dorsally located. Jugal-quadrate contact also occurs in the cryptodire Archelon .

Discussion: The jugal-quadrate contact is a diagnostic character of the Taphrosphyini in the MPC, but the character is determinable with certainty in only 5 of 11 taxa, although Labrostochelys probably has it. The contact also differs among the known taxa. In Azabbaremys it is a wide contact with both bones relatively thick. In Phosphatochelys the bones are very thin and the contact could almost be kinetic.

23. Jugal, exposure on triturating surface: none ( Pelusios ) 5 0; large exposure visible in ventral view ( Bothremys ) 5 1.

Morphology: The upper triturating surface in turtles is made up mostly by the maxilla with an anterior contribution by the premaxilla. In many forms there is also some contribution posteriorly or posteromedially from the palatine. However, contribution by the jugals to the triturating surface is rare. Character state 1 is the ventral exposure of the jugal on the triturating surface in palatal view (figs. 10, 130). A small amount of jugal may be exposed, as in Cearachelys (fig. 77), but this is not the character as delimited here.

Primitive condition: In Proganochelys , all primitive cryptodires, chelids, pelomedusids, euraxemydids, and podocnemidids the jugals are excluded from the triturating surface.

Homoplasy: The MPC shows this character as a synapomorphy of the Bothremydina , with a reversal in Chedighaii . The character also occurs independently in the Taphrosphyini in CNRST-SUNY 199, an undescribed form from Mali.

The only taxon outside of the Bothremydidae in which jugal participation in the triturating surface is known is Sandownia ( Meylan et al., 2000) and a probably related undescribed form from the Paleocene of Morocco. In Sandownia jugal participation is long and narrow and lateral to the maxilla and pterygoid. There is no jugal-palatine contact on the triturating surface as in bothremydids.

Discussion: Within the Cearachelyini , a slightly exposed jugal occurs in Cearachelys and one species of Galianemys , G. whitei , but not in the other species, G. emringeri . The three skulls of Cearachelys (fig. 77) show some variation in the degree of jugal exposure, suggesting that this character is one of degree rather than the all or none used in the character coding. Nonetheless, we have coded the small exposure as ‘‘0’’ and only recognized the more extreme condition as the character.

This character is used in Meylan (1996), Lapparent de Broin and Werner (1998), and Tong et al. (1998).

24. Squamosal, posterior projection: absent ( Galianemys ) 5 0; present, projecting posteriorly, forming distinct process ( Bothremys ) 5 1.

Morphology: In the Bothremydini and some Taphrosphyini , the squamosal forms a posterior process that extends well posteri- or to the opisthotic (figs. 7, 101, 102, 122, 123).

Primitive condition: In the outgroups, either there is no projection ( Proganochelys , euraxemydids) or the opisthotic projects no farther than the squamosal (chelids, pelomedusids). In most Bothremydidae (except Bothremydini and Labrostochelys and CNRST-SUNY 199), the squamosal is even with or projects only a short distance beyond the opisthotic.

Homoplasy: This character occurs in Labrostochelys and CNRST-SUNY 199 outside the Bothremydini , although it is longer and deeper in Labrostochelys , consistent with the hypothesis that it is nonhomologous.

Discussion: The character is a synapomorphy for the Bothremydini but remains uncertain in several taxa for which this part of the skull is poorly known.

Another character, character 102, processus paroccipitalis not projecting beyond squamosal, is distinct from this one. In the opisthotic character, the processus paroccipitalis is very similar in all Bothremydidae ; it is relatively small compared to those in outgroups like pelomedusids. The squamosal projection is present with a small opisthotic.

25. Squamosal, posteroventral vertical flange: absent ( Galianemys ) 5 0; present ( Labrostochelys ) 5 1.

Morphology: The squamosal forms the posterolateral corner of the skull posterior to the quadrate. It is generally a cone-shaped bone with the hollow of the cone contributing to the antrum postoticum if one is present. In Proganochelys the antrum postoticum is absent and the squamosal is a simple curved plate with a nearly flat ventral surface. When the character is present, the flange (figs. 168, 169, 172, 177–209, 287) is a thin sheet of bone running anteroposteriorly on the ventral surface of the squamosal. It appears to be related to the attachment of the depressor mandibulae musculature.

Primitive condition: The squamosal flange is absent in Proganochelys and all outgroups.

Homoplasy: None observed (fig. 299).

Discussion: The ventral flange developed as a vertical plate on the squamosal is unique to a monophyletic group (fig. 299) within the subtribe Taphrosphyina : Taphrosphys , Labrostochelys , Ummulisani , Rhothonemys , and Phosphatochelys .

26. Squamosal, lateral tubercle: absent ( Galianemys ) 5 0; present ( Labrostochelys ) 5 1.

Morphology: At the posterodorsal edge of the cavum tympani in Taphrosphys , Labrostochelys , Ummulisani , and Phosphatochelys is a small tubercle produced by the squamosal that is directed laterally and ventrally along the suture with the quadrate (fig. 287). The tubercle may be at the posterior edge of the scale covering the cavum tympani. Posterior and ventral to the tubercle, the bone trends medially and was presumably covered by some part of the depressor mandibulae.

Primitive condition: The tubercle is absent in all outgroups.

Homoplasy: None apparent.

Discussion: The function of this peculiar structure is unknown. It appears to mark the division between the scaled surface and the muscular covering.

27. Postorbital, fossa orbitalis posterior enlargement: absent ( Galianemys ) 5 0; present ( Bothremys ) 5 1.

Morphology: The fossa orbitalis of bothremydids is well defined by bone: the postorbital and jugal posteriorly, the postorbital and parietal dorsally, and the maxilla and palatine ventrally. In many bothremydids, the internal fossa is much larger than the orbital rim, and some of the bones form pockets or concavities that seem to be significantly larger than a spherical eyeball would require (fig. 144). Although there is some variation in the extent of this enlargement and which bones form it, we use a single character state to treat it. All of the taxa that have the character have a postorbital forming the septum orbitotemporale distinctly posterior to the orbital rim (figs. 211, 278), and some have a more pronounced ventral concavity formed by the maxilla (fig. 211) as well. Even those Taphrosphyini that have a small or mostly absent septum orbitotemporale have ridges on the parietal (fig. 211), which indicate the position of the posterior limits of the fossa orbitalis.

Primitive condition: The pelomedusids (fig. 25) and chelids ( Gaffney, 1979a: fig. 55) show the primitive condition of a fossa orbitalis close to the size of the orbital rim.

Homoplasy: The condition may be a synapomorphy for the Bothremydidae with loss in Cearachelys , and within the Taphrosphyini in the MPC.

Discussion: The identification of this character in Chedighaii barberi is based on the whole skull endocast YPM PU 12951 (figs. 166, 167), also described by Gaffney (1977b).

The principal difficulty with this character is that it may be related to relative skull/orbit size. Skulls with relatively small orbital openings (e.g., Bothremys ) may have a larger fossa orbitalis than skulls with relatively large orbits (e.g., Phosphatochelys ). Nonetheless, identification of the character is based on the position of the ridges, particularly parietal and postorbital. Therefore, both Phosphatochelys and Bothremys are identified as having this character.

28. Postorbital, septum orbitotemporale: absent ( Proganochelys ) 5 0; postorbital wall present ( Galianemys ) 5 1; postorbital wall at least partially open ( Phosphatochelys ) 5 2.

Morphology: In pleurodires the posterior wall of the fossa orbitalis formed mostly by the jugal, palatine, and postorbital is a distinct structure (state 1), the septum orbitotemporale (figs. 23, 25, 43, 78). Proganochelys and cryptodires lack this wall (state 0). In some Taphrosphyini ( T. congolensis , Phosphatochelys , Azabbaremys ) this wall is reduced and mostly open (state 2; fig. 300). In this state (fig. 279), the ventral part of the postorbital is missing, lacking the palatine contact, and much of the jugal vertical component is reduced. The sulcus palatinopterygoideus is not defined laterally except for the pterygoid portion of the processus trochlearis pterygoidei. The partially open condition of this wall in the Taphrosphyini is a different condition from that in Proganochelys and cryptodires, as a portion of the postorbital wall and the sulcus palatinopterygoideus is still determinable in every pleurodire.

Primitive condition: For pleurodires, a well-developed septum orbitotemporale is primitive (fig. 300). The septum orbitotemporale is part of the complex forming the processus trochlearis pterygoidei and the sulcus palatinopterygoideus (figs. 23, 25), all pleurodiran synapomorphies.

Homoplasy: None known. Some cryptodires (e.g., trionychids; Gaffney, 1979a: fig. 144) may develop a partial wall along the posterior orbital rim, but it is clearly morphologically distinct from the pleurodiran condition.

29. Postorbital, size: relatively short ( Euraxemys ) 5 0; relatively long ( Galianemys ) 5 1.

Morphology: Considering the simple nature of the character definitions and the relatively large amount of variation in pleurodire skull roofs, this character may seem at first to be useless. However, there is an objective kernel in all this corn. The short postorbital, either entering the temporal margin (fig. 6B, F) or being bordered by the quadratojugal-parietal contact (fig. 6D, E), appears qualitatively distinguishable from the long, narrow postorbital of the Cearachelyini , Bothremydini , and Azabbaremys (figs. 6H–J, 7A, B, E, H–J, 8J).

Primitive condition: The short postorbital occurs in the outgroups Proganochelys , chelids, pelomedusids, and Araripemys . Primitively, the postorbital was both short and not exposed by emargination. Exposure due to temporal emargination is treated elsewhere.

Homoplasy: The longer than wide postorbital (state 1) appears only once within the Bothremydidae , but it is reversed within the Taphrosphyini for the Taphrosphys -Labrostochelys-Ummulisani-Phosphatochelys group. However, it also occurs in the Cryptodira .

Discussion: Interestingly, the long, narrow postorbital within Bothremydidae occurs along with a well-roofed skull lacking an extensive temporal emargination. The Kurmademydini have a short postorbital, because they are extensively emarginated. One might speculate that this was the primitive Bothremydidae condition and that the elongate postorbital is correlated with the presumably re-evolved roofed skull found in all other bothremydids.

This character is used in Meylan (1996), Tong et al. (1998), and Lapparent de Broin (2000b).

30. Premaxilla, protrudes anteriorly beyond labial ridge: no, or slightly ( Galianemys ) 5 0; yes, in ventral view projects anteriorly ( Bothremys ) 5 1

Morphology: The premaxilla, seen in ventral view, projects anteriorly beyond the labial ridge (figs. 10E, G–I, 11F, G, J).

Primitive condition: A vertical or slightly inclined wall of the premaxilla between the labial ridge and the apertura narium externa is present in Proganochelys ( Gaffney, 1990: fig. 27) and nearly all turtles and represents the primitive condition.

Homoplasy: This character has a CI of 0.20, showing that it has a lot of homoplasy. However, there are three independent origins within the Taphrosphyini alone and only one in the Bothremydini . In the Bothremydini the premaxilla and maxillae have similar anterior protrusions beyond the labial ridge of the triturating surface. In Labrostochelys there is a long triangular projection of the premaxilla that is unlike any morphology known in any fossil or living turtle. Other Taphrosphyini have thick premaxillae, different from those in the Bothremydini . What is treated as one character could be argued to be different character states rather than one.

Discussion: Although Chedighaii hutchisoni completely lacks the premaxillae, the receded prefrontals and inclined maxillae strongly suggest that the premaxilla was protruding. We have coded this taxon as ‘‘?’’ anyway, but it is nice to see that the MPC has decided Chedighaii should have a protruded premaxilla.

31. Premaxilla, midline depression: absent, shallow, or indistinct ( Euraxemys ) 5 0; distinct and wide ( Bothremys kellyi ) 5 1; distinct and narrow ( Rosasia ) 5 2.

Morphology: When the lingual ridge of the triturating surface is well defined and extends onto the premaxillae, a midline depression or concavity between the pair of ridges is sometimes formed on the triturating surface of the premaxilla. In Bothremydidae this depression is usually deep and well defined. It may be very broad, widening posteriorly as in Kurmademys , Zolhafah , Bothremys kellyi , and B. arabicus (state 1, figs. 9F, 10C, F, I) or it may be narrow, with essentially parallel sides as in Bothremys cooki , B. maghrebiana , Rosasia , and Araiochelys (state 2, fig. 10D, E, G, H).

Primitive condition: In Proganochelys the lingual ridge does not extend up onto the triturating surface of the premaxilla. This is also the case in many pleurodires, including chelids, pelomedusids, Araripemys , euraxemydids, and most podocnemidids (a lingual ridge is present in Erymnochelys and Peltocephalus ). These premaxillary depressions are generally absent in cryptodires. Similar depressions are seen only in cryptodires with multiple ridges on the triturating surface (e.g., Meiolania , Geochelone , Kachuga ) in which they are formed at least in part by lateral accessory ridges of the triturating surface.

Homoplasy: In the MPC the deep, wide midline concavity (state 1) occurs as a synapomorphy for Bothremydidae , but there are a number of reversals. State 1 is lost within the Bothremydini , but re-evolves in Chedighaii and in Bothremys arabicus and B. kellyi . There is also a reversal of the wide condition within Taphrosphyini where Azabbaremys loses the concavity (state 0) and Labrostochelys acquires a narrow one (state 2).

Discussion: The extremes of this character, states 1 and 2, are easily recognized, but because it is a gradational character, intermediate stages are a problem. As with a number of these more subjective, gradational characters, the senior author has chosen to try recognizing just the most extreme conditions as states and to ignore the variation between them. However, this can be considered too subjective in some cases. Deleting this character from the analysis has a significant effect on the MPC, so this issue is important. When this character is deleted, the result is a loss of resolution for the four species in Bothremys , which become a multichotomy with Chedighaii , and a loss of resolution for Zolhafah and Rosasia , which become a multichotomy with the remaining Bothremydina . These are obviously poorly supported nodes and it is worth looking at the character in more detail. The taxa that it resolves in Bothremys have the character states clearly recognized. B. kellyi and B. arabicus have distinctly wide median depressions, and B. cooki and B. maghrebiana do have distinctly narrow ones. It is outside the Bothremydina that coding the depressions becomes more subjective. Rather than delete the character entirely from the dataset, it seems better to use it in the analysis and present the alternative cladogram (fig. 290).

The midline space between the lateral triturating surfaces seen in state 1 is reduced in state 2, perhaps by increasing durophagy within the clades in which state 2 occurs.

This character is used in Antunes and Broin (1988) and is probably equivalent to the ‘‘anterior palatine sulcus’’ of Lapparent de Broin and Werner (1998).

32. Premaxilla, midline dorsal process: present, meeting nasals ( Proganochelys ) 5 0; absent or low ( Galianemys ) 5 1; present, at least partially separating nares ( Araiochelys ) 5 2.

Morphology: In Bothremys , Araiochelys , Rhothonemys , and Labrostochelys the apertura narium externa is at least partially divided by a dorsal process, half formed by each premaxilla, which rises from the lower margin of the apertura (figs. 130, 141, 143). In Araiochelys (fig. 125D) this almost completely divides the apertura.

Primitive condition: Proganochelys has a dorsal premaxillary process, as in many other amniotes, which completely divides the apertura narium externa. However, chelids, pelomedusids, euraxemydids, Araripemys , podocnemidids, Cearachelyini , and Kurmademydini all lack one, and this is the presumed primitive condition for Pleurodira .

Homoplasy: The dorsal premaxillary process is lost once in the MPC, at the Casichelydia node, and reversed twice, within the tribes Bothremydini and within Taphrosphyini .

Discussion: An earlier version of this character was ‘‘figure eight shaped apertura narium externa’’, but this version may be a little more objective. There are many stories in the nose of pleurodires, and this is just one of them.

This character is used in Gaffney and Kitching (1995) and Rougier et al. (1995).

33. Premaxilla, dorsal sulcus: smooth surface ( Bothremys ) 5 0; sulcus parallel to margin of apertura narium externa ( Phosphatochelys ) 5 1.

Morphology: A narrow, parallel-sided trough occurs along the anterior edge of the apertura narium externa on the dorsal surface of the premaxilla in Taphrosphys congolensis , T. ippolitoi , and Phosphatochelys (fig. 279). The sulcus has a variable number of foramina in its floor that extend ventrally into the premaxilla. The sulcus presumably held an artery or vein.

Primitive condition: Neither Proganochelys nor any of the pleurodiran outgroups have this sulcus. The floor of the fossa nasalis in this area in various pleurodires may have a lip and scattered foramina, but not this well-defined sulcus.

Homoplasy: None known.

Discussion: This is a fairly ‘‘minor’’ character that intuitively seems particularly liable to individual variation, but it occurs only once in the MPC and does help define a group within the Taphrosphyina .

34. Maxilla, triturating surfaces: relatively narrow, parallel sides ( Taphrosphys ) 5 0; triangular, wider posteriorly than anteriorly ( Galianemys ) 5 1: triangular, very wide posteriorly ( Bothremys ) 5 2.

Morphology: The bothremydid tribes Cearachelyini , Kurmademydini , and Bothremydini all have wide triturating surfaces formed by maxilla, premaxilla, palatine, and sometimes jugal. They are expanded posteriorly to form a roughly triangular shape in ventral view in which the lingual ridge converges anteriorly toward the labial ridge. In state 1 the triangle is narrow posteriorly (figs. 9F, H–J, 10B, E). In state 2, the triangle is broader and the midline depression (Character 32) is narrower (fig. 10A, C, D, F, G, I, K). Some Taphrosphyini have relatively wide triturating surfaces ( Azabbaremys , Nigeremys ), but the lingual and labial ridges are parallel.

Primitive condition: Although there is some variation in triturating surface shape in the outgroups, all are narrower and more parallel-sided in comparison to bothremydids.

Homoplasy: With a CI of 0.33, this character has homoplasy problems. In the MPC (fig. 301), the derived condition of a wide triturating surface of any sort originates once at Bothremydidae and reverses once at Taphrosphyini . Within the Bothremydini , the very wide condition (state 2) is reversed three times or originates twice and is lost twice. Triangular, or at least very wide, triturating surfaces occur within Podocnemididae and in many cryptodire groups.

Discussion: It is possible to subdivide the wide condition into two states, because some Bothremydini have significantly wider palates ( Foxemys , Zolhafah , Rosasia , Bothremys , and Chedighaii hutchisoni ) than do other bothremydids. Comparing these widths using the width of posterior triturating surface/skull length as a ratio produces two possible ranges: 16–28 for state 1 and 31–40 for state 2. However, within Cearachelys the ratio for two specimens shows a fairly wide range of 21–28, suggesting that the role of individual variation could be considerable. The range of variation in recent broad-jawed taxa (some chelids show a wide range within one species) supports the suggestion that without a larger sample, subtle differences in triturating surface width would best be ignored. We thereforeadopt a fairly simplistic view of what is undoubtedly a more complex character (see also discussion under Araiochelys in the Systematics section for maxilla width in Araiochelys compared with Bothremys maghrebiana ). It might also be argued that these states should be ordered, and the MPC is consistent with this, given a few not unreasonable reversals (fig. 301). If ordered, the same MPC results.

This character, or a similar version of it, is used by Antunes and Broin (1988), Lapparent de Broin and Werner (1998), and Tong et al. (1998).

35. Maxilla, triturating surface pits: absent ( Kurmademys ) 5 0; present ( Bothremys ) 5 1.

Morphology: The paired conical depressions in the triturating surface are known in Bothremys , Zolhafah , Rosasia , and Araiochelys (figs.123, 128, 133, 137; see also text under Maxilla and Jugal for these taxa). The pit is formed mostly by the maxilla with a widely varying contribution from the jugal.

Primitive condition: The absence of triturating pits is primitive, as all pleurodires outside bothremydids lack them.

Homoplasy: A shallow pit is present in CNRST-SUNY 199 (fig. 302), an undescribed Taphrosphyini skull. Its morphology, however, is different from the pits in Bothremydini . Some Cearachelys have a shallow pit.

Outside pleurodires, a very similar paired set of triturating pits occurs in an undescribed cryptodire from the Paleocene of Morocco, represented by a series of skulls: AMNH 30001, AMNH 30558, and AMNH 30554.

Discussion: This character (fig. 302) is a synapomorphy for the subfamily Bothremydini , being absent in Chedighaii . Character 23 (jugal exposed in triturating surface) overlaps with this character in that the formation of the pit exposes the jugal overlying the maxilla in the cheek. This exposure, however, varies widely in extent. The overlap is incomplete in that Cearachelys has some jugal exposure but no deep pit. In any case, deleting this character results in the same MPC.

36. Maxilla, accessory ridge on triturating surface: absent ( Galianemys ) 5 0; present ( Euraxemys ) 5 1.

Morphology: A ridge on the triturating surface between the lingual and labial ridges can be seen in the Euraxemydidae (figs. 42, 47). It also occurs in Sankuchemys (fig. 65). The ridge lies parallel to the lingual and labial ridges.

Primitive condition: Proganochelys , basal cryptodires, and most chelids and pelomedusids lack accessory ridges. Within chelids and pelomedusids, accessory ridges do pop up so to speak, but it is unlikely that this is primitive for these groups.

Homoplasy: Accessory ridges are frequent within the Podocnemididae (especially Podocnemis ), chelids, and pelomedusids, and they are also common within several groups of cryptodires. However, in the MPC accessory ridges are a synapomorphy for the Euraxemydidae . Within the Bothremydidae only Sankuchemys , one Foxemys , and the undescribed CNRST-SUNY 199 have one.

Discussion: The triturating surface of turtles has produced accessory ridges on multiple occasions within turtles. However, within the Pelomedusoides, distribution of this character proves useful and it is included in the dataset.

37. Maxilla, labial ridge depth below orbit: relatively shallow (roughly equal to or less than orbital diameter) ( Galianemys ) 5 0; very deep (greater than orbital diameter) ( Bothremys ) 5 1.

Morphology: In Bothremys and Chedighaii hutchisoni the maxilla ventral to the orbit is very deep (figs. 127, 130, 136, 146, 149) compared to Kurmademydini , Cearachelyini, and other Bothremydini .

Primitive condition: All outgroups have a relatively shallow labial ridge.

Homoplasy: In the MPC, this character has evolved independently within the Bothremydini and the Taphrosphyini . There is morphologic support for this in that the Taphrosphyini maxilla is a very thin sheet of maxilla in contrast to the thick, wedge-shaped maxilla of the Bothremydini .

Discussion: Chedighaii barberi is scored as ‘‘?’’ because the orbit is broken along most of the edges in Alabama 2001.2; however, the endocast YPM PU 12951 shows that the orbits were small, as in C. hutchisoni . The labial ridge in C. barberi is shallower than in C. hutchisoni , but it is still deep compared to other Bothremydini .

38. Maxilla, maxilla-quadratojugal contact: absent (quadratojugal present) ( Euraxemys ) 5 0; present ( Galianemys ) 5 1; absent (quadratojugal absent, chelids only) ( Emydura ) 5 2.

Morphology: The lateral face of the turtle skull is made up by the maxilla anteriorly, the jugal and postorbital posterior to the orbit, and the quadratojugal, quadrate, and squamosal posteriorly. In most turtles the jugal is positioned posteroventrally to the orbit and reaches the cheek margin ( Gaffney, 1990: fig. 16). In this position it prevents maxilla-quadratojugal contact. If the jugal is retracted from the cheek margin, the quadratojugal may meet the maxilla ventral to the jugal. This is the condition seen in the tribes Kurmademydini , Cearachelyini, and Bothremydini (figs. 3, 4). In the Taphrosphyini the quadratojugal is small and dorsal (fig. 5).

Primitive condition: Separation of the maxilla and quadratojugal by the jugal occurs in all the relevant outgroups (fig. 3).

Homoplasy: Maxilla-quadratojugal contact occurs in some cryptodires independently. Within the Bothremydidae the absence of a maxilla-quadratojugal contact is a Taphrosphyini synapomorphy. In this tribe the quadratojugal is retracted dorsally and lacks the ventral portion (see character 13), typically present in other Pelomedusoides. Therefore, there is morphological support for the nonhomology of the primitive condition in Taphrosphyini versus the pleurodiran outgroups outside Bothremydidae .

Discussion: Although the character is not determinable in a number of bothremydids, its distribution in the MPC shows it as a synapomorphy for Bothremydidae and its reversal as a synapomorphy for Taphrosphyini .

39. Maxilla, maxilla-quadrate contact, cheek emargination: absent, little or no emargination ( Galianemys ) 5 0; present, no emargination ( Azabbaremys ) 5 1; absent, barely separated by narrow fissure ( Phosphatochelys ) 5 2; absent, deep emargination, quadratojugal present ( Euraxemys ) 5 3; absent, deep emargination, quadratojugal absent ( Emydura ) 5 4.

Morphology: The lateral surface of the turtle skull is made up of the maxilla anteriorly, the jugal and postorbital posterior to the orbit, and the quadratojugal, quadrate, and squamosal posteriorly ( Gaffney, 1979a). In most turtles both the jugal and the quadratojugal reach the cheek margin ( Gaffney, 1990: fig. 16). In this position they prevent maxilla-quadrate contact. If both the jugal and quadratojugal are placed dorsally from the cheek margin, then the maxilla may contact the quadrate, and this is state 1 (fig. 4D, H). Within the Bothremydidae , an anteriorly elongate quadrate and an elongate maxilla meet to prevent jugal and quadratojugal exposure on the cheek margin. This contact is present in most Taphrosphyini , but there is some variation in form. In Azabbaremys (figs. 5J, 215) the contact results from a broad, posterior process of the maxilla. In Labrostochelys and Taphrosphys there is a narrow quadrate-maxilla contact, with an anterior process of the quadrate not seen in Azabbaremys or Phosphatochelys . Within the Bothremydini , Rosasia , Araiochelys , and Bothremys have a broad quadrate-maxilla contact.

State 2 (fig. 5D) is the narrow fissure present in Phosphatochelys and Ummulisani , which separates the quadrate and maxilla. This does not seem to be homologous with the deep emargination condition with the quadratojugal present, state 3 (fig. 3D). State 4 (fig. 3A) is the condition of a deep emargination with the quadratojugal entirely missing, known only in chelids.

Primitive condition: Intervention of the jugal and quadratojugal between the maxilla and quadrate occurs in all outgroups.

Homoplasy: We are unaware of any turtles outside of the Bothremydidae in which contact between the maxilla and quadrate occurs. The MPC shows the quadrate-maxilla contact occurring within the Taphrosphyini and within the Bothremydini (fig. 296A). Foxemys and Polysternon lack the contact and have the primitive condition for the quadratojugal, so it can be interpreted that the presence of the quadrate-maxilla contact is not primitive for Bothremydini . However, a large quadratojugal is present in Chedighaii hutchisoni , although not well preserved (and not determinable in C. barberi ), so presence of a maxilla-quadrate contact in the other Bothremydini is equivocal. The MPC is still consistent with a single origin of a maxilla-quadrate contact for the Bothremydodda ( Taphrosphyini + Bothremydini ) with reversals in the Foxemydina and Nigeremydina (see below).

Discussion: Character state 1 may not be homologous between the Taphrosphyini and the Bothremydini , due to different morphologies as well as ambiguity in the MPC (fig. 296A). Alternatively, it may be a synapomorphy for the tribes Bothremydini + Taphrosphyini and may be reversed in Foxemys , Polysternon , and Chedighaii . Character state 3 is a synapomorphy for Pelomedusoides and is lost in the subfamily Bothremydinae (tribes Cearchelyini + Bothremydini + Taphrosphyini ).

This character or a version of it is used in Antunes and Broin (1988).

40. Maxilla, orbitonarial bar width: roughly equal to or slightly less than diameter of orbit ( Galianemys ) 5 0; wider than orbit ( Bothremys ) 5 1; more than twice orbital diameter ( Labrostochelys ) 5 2; very narrow, much less than diameter of orbit ( Phosphatochelys ) 5 3.

Morphology: The bone between the orbit and the lateral margin of the apertura narium externa, the orbitonarial bar, is formed by the maxilla and prefrontal, but usually the maxilla forms most of it. The width of this bar varies, and this character is an attempt to use the extremes of this variation. The extremely narrow orbitonarial bar (state 3, fig. 5D) present in Phosphatochelys , Ummulisani , and Rhothonemys is unique in pleurodires. The Bothremydini have a wider bar than in most pleurodires, and Bothremys is the widest (state 1, fig. 4H). The Bothremys arabicus condition is unknown, however. Labrostochelys is unique among pleurodires in its extremely long snout and very wide orbitonarial bar (state 2, fig. 5G).

Primitive condition: The intermediate state (state 0) is fairly consistent among the outgroups, Proganochelys , chelids, and pelomedusids, but it is unlikely that all the intermediate states identified here are actually homologous.

Homoplasy: As restricted here, the states are not homoplastic in the MPC, but variation of the orbitonarial width is considerable throughout turtles.

Discussion: Although this character is not defined with precision, it seems to be the best way to obtain information from this morphology. Measuring some parameters might enhance the use of the character and allow more states to be distinguished, but there is a great deal of variation of this character in turtles and a very restricted usage seems best. Nonetheless, this is a relatively subjective character set and must be used with caution.

41. Maxilla, dorsal process onto skull roof: maxilla more lateral and ventral ( Galianemys ) 5 0; maxilla extending dorsomedially onto skull roof, restricting lateral extent of prefrontal ( Bothremys cooki ) 5 1.

Morphology: The dorsal process of the maxilla extends onto the skull roof in two Bothremys species, B. maghrebiana and B. cooki , restricting the prefrontal exposure in dorsal view (fig. 7G, H). The maxilla is more extensive medially in B. cooki than in B. maghrebiana , and in both species the prefrontal is more L-shaped in contrast to the more rectangular prefrontal of other pleurodires.

Primitive condition: Rectangular prefrontals and no dorsal extension of the maxilla are found throughout all outgroups.

Homoplasy: None apparent.

Discussion: Presumably the wide orbitonarial bar and very short snout in Bothremys are related to the dorsal extent of the maxilla.

42. Maxilla, ventral rim of orbit: rim with distinct margin ( Galianemys ) 5 0; rim absent, continuous slope ( Bothremys ) 5 1.

Morphology: In most turtles, the floor of the fossa orbitalis is nearly horizontal and the lateral surface of the maxilla is nearly vertical, with the two surfaces meeting at a distinct edge, often forming a ridge. In Bothremys cooki and B. maghrebiana the two surfaces are continuous without a distinct edge or break in slope, state 1 (figs. 129, 138).

Primitive condition: The outgroups, Proganochelys , chelids, pelomedusids, euraxemydids, and Cearachelyini , all have a low ridge marking the change in slope between the fossa orbitalis floor and the external surface of the maxilla.

Homoplasy: The absent rim of Bothremys cooki and B. maghrebiana is closely approached by Foxemys , which is coded as having this character. However, Foxemys does have more of a rim than does Bothremys , although the distinction is somewhat subjective. One specimen of B. maghrebiana, MHNL 20-268370, has a low rim (see discussion under Systematics, B. maghrebiana ), showing that there is some intraspecific variation of this character.

Discussion: Although it would seem likely that this character should be related to size of the fossa orbitalis, all Bothremys have a large fossa and only two of the species lack the rim.

43. Maxilla, exposure in orbital floor: maxilla broadly exposed ( Galianemys ) 5 0; maxilla narrowly or not exposed ( Azabbaremys ) 5 1.

Morphology: In most turtles the maxilla forms the major part of the floor of the fossa orbitalis. In Taphrosphys ippolitoi , Labrostochelys, CNRST-SUNY 199, and Azabbaremys the maxilla does not have an extensive medial process as in other turtles (fig. 280). The orbital floor in these forms is formed mostly by the palatine.

Primitive condition: The wide occurrence of a medial maxillary process in the outgroups shows it to be primitive for pleurodires.

Homoplasy: The character must be lost in Phosphatochelys , Arenila , and Ummulisani .

Discussion: As the character is only determinable in one of three Taphrosphys species, a new discovery could alter this distribution; as it is, this character is a subtribe Taphrosphyina synapomorphy.

44. Vomer, maxilla-vomer contact: present ( Proganochelys ) 5 0; absent ( Azabbaremys ) 5 1.

Morphology: The vomer is restricted to the premaxilla and allows premaxilla exposure on the apertura narium interna (fig. 215B). The premaxilla may be exposed on the apertura narium interna if there is no vomer-maxilla contact (figs. 65, 155) or if the vomer is simply absent (fig. 21). The character could be coded with a third state for this latter condition. This also has the effect of counting the vomer absent condition (character 45) twice. We have combined both conditions as one state.

Primitive condition: A vomer with a narrow anterolateral maxillary contact is present in Proganochelys , early cryptodires, and chelids, and it is presumed to be primitive for pleurodires. The absence of a contact would then be the derived state.

Homoplasy: This is a highly variable character with a CI of 0.16. Keeping this character demonstrates the complete objectivity of this work, and any decent amount of cooking would remove it. The loss and subsequent reappearance of the vomer is suspected within the Podocnemididae .

Discussion: The Kurmademydini and podocnemidids have this character, so it may be primitive at the level of the superfamily Podocnemidoidea within Pelomedusoides, although with multiple reversals. It also unites Azabbaremys and CNRST-SUNY 199. Due to the fragile nature of the vomer area, there are many missing data for taxa that otherwise have well-preserved skulls. Better material may eventually clarify the distribution of this character.

45. Vomer: present, paired ( Proganochelys ) 5 0; present, single ( Galianemys ) 5 1; absent ( Pelusios ) 5 2.

Morphology: The living Pelomedusoides are unusual among living turtles for lacking a vomer (its apparent presence in some living podocnemidids is interpreted here as a neomorph). However, the fossil record shows that the vomer was widely present in extinct Pelomedusoides. Pelusios and Pelomedusa lack a vomer but the condition in Araripemys is indeterminate, although it probably lacked a large, ‘‘normal’’ vomer. Hamadachelys , based on two specimens, has a vomer, and some extinct podocnemidids have well-developed vomers. Where determinable, all bothremydids have vomers.

Primitive condition: A vomer is primitive for pleurodires.

Homoplasy: Within Podocnemididae the vomer may be lost twice. It is lost independently in Pelomedusidae .

Discussion: In the MPC, the vomer loss does not define any groups, other than Pelomedusidae . It is very easy for the small, loosely attached vomer to be missing in fossil skulls and to fall out of recent ones. Nonetheless, representation within the Bothremydidae is good, and there is no indication of loss within the group.

This character is used in Gaffney et al. (1991), Lapparent de Broin and Werner (1998), Meylan (1996), and Rougier et al. (1995).

46. Vomer, vomerine teeth: present ( Proganochelys ) 5 0; absent ( Pelusios ) 5 1.

See Gaffney (1990), Gaffney et al. (1991), Gaffney and Kitching (1995), and Rougier et al. (1995).

47. Vomer, central bar: thin, sutured at both ends ( Proganochelys ) 5 0; thin, attached anteriorly only ( Azabbaremys ) 5 1; vomer absent ( Emydura ) 5 2.

Morphology: In Azabbaremys and CNRST-SUNY 199 the vomer tapers posteriorly and ends before reaching the palatine attachment (fig. 216).

Primitive condition: A narrow vomer, attached at both ends, occurs in the outgroups (fig. 9).

Homoplasy: None known.

Discussion: When first seen in Azabbaremys , the senior author thought this was due to preservational damage, but there is no indication of that, and CNRST-SUNY 199 has a very similar morphology with a clearly well-preserved vomer that is sutured anteriorly only.

48. Palatine, foramen palatinum posterius: in floor of orbit ( Chelydra ) 5 0; behind orbit, in floor of sulcus palatinopterygoideus ( Pelusios ) 5 1.

Morphology: The foramen palatinum posterius is described and figured in Gaffney (1979a: figs. 53–65) for a number of turtles. In cryptodires, the foramen lies in the floor of the fossa orbitalis, but in pleurodires it is separated from the fossa due to the presence of the septum orbitotemporale and the sulcus palatinopterygoideus (figs. 23. 24). The foramen is usually in the floor of the sulcus palatinopterygoideus.

Primitive condition: Proganochelys and cryptodires lack this character.

Homoplasy: None known.

Discussion: This character is synapomorphic at least for Eupleurodira, being unknown for shell-only taxa. A decently preserved skull is necessary for determining the character. The orbit or sulcus palatinopterygoideus needs to be visible; a ventral view of the foramen palatinum posterius is insufficient.

This character is used by Gaffney and Meylan (1988), Gaffney et al. (1991), and Fuente and Iturralde-Vinent (2001).

49. Palatine, dorsally arched palate: absent ( Galianemys ) 5 0; present ( Azabbaremys ) 5 1.

Morphology: The area between the lingual ridges of the triturating surfaces forms the roof of the mouth and the choanal passages. In Azabbaremys, CNRST-SUNY 199, Nigeremys , Phosphatochelys , and probably Rhothonemys , this area, particularly the part formed by the palatines posterior to the apertura narium interna, is an anterodorsally sloping surface that rises well above the level of the maxillary triturating surfaces (fig. 215C). The anterolateral margins of the apertura narium interna are relatively low, so the result is an apertura that is deeper than in other bothremydids.

Primitive condition: Although the outgroups and other pleurodires have many different shapes to the central palatal surface, all are shallower than in these taxa. A relatively flat palate is found in most relevant outgroups.

Homoplasy: There is a reversal in Labrostochelys ; otherwise, all Taphrosphyini that can be determined (only 5 out of 11) have the character present.

Discussion: Arenila is scored as ‘‘?’’ because the area is damaged, but the MPC suggests that it should have the deep choanal passage/ arched palate, based on the morphology that is present. The best skull of Taphrosphys , the type of T. ippolitoi , is damaged in this region, but the palatines as preserved have an anterodorsal slope, suggesting that it also has an arched palate. If this proves to be the case, this character will be a Taphrosphyini synapomorphy with one reversal.

50. Palatine, palatine contribution to triturating surface: little or none ( Euraxemys ) 5 0; moderate to extensive ( Bothremys ) 5 1.

Morphology: The upper triturating surface of turtles is usually made up mostly by the maxilla with a smaller contribution by the premaxilla anteromedially. The palatine is typically present just medial to the triturating surface and may also provide a small contribution. In many bothremydids, the palatine forms a significant part of the posteromedial part of the triturating surface (figs. 9F, 10H), and the maxilla-palatine suture is more anterolateral than in the primitive condition.

Primitive condition: The outgroups have little or no palatine contribution to the triturating surface.

Homoplasy: The character is a Bothremydidae synapomorphy, lost in the Taphrosphyini . The occurrence in Araripemys is probably independent on morphological grounds, as the triturating surface is narrow. The character is unclear in Sankuchemys ; it may be present but the specimen is ambiguous and has been coded ‘‘?’’.

Discussion: Although there is a gradational aspect to this character, we have tried to restrict it to the most extreme condition, particularly as seen in the Bothremydini . Nonetheless, some decisions about scoring are subjective. This character is generally correlated with the wide triturating surface in many bothremydids, but it is not a consistent correlation ( Araiochelys and Araripemys are exceptions).

This character is used in Meylan (1996) and Tong et al. (1998).

51. Quadrate, antrum postoticum: absent, open incisura columellae auris ( Proganochelys ) 5 0; small ( Bothremys ) 5 1; absent, closed incisura columellae auris ( Azabbaremys ) 5 2; moderate to large ( Galianemys ) 5 3.

Morphology: The antrum postoticum shows a wide range of variation in Pelomedusoides (figs. 176, 281–286). The antrum postoticum is smaller in many bothremydids than it is in other pleurodires. In Azabbaremys , Nigeremys , Ummulisani , and Arenila , it is completely filled with bone and no indication of the antrum is present in the cavum tympani (state 2, fig. 281C). This condition is clearly identifiable. The remaining taxa have a varying size for the antrum postoticum, and we have identified the small extreme (state 1, fig. 286D) and the moderate to large (state 3, fig. 281A, 282, 284) as character states. It is useful to compare figure 284C with figure 286D; both are broken, exposing the internal shape of the antrum postoticum.

The small condition, state 1, is found in Labrostochelys , Araiochelys , Bothremys maghrebiana , and Chedighaii barberi (it is indeterminate for the other species of Bothremys and Chedighaii ). However, within state 3 (moderate to large), Taphrosphys (fig. 176), for example, has an antrum that is larger than that in Labrostochelys but smaller than those in the Kurmademydini and Cearachelyini (figs. 283, 284). We have been unable to distinguish this degree of gradation because it seems to be too subjective.

Primitive condition: Proganochelys lacks an antrum postoticum, but a large antrum is apparently primitive for pleurodires as it occurs in chelids, pelomedusids (fig. 282A), and Araripemys (fig. 282C). However, euraxemydids and podocnemidids have a smaller antrum, as do the Kurmademydini and Cearachelyini . All these are coded within the ‘‘moderate to large’’ character state 3.

Homoplasy: Although the CI is a respectable 0.5, there is actually more homoplasy within bothremydids, which is hidden by the lumping of a lot of size variation into the ‘‘moderate to large’’ character state 3. It is difficult to divide this continuous variation objectively (except for absence) and is therefore hard to identify independent acquisitions morphologically. As defined here, the small condition (state 1) is acquired twice, once for Araiochelys + Bothremys + Chedighaii within the Bothremydini and once in the Taphrosphyini in Labrostochelys . The absent condition, state 2, appears independently three times, in Ummulisani , Azabbaremys , and Nigeremys .

Discussion: Although the size of the antrum postoticum varies widely in pleurodires, from very large to completely absent, it has been difficult to objectively divide this variation into discrete character states. The difference between ‘‘moderate’’ and ‘‘small’’ is subjective and based on what we perceive as a gap in antrum size. More specimens may alter this division. Attempts to measure the variation have been unsatisfactory but could be worth future efforts.

This character is used in Gaffney and Meylan (1988), Meylan (1996), and Tong et al. (1998).

52. Quadrate, incisura columellae auris: no posterior bony restrictions ( Euraxemys ) 5 0; eustachian tube and stapes separated by bone or a narrow fissure ( Foxemys ) 5 1; eustachian tube and stapes enclosed together by bone ( Podocnemis ) 5 2.

Morphology: The incisura columellae auris can be open to a varying degree (state 0, fig. f281A) or it can be closed by a meeting or near meeting of dorsal and ventral processes of the quadrate. In state 1, the incisura is closed, or nearly closed, separating the stapes and eustachian tube (fig. f281B, C). In state 2, the dorsal and ventral processes meet or nearly meet, posterior to the eustachian tube, enclosing both stapes and eustachian tube in the same oval opening ( Gaffney, 1979a: figs. 134, 140).

Primitive condition: The open incisura columellae auris of euraxemydids, early cryptodires, and Proganochelys is the presumed primitive condition for pleurodires. However, the apparently independent occurrence of state 2 in chelids and pelomedusids, as well as podocnemidids in the MPC (fig. 303), suggests an alternative that the enclosure of stapes and eustachian tube may be primitive for eupleurodires and lost in euraxemydids, Brasilemys , Araripemys , and Teneremys .

Homoplasy: None for state 1, but state 2 occurs three times independently (see above) in the MPC (fig. 303).

Discussion: The main objection to this character is that state 1 might be interpreted as redundant with character 53, incisura columellae auris completely closed to form a canal containing the stapes. The decision to make two characters using different morphologic criteria in this area is based on the open, but slitlike condition of the incisura in a few bothremydids, Cearachelys (fig. 283), Foxemys (fig. 281B), and Polysternon . Therefore, the two characters are not strictly redundant, and we use state 1 to try to reflect another aspect of the cavum tympani morphology.

The position of the eustachian tube in a fossil turtle is not, strictly speaking, determinable. The position of this soft tissue structure may not be indicated in bone. However, examination of thin sections and dissections of recent turtles by the senior author shows that its position and structure are relatively consistent in all turtles, particularly pleurodires. Although the eustachian tube in recent pleurodires does tend to take the shape of surrounding bones, it is unlikely that it would be thin and narrow enough to fit into the fissure or slitlike incisura columellae auris found in Cearachelys , Foxemys , and Polysternon . Its position in forms like the Euraxemydidae , with a wide incisura, is not determinable, but a distinct groove (the sulcus eustachii) showing its position is present on the posterior surface of the quadrate in bothremydids. Therefore, it is possible to argue that in bothremydids, the stapes and eustachian tube were separated by bone, either completely by bone or with a narrow fissure in the bone remaining, in contrast to all other pleurodires.

Although Araripemys is open posteriorly and we have coded it as ‘‘0’’, there is a small dorsal and ventral process (fig. 282) that almost completely closes the incisura. In euraxemydids the incisura is more open than in Araripemys : however, there is a partial restriction with small dorsal and ventral processes. Thus, euraxemydids (also coded as ‘‘0’’) could be coded separately from Araripemys as a different state, or Araripemys could be coded as ‘‘2’’. Using this approach, however, does not change the MPC.

A version of this character is used in Antunes and Broin (1988), Meylan (1996), Lapparent de Broin and Werner (1998), and Tong et al. (1998).

53. Quadrate, stapes contained in bony canal: stapes not completely contained in bone ( Euraxemys ) 5 0; stapes completely enclosed by bony incisura columellae auris ( Bothremys ) 5 1.

Morphology: The quadrate of most bothremydids completely surrounds the stapes, causing the incisura columellae auris to be a bony canal. This character is easily determined; that is, there are no cases of ambiguity. As can be seen in figures 176, 281, and 286, the canal is encased by bone on all sides, even in the Kurmademydini (fig. 282B) and Galianemys (fig. 284), which are more generalized in most characters than in other bothremydids. The narrow fissure condition seen in Foxemys (fig. 281B), Polysternon , and Cearachelys (fig. 283) is exclud- ed from this character.

Primitive condition: An open quadrate, with the incisura columellae auris only partially surrounding the stapes and not forming a canal, is found in all the relevant outgroups (fig. 282).

Homoplasy: Two reversals of this character occur in Cearachelys and the Foxemydina (fig. 304). The reversal is a synapomorphy for Foxemys + Polysternon . The morphology of the open incisura columellae auris in these bothremydids does differ from the open condition in outgroups like podocnemidids, pelomedusids, euraxemydids, and chelids (see character 52 for discussion). Lapparent de Broin (2000b) considered the Polysternon- Foxemys condition a reversal as well. A number of cryptodires have a completely closed incisura columellae auris ( Gaffney, 1979a, 1996), but they do not develop the bony canal as seen in bothremydids.

Discussion: The bony canal for the stapes in bothremydids is an unusual feature, synapomorphic at the level of Bothremydidae (fig. 304), and it is not found in any other pleurodires.

This character is used in Gaffney and Meylan (1988), Lapparent de Broin and Werner (1998), Tong et al. (1998), and Lapparent de Broin (2000b).

54. Quadrate, sulcus eustachii: without ventral process ( Bothremys ) 5 0; with ventral process overhanging dorsal margin ( Labrostochelys ) 5 1.

Morphology: The groove on the posterior surface of the quadrate, presumably marking the position of the eustachian tube (the sulcus eustachii), forms an indentation on the posterior margin of the cavum tympani. In Taphrosphys , Labrostochelys , and Phosphatochelys , the anterodorsal margin of the notch has a small, ventrally dependent process. In Labrostochelys (fig. 287) the process is narrow, but in Phosphatochelys (fig. 199) and Taphrosphys it is wider, forming a small, anteroposteriorly broad flange.

Primitive condition: All outgroups and other pleurodire taxa lack this process.

Homoplasy: None known. We know of no other turtle with this process.

Discussion: The small ventral process at the posterior edge of the cavum tympani, above the sulcus eustachii, seems to be related to some soft structure, probably the attachment of the eustachian tube, which enters the cavum tympani at this edge. The process is broken off in the Taphrosphys specimens, but its presence is determinable from the broken base.

55. Quadrate, trough on incisura columellae auris ridge: absent ( Bothremys ) 5 0; present ( Galianemys ) 5 1.

Morphology: When the incisura columellae auris is completely closed by bone, as in most bothremydids, there may be a ridge in the position across the surface of the cavum tympani, extending from the remnant of the incisura (the bony stapedial canal) to the sulcus eustachii. In Kurmademys (fig. 282B) and Galianemys (fig. 284) this ridge has a shallow groove along its lateral surface. This is thought to be a suture in Lapparent de Broin (2000b), but it is only a surface feature.

Primitive condition: A closed incisura columellae auris is absent in all the pleurodire outgroups, so none has a ridge or a trough. In the other bothremydids, there may be a low, poorly defined trough ( Labrostochelys ) but no distinct ridge.

Homoplasy: In the MPC, the character evolves once in Kurmademys and independently in Galianemys . However, the incisura is open in Cearachelys , and the character may be primitive for Bothremydidae and lost in Cearachelys and in the infrafamily Bothremydodda (the tribes Bothremydini + Taphrosphyini ).

Discussion: This character is a distinctive one for Galianemys . Although similar to the condition in Kurmademys , Galianemys has the distal ends of the groove separating, making it different from Kurmademys .

56. Quadrate, fossa precolumellaris: very small to absent ( Galianemys ) 5 0; present but shallow ( Euraxemys ) 5 1; deep and well defined ( Pelusios ) 5 2.

Morphology: Just anterior to the incisura columellae auris, where the quadrate forms the medial wall of the cavum tympani, many pleurodires have a depression, the fossa precolumellaris. Described and figured by Williams (1954b) in Podocnemididae (also in Gaffney, 1979a) as the precolumellar fossa, we have formalized it as the fossa precolumellaris. The distinction between shallow (state 1, fig. 281A) and deep (state 2, fig. 282A, B) is not as arbitrary as it might seem. The euraxemydid condition (state 1) shows a small dimple, which we think should be differentiated from the deep condition (state 2) seen in pelomedusids, chelids, Araripemys , and Hamadachelys .

Within Podocnemididae , the fossa varies in size, as described by Williams (1954b). All bothremydids except Kurmademys (it is indeterminate for Sankuchemys ) lack a fossa precolumellaris, so that the cavum tympani anterior to the incisura columellae auris is smooth (fig. 281B, C).

Primitive condition: The fossa precolumellaris is absent in Proganochelys and cryptodires, but a deep one is present in chelids and pelomedusids, the presumed primitive condition for pleurodires (a variable one is present in podocnemidids).

Homoplasy: Most bothremydids lack a fossa precolumellaris, and this is the primitive condition for turtles; however, this is a reversal and must have originated independently within bothremydids, as pleurodiran outgroups have a deep fossa.

Discussion: Kurmademys has a deep fossa precolumellaris, as in pleurodire outgroups Pelomedusidae and Chelidae . All other bothremydids lack the fossa, which is a synapomorphy for the subfamily Bothremydinae (the tribes Taphrosphyini + Bothremydini + Cearachelyini ).

This character, or one like it, is used by Meylan (1996), Lapparent de Broin and Werner (1998), Tong et al. (1998), and Lapparent de Broin (2000b).

57. Quadrate, shelf below cavum tympani: absent ( Galianemys ) 5 0; present, lower portion of cavum tympani unusually deep ( Bothremys ) 5 1.

Morphology: In the bothremydid tribes Bothremydini and Taphrosphyini the cavum tympani is recessed deeply into the quadrate, and its lower edge is deep, forming a sloping shelf along the ventral margin of the cavum (figs. 285, 286). In Labrostochelys and Phosphatochelys the cavum is not recessed, but the shelf persists as a ventrally sloping surface.

Primitive condition: All outgroups lack a ventral shelf on the quadrate.

Homoplasy: None known.

Discussion: The principal problem with this character is variation in the slope of the shelf within the tribe Taphrosphyini ; howev- er, scoring Labrostochelys and Phosphatochelys as absent produces the same MPC.

58. Quadrate, medial process contacting braincase elements and underlying cranioquadrate space: absent ( Chelydra ) 5 0; present ( Pelusios ) 5 1.

Morphology: This character is figured and described in Gaffney (1975 b, 1979a). The quadrate of pleurodires has a medial process that contacts the prootic and basisphenoid, although on the ventrally exposed surface, these contacts are not always exposed. The basisphenoid itself may cover the prootic or the prootic may contact the pterygoid. The medial process underlies the cranioquadrate space, which in the adult is the canalis cavernosus.

Primitive condition: The character is absent in Proganochelys and Cryptodira .

Homoplasy: None known.

Discussion: This character is a synapomorphy for Pleurodira . This character is used in Gaffney and Meylan (1988), Gaffney et al. (1991), and Fuente and Iturralde-Vinent (2001).

59. Quadrate, quadrate-basioccipital contact: absent ( Euraxemys ) 5 0; present ( Galianemys ) 5 1.

Morphology: The medial process of the quadrate in the Podocnemididae and Bothremydidae is larger and more extensive posteromedially than in other pleurodires (fig. 9H–J). The shape of the basioccipital is not very different in pleurodires with respect to this character; it is primarily a difference in quadrate shape that causes the contact. The quadrate covers the processus interfenestralis of the opisthotic and the prootic ventrally, resulting in a significant difference in the basicranium between podocnemidoids (podocnemidids + bothremydids) and all other pleurodires.

Primitive condition: Proganochelys and cryptodires completely lack the medial process of the quadrate ( Gaffney, 1975 b, 1979a). Within pleurodires, a smaller medial quadrate process is present in chelids, pelomedusids, and Araripemys ; as they lack a basioccipital-quadrate contact, the prootic is exposed. This appears to be the primitive condition for pleurodires.

Homoplasy: None known.

Discussion: This character supports the monophyly of the families Podocnemididae + Bothremydidae as the superfamily Podocnemidoidea and represents a major difference between their basicrania and those in all other pleurodires.

This character is in Antunes and Broin (1988), Meylan (1996), Lapparent de Broin and Werner (1998), and Tong et al. (1998).

60. Quadrate, condylus mandibularis position: near or in line with the basioccipital-basisphenoid suture ( Galianemys ) 5 0; distinctly anterior to plane of basioccipital-basisphenoid suture ( Pelusios ) 5 1; distinctly posterior to condylus occipitalis ( Nigeremys ) 5 2.

Morphology: This character is an attempt to use the variation in position of the condylus mandibularis relative to the elements of the basicranium. Although this variation is somewhat continuous, it is apparent that most bothremydids, compared to other pleurodires, have the condylus mandibularis far posterior, relative to the condylus occipitalis. This may also be described as a shortened or telescoped basicranial and otic region ( Lapparent de Broin and Werner, 1998). The character has been determined by drawing a line between each condylus mandibularis and seeing where the basioccipital-basisphenoid suture falls with respect to this line. If the line falls slightly behind, on, or slightly anterior to the suture, it is character state ‘‘0’’ (fig. 9H). If the line is well anterior to the basioccipital-basisphenoid suture, that is, well onto or anterior to the basisphenoid, it is character state ‘‘1’’ (fig. 9B). In Nigeremys and Arenila the line across the condylus mandibularis falls so far behind the basioccipital-basisphenoid suture that it is posterior to the condylus occipitalis, and this is character state ‘‘2’’ (fig. 11H, I).

Primitive condition: Proganochelys and early cryptodires have state ‘‘0’’, the primitive state at the level of all turtles. However, as all the pleurodire outgroups have state ‘‘1’’, this seems to be primitive for pleurodires.

Homoplasy: This character has a CI of 0.33, showing that there is a lot of homoplasy. Bothremydids are characterized by a reversal to state ‘‘0’’ at the level of the subfamily Bothremydinae (consisting of the tribes Cearachelyini , Bothremydini , and Taphrosphyini ). Within this group a number of the Taphrosphyini have a reversal to state ‘‘1’’ independently of Polysternon , which is also coded ‘‘1’’. However, this character is a relatively simplified summary of a lot of complex morphology, making more detailed comparisons to test homoplasy inconclusive.

Discussion: The difficulty with this character is scoring intermediate taxa (e.g., Azabbaremys and Labrostochelys ) that have a condylus mandibularis just anterior to the basioccipital-basisphenoid suture. We have scored both as ‘‘0’’ because the condylus mandibularis line falls close to the suture, although they could be scored ‘‘1’’ because the line is just anterior to the suture. We have run the dataset both ways with the same MPC resulting.

A similar character is used in Lapparent de Broin and Werner (1998).

61. Quadrate, fully formed cavum tympani: no ( Proganochelys ) 5 0; yes ( Pelusios ) 5 1.

See Gaffney (1990), Gaffney and Kitching (1995), and Rougier et al. (1995).

62. Quadrate, cavum tympani with acute posterior edge: no ( Proganochelys ) 5 0; acute edge, also enclosing stapes ( Pelusios ) 5 1.

See Gaffney (1990), Gaffney and Kitching (1995), and Rougier et al. (1995).

63. Quadrate, middle ear with complete lateral wall: no ( Proganochelys ) 5 0; yes ( Pelusios ) 5 1.

See Gaffney (1990), Gaffney et al. (1991), Gaffney and Kitching (1995), and Rougier et al. (1995).

64. Quadrate, cavum tympani curved dorsally: no ( Proganochelys ) 5 0; yes ( Palaeochersis ) 5 1.

See Gaffney (1990), Gaffney and Kitching (1995), and Rougier et al. (1995).

65. Quadrate, covers opisthotic laterally: no ( Proganochelys ) 5 0; yes ( Australochelys ) 5 1.

See Gaffney (1990), Gaffney et al. (1991), Gaffney and Kitching (1995), and Rougier et al. (1995).

66. Quadrate, pocket for stapes articulation: present ( Proganochelys ) 5 0; absent ( Australochelys ) 5 1.

See Gaffney (1990), Gaffney and Kitching (1995), and Rougier et al. (1995).

67. Quadrate, cranioquadrate space: relatively open ( Proganochelys ) 5 0; well-defined canal ( Australochelys ) 5 1.

See Gaffney (1990), Gaffney and Kitching (1995), and Rougier et al. (1995).

68. Pterygoid, fossa pterygoidea: absent or small ( Pelusios ) 5 0; moderate ( Galianemys whitei ) 5 1; deep and narrow ( Foxemys ) 5 2.

Morphology: The fossa pterygoidea in bothremydids is an anteromedial–posterolaterally elongate depression formed by the pterygoid, quadrate, and basisphenoid (figs. 101, 102). The foramen posterius canalis carotici interni, the foramen nervi facialis, and the foramen nervi vidiani may open within this depression. The prootic may be exposed in the deepest part of the depression in some forms ( Kurmademys , fig. 276E; Galianemys emringeri , fig. 277E).

The distinction between the two states of fossa development is not arbitrary because there is a gap in the size variation. Kurmademys (fig. 55), Cearachelys (two of three specimens) (fig. 74), Galianemys emringeri (fig. 82), and Rosasia (fig. 119) have a fossa pterygoidea that is not as deep as it is wide. The margin is recessed and the center is relatively shallow (state 1). In Foxemys (fig. 102), Polysternon (fig. 109), Nigeremys , and Arenila (fig. 225), the fossa pterygoidea is about as deep as it is wide, particularly at its center (state 2).

Lapparent de Broin and Werner (1998) and Lapparent de Broin (2000b) treated this structure as homologous to what we call the cavum pterygoidei (here restricted to the epifamily Podocnemidinura, which is the

Podocnemididae + Hamadachelys + Brasilemys ) and referred to this depression as the ‘‘podocnemidid fossa’’. They asserted that it becomes enlarged to form ‘‘the true enlarged carotid canal of the Podocnemididae’’ ( Lapparent de Broin, 2000b: 69). However, unlike the cavum pterygoidei of podocnemidids, there is no bony covering beneath the fossa pterygoidea and it never ends anteriorly in a single large opening, both features of the cavum pterygoidei. Thus, we treat the fossa pterygoidea as an independent structure from the cavum pterygoidei. Furthermore, the cladogram shows no homologous relationship between these concavities; they are both derived independently.

Primitive condition: Depressions of this kind are absent in the pterygoids of chelids, pelomedusids, Araripemys and euraxemydids. The absence of a fossa pterygoidea is the primitive condition.

Homoplasy: Lapparent de Broin (2000b) argued that the cavum pterygoidei of podocnemidids is homologous to the fossa pterygoidea of bothremydids and that the two are derived from the ‘‘podocnemidoid fossa’’ of Brasilemys . If this is correct, then the depression has occurred one time in the Pelomedusoides and been enlarged in the Podocnemididae and reduced and lost within most of the Bothremydidae . However, the distribution of the fossa in the MPC (fig. 305) shows that the fossa has originated five times within the Bothremydidae , all independently of the Podocnemididae . A similar but groovelike fossa occurs on the pterygoid of cheloniids ( Gaffney, 1979a: fig. 210).

The CI of this character in the MPC is 0.33, reflecting the high degree of homoplasy present in this character. A moderately developed fossa pterygoidea (state 1) originated three times and the deep fossa (state 2) three times within the Bothremydidae . There is no evidence from the cladogram (fig. 305) that the two states should be treated as additive, even though this might be expected from the morphology, that is, moderate (state 1) and deep (state 2).

Discussion: Although the fossa pterygoidea is a prominent feature of the palate, it is too sporadic in distribution to be of much phylogenetic significance. The deep fossa unites ( Foxemys , Polysternon ) with ( Nigeremys , Arenila ), but the moderate condition seems to have arisen independently in each case. The fossa seems to be the attachment site of a portion of the pterygoideus muscle.

A form of this character is used in Antunes and Broin (1988), Meylan (1996), Lapparent de Broin and Werner (1998), Tong et al. (1998), and Lapparent de Broin (2000b).

69. Pterygoid, cavum pterygoidei: absent ( Pelusios ) 5 0; present ( Podocnemis ) 5 1.

Morphology: Used by many previous authors, the cavum pterygoidei is a more formalized name for the ‘‘pterygoideus muscle chamber’’ or ‘‘enlarged carotid channel’’ of Gaffney (1979a: fig. 86). This is a relatively large opening from the palate into the braincase located at the posterior end of the pterygoid that contains a subdivision of the pterygoideus muscle ( Schumacher, 1954, 1955a, 1955b, 1973). It is differentiated from the fossa pterygoidea, character 68, by having at least a partial covering ventrally and an anteromedial opening into the braincase.

Primitive condition: All outgroups lack this character.

Homoplasy: None known, but see Quadrate, fossa pterygoidea (character 68).

Discussion: The two sister groups of the Podocnemididae , Brasilemys ( Lapparent de Broin, 2000b) and Hamadachelys ( 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 .

This character is used by Gaffney and Meylan (1988), Meylan (1996), Lapparent de Broin and Werner (1998), and Lapparent de Broin (2000b).

70. Pterygoid, processus trochlearis pterygoidei: absent ( Proganochelys ) 5 0; present ( Pelusios ) 5 1.

Morphology: The processus trochlearis pterygoidei (figs. 23, 24) is an important pleurodire synapomorphy described in Schumacher (1954, 1955a, 1955b, 1956) and Gaffney (1975 b, 1979a).

Primitive condition: All turtles outside Pleurodira lack the processus entirely.

Homoplasy: None known.

Discussion: This is a synapomorphy for Pleurodira .

This character is used in Gaffney and Meylan (1988), Gaffney et al. (1991), Rougier et al. (1995), Lapparent de Broin and Werner (1998), and Fuente and Iturralde-Vinent (2001).

71. Pterygoid, posteroventral flange along lateral edge, medial to processus trochlearis pterygoidei: absent ( Chelydra ) 5 0; present ( Pelusios ) 5 1.

Morphology: In pleurodires, the pterygoid forms a very thin sheet of bone, just medial to the processus trochlearis pterygoidei, which extends ventrally below the level of the rest of the pterygoid. These are figured in Gaffney (1979a: figs. 134, 142) and in figure 136B. Antunes and Broin (1988) Lapparent de Broin and Werner (1998) referred to this flange as a ‘‘pterygoid wing’’.

Primitive condition: Proganochelys and cryptodires lack these.

Homoplasy: None known.

Discussion: Lapparent de Broin and Werner (1998) argued that the flange or wing is found in most pleurodires, but that bothremydids and Araripemys lack them. We have concluded, however, that the flange is present in all pleurodires and that its supposed absence in some is a consequence of preservation. The flange or wing is very thin and fragile, often being damaged or broken off entirely in recent pleurodire specimens. The flange does occur in better preserved specimens of Araripemys (THUg 1907, fig. 33B) and in a number of bothremydids: Bothremys maghrebiana (fig. 141B), Labrostochelys (fig. 193B), and Cearachelys (fig. 71B).

72. Pterygoid, processus pterygoideus externus: without vertical plate ( Pelusios ) 5 0; with vertical plate ( Chelydra ) 5 1.

See Gaffney et al. (1987), Gaffney et al. (1991), Rougier et al. (1995), and Gaffney (1996).

73. Pterygoid, trigeminal ridge: absent ( Galianemys ) 5 0; ridge extending posteroventrally from foramen nervi trigemini to condylus mandibularis ( Phosphatochelys ) 51.

Morphology: In Taphrosphys , Ummulisani , and Phosphatochelys a thin ridge extends from the ventral margin of the foramen nervi trigemini posteroventrally along the quadrate process of the pterygoid, to the vicinity of the condylus mandibularis (fig. 202). This ridge may be an attachment area for a part of the pterygoideus muscle.

Primitive condition: All outgroups lack this character.

Homoplasy: This character occurs only in Taphrosphys and Phosphatochelys .

Discussion: This character supports the subgroup of Taphrosphyina consisting of Taphrosphys , Ummulisani , and Phosphatochelys .

74. Pterygoid, position of the foramen posterius canalis carotici interni: in ventral surface of basisphenoid ( Proganochelys ) 5 0; in pterygoid-quadrate suture ( Araiochelys ) 5 1; in medial wall of basisphenoid ( Kurmademys ) 5 2; in prootic ( Pelusios ) 5 3; in pterygoid-basisphenoid (also prootic in some cases) suture ( Euraxemys ) 5 4; in pterygoidbasisphenoid-quadrate suture ( Taphrosphys ) 5 5; in quadrate only ( Labrostochelys ) 5 6.

Morphology: The internal carotid artery of turtles enters the skull through the foramen posterius canalis carotici interni ( Albrecht, 1967, 1976; Gaffney, 1979a), which may be located in the basisphenoid, adjacent palatoquadrate elements, or in sutures between these elements. In Proganochelys and Kayentachelys , the internal carotid enters the skull via the ventral surface of the basisphenoid (state 0; Gaffney, 1990: fig. 20). Pleurodires modify this direct entry by interposing other elements (see figs. 276 and 277 for all states of the foramen posterius canalis carotici interni in pleurodires). Phylogenetically, based on the MPC (fig. 296B), the first element to contain the foramen was the prootic. In chelids, pelomedusids, and Araripemys , the carotid enters the prootic before entering the basisphenoid so that the foramen posterius canalis carotici interni is formed in the prootic (state 3). This character is synapomorphic for Pleurodira in the MPC; of course it is not known in the shell-only taxa. In all pleurodires (except possibly Kurmademys ), the carotid goes through the prootic, even though this may be covered ventrally by other bones.

The formation of the foramen posterius canalis carotici interni by the basisphenoid and pterygoid (fig. 277A, state 4) occurs in Euraxemydidae , Podocnemididae , Sankuchemys, Cearachelyini , Arenila , and some Bothremydini , so that in the MPC this condition is primitive for Bothremydidae and a synapomorphy for the magnafamily Podocnemidera (fig. 296B). This makes morphologic sense as well because the large medial process of the quadrate seen in many Bothremydini and Taphrosphyini may be interpreted as a modification of state 4, with the quadrate covering some of the basisphenoid to produce a foramen posterius canalis carotici interni formed by the quadrate and pterygoid, recognized here as a different condition, state 1 (fig. 277B).

In podocnemidids ( Gaffney, 1979a: fig.86), Hamadachelys , and Brasilemys , the carotid enters the skull at the anterior margin of the cavum pterygoidei. The foramen posterius canalis carotici interni is formed mostly by the basisphenoid, but the pterygoid and prootic are close. This character could be interpreted as unique to this group, but we have included it in state 4 because the bones involved in it are the same as in other taxa with state 4.

The three species of Taphrosphys , Bothremys kellyi , and Zolhafah have the foramen posterius canalis carotici interni formed at the junction of three bones: pterygoid, quadrate, and basisphenoid (fig. 277D, state 5).

Kurmademys is unique in having the carotid enter the basisphenoid on a laterally facing surface that forms the anteromedial edge of a moderate fossa pterygoidea (state 2, fig. 63). This is morphologically distinct from the primitive chelonian condition (state 0) in being more laterally placed on a much wider basisphenoid. Scoring it as state 0, however, does not change the MPC.

Labrostochelys (fig. 277F) (and one specimen of Ummulisani ) is unique in having only the quadrate form the foramen posterius canalis carotici interni. Nonetheless, it is not very different morphologically from state 1, formation by pterygoid and quadrate, as the pterygoid is not far from the foramen in Labrostochelys . The quadrate has a suture extending from the foramen posterius canalis carotici interni to the pterygoid, suggesting that the canal is only barely covered by quadrate. A separate state, state 6, is created for the quadrate only. However, the condition in Labrostochelys is morphologically so similar to state 1, pterygoid plus quadrate, that Labrostochelys is scored as polymorphic. This decision does make a difference in the cladogram, however. Scoring both Labrostochelys and Ummulisani as state 6 makes the three species of Taphrosphys a multichotomy with the Ummulisani-Rhothonemys-Labrostochelys clade, reflecting the low support for the species of Taphrosphys .

Teneremys ( Broin, 1980: pl. 2) seems to have a foramen posterius canalis carotici interni formed by both basisphenoid and pterygoid (although this has not been substantiated), but it is unique in having the foramen at the anterior end of the basisphenoid.

Primitive condition: The formation of the foramen posterius canalis carotici interni by a narrow basisphenoid as described in Gaffney (1990) is presumed to be primitive for turtles. Within Pleurodira , state 3, formation of the foramen by the prootic, is interpreted as primitive within this group.

Homoplasy: Although this is a complex character (fig. 296B) with seven states, its CI is a respectable 0.5. None of the states reverses back to state 0. States 0 and 3 do not reverse at all, although they are each lost once. The more general state 4 is lost four times, possibly more. State 1 arises independently within Bothremydini and Taphrosphyini . State 5 arises three times. State 6 occurs in only Labrostochelys and one of three specimens of Ummulisani .

Discussion: The formation of the internal carotid entry and its canal is more complex in pleurodires than in cryptodires ( Gaffney, 1979a, 1996), and this character is only one way of coding this variation. The formation of a fossa pterygoidea or a cavum pterygoidei alters the bones around the foramen posterius canalis carotici interni, obscuring a more general similarity. Combining the quadrate plus pterygoid with prootic plus quadrate plus pterygoid condition into one state was done because the prootic is exposed when the fossa pterygoidea is present (see character 68, fossa pterygoidea).

Some aspect of this character is used in Meylan (1996), Lapparent de Broin and Werner (1998), Tong et al. (1998), and Lapparent de Broin (2000b).

75. Pterygoid, posterior margin of pterygoid forms part of foramen posterius canalis carotici interni: no, pterygoid does not enter foramen ( Pelusios ) 5 0; yes, pterygoid enters anterior margin of foramen ( Galianemys ) 5 1.

Morphology: The pterygoid enters the anterior margin of the foramen posterius canalis carotici interni, apparently correlated with a posterior extension of the pterygoid (figs. 276, 277). Even in podocnemidids ( Gaffney, 1979a: fig.86), the pterygoid enters into the formation of the foramen.

Primitive condition: In chelids, pelomedusids, and Araripemys , the foramen posterius canalis carotici interni lies almost entirely within the prootic, sometimes with the basisphenoid entering it. In nearly all bothremydids the pterygoid is slightly more posterior and covers part of the foramen posterius canalis carotici interni. The more anterior position of the pterygoid posterior edge is primitive based on comparisons with Proganochelys .

Homoplasy: Kurmademys and Labrostochelys have unique conditions of the foramen posterius canalis carotici interni and lack a pterygoid contribution.

Discussion: In this analysis, this character is considered distinct from the preceding character 74, which also deals with the formation of the foramen posterius canalis carotici interni, because there are multiple patterns involving the foramen that are not expressed with only one character, even with multiple states. This character seeks to use the pattern that occurs in Euraxemydidae , Podocnemididae , and most Bothremydidae (fig. 306), that is, a slight posterior extension of the pterygoid reaching the foramen posterius canalis carotici interni. This condition seems to be a more general one (fig. 306) than the states identified in character 74 (fig. 296B); the other attempts to codify carotid variation. There are many carotid stories, and this is only one of them.

76. Pterygoid, foramen caroticum laterale: present ( Emydura ) 5 0; absent ( Pelusios ) 5 1.

Morphology: The foramen caroticum laterale is the anterior opening of the canalis caroticus lateralis (a lateral branch of the canalis caroticus internus) into the sulcus cavernosus that is usually just lateral to the foramen anterius canalis carotici interni and just medial to the foramen cavernosum ( Gaffney, 1979a: 101; see also Albrecht, 1967, 1976). It is an opening in the dorsal surface of the pterygoid adjacent to the basisphenoid. This opening usually transmits the palatine artery. Its presence seems to be primitive for Pelomedusoides, as it is found in chelids and Araripemys (fig. 34); however, this is not well supported because of its absence in pelomedusids.

Primitive condition: Proganochelys has a partially open cranioquadrate space, so the foramen caroticum laterale is not defined by bone ( Gaffney, 1990). It is present in many casichelydians ( Gaffney, 1979a: 101, figs. 54–65). Among pleurodires, it has been identified in Podocnemis , Peltocephalus , Erymnochelys , and all chelids except Chelus . It is considered to be absent in Pelusios and Pelomedusa ( Albrecht, 1976) .

Homoplasy: The foramen caroticum laterale appears to have been lost at least three times within the Pleurodira : once in the genus Chelus , once in the Pelomedusidae , and at least once for the Bothremydidae .

Discussion: This character can be scored only in specimens in which the braincase is very well preserved and has been prepared in detail. Thus, although the foramen caroticum laterale has yet to be found in bothremydids, its absence can be confirmed in only six species. Nonetheless, these six ( Kurmademys , Cearachelys , Galianemys , Bothremys cooki , B. maghrebiana , and Taphrosphys sulcatus ) include taxa in each tribe and are well distributed taxonomically.

77. Pterygoid, midline contact: midline contact of pterygoids relatively long ( Galianemys ) 5 0; midline contact very short ( Dirqadim ) 5 1.

Morphology: The pterygoids meet on the midline at their anterior end, anterior to the basisphenoid. In some taxa, the contact is very short ( Labrostochelys , fig. 11G) and the basisphenoid is close to the palatines.

Primitive condition: Most outgroups have a longer midline contact.

Homoplasy: Rampant, with a CI of 0.14. Within pleurodires, the character arises independently six times in the MPC.

Discussion: The only group this noisy feature supports is Nigeremys + Arenila .

This character is used in Meylan (1996) and Lapparent de Broin (2000b).

78. Epipterygoid, absent: present ( Chelydra ) 5 0; absent ( Pelusios ) 5 1.

Morphology: Described and figured by Gaffney (1975a, 1975 b, 1979a, 1979b), this bone is absent in pleurodires and some cryptodires.

Primitive condition: The bone is present in Proganochelys .

Homoplasy: Epipterygoid loss or fusion occurs within cryptodires in baenids and Dermochelys , but apparently only once in pleurodires.

Discussion: This character is a Pleurodira synapomorphy, but it is known only for taxa with good skulls, so it could be present from Pleurodira to Eupleurodira.

This character is used by Gaffney and Meylan (1988) and Gaffney et al. (1991).

79. Supraoccipital, supraoccipital-quadrate contact: absent ( Pelusios ) 5 0; present on dorsal surface of otic chamber ( Bothremys ) 5 1.

Morphology: The otic region of turtles in dorsal view can be seen to be made up of contributions of four elements: the quadrate laterally, the prootic anteriorly, the supraoccipital medially, and the opisthotic posteriorly. Although there is variation in the contributions made by these elements, in nearly all turtles the prootic and opisthotic make contact and thus intervene between the supraoccipital and quadrate. In most bothremydids (fig. 143A) the supraoccipital and quadrate meet on the dorsal surface of the otic region, preventing contact of the prootic and opisthotic.

The contacts of the otic chamber bones as seen on the dorsal surface do not necessarily reflect underlying contacts. As might be expected, the prootic-opisthotic contact is still present in some broken specimens, beneath the supraoccipital-quadrate contact.

Primitive condition: Prootic-opisthotic contact occurs in Proganochelys and all relevant outgroups.

Homoplasy: Although this is an important character in Bothremydidae (fig. 307), it has a low CI of 0.25. All bothremydids have a supraoccipital-quadrate contact except for Zolhafah and the Taphrosphyini , which represent reversals. Within the tribe Taphrosphyini , the undescribed skull CNRST-SUNY 199 is reversed to acquire the supraoccipital contact. No other pleurodires have the contact. However, some baenids ( Gaffney, 1979a: fig. 56) have a supraoccipital-quadrate contact.

Discussion: This character is a synapomorphy for the Bothremydidae , and its reversal is a synapomorphy for the Taphrosphyini (fig. 307).

This character is used in Tong et al. (1998).

80. Supraoccipital, crista supraoccipitalis: low to absent ( Proganochelys ) 5 0; a distinct sagittally oriented plate ( Pelusios ) 5 1.

See Gaffney (1990), Gaffney and Kitching (1995), and Rougier et al. (1995).

81. Supraoccipital, wide occipital plate: wide occipital plate with posteriorly concave depression ( Proganochelys ) 5 0; narrower occiput ( Pelusios ) 5 1.

See Gaffney (1990), Gaffney and Kitching (1995), and Rougier et al. (1995).

82. Exoccipital, foramen jugulare posterius: not formed in bone ( Proganochelys ) 5 0; formed by bone and open or partially closed ( Galianemys ) 5 1; completely closed by bone ( Azabbaremys ) 5 2.

Morphology: The foramen jugulare posterius is an opening surrounded mostly by the exoccipital that leads into the recessus scalae tympani. When completely closed (state 2, fig. 195), the lateral margin of this foramen is usually formed by the exoccipital but sometimes by the opisthotic (fig. 203). The foramen may be open laterally and continuous with the fenestra postotica (figs. 46, 98).

Primitive condition: The foramen jugulare posterius does not exist in bone in Proganochelys ( Gaffney, 1990: 80) . In most cryptodires it is enclosed by the exoccipital alone or by the exoccipital in combination with the opisthotic or pterygoid; in other cryptodires, it is continuous laterally with the fenestra postotica. Among living pleurodires it is closed laterally in Pelusios , Podocnemis , and in chelids other than Pseudemydura ( Gaffney, 1979a) . In Araripemys it is both open (fig. 36) and closed (fig. 37). In Euraxemys (fig. 46), it is open laterally. The fact that this opening is not defined by bone in Proganochelys has led us to score the open condition as primitive for turtles, but the closed condition seems to be primitive for pleurodires.

Homoplasy: This is a variable character with a CI of 0.33, and it has undoubtedly undergone multiple reversals and/or multiple acquisitions. Morphologically, it would be expected that the open or less complete condition would be primitive to the closed, more ossified condition, and this seems to be the case for Cryptodira ( Gaffney, 1996) . However, the closed condition in chelids and pelomedusids, as well as in podocnemidids, Kurmademys , and some Araripemys , suggests that the reverse is true. The open condition occurs independently three times in Pleurodira : Euraxemydidae , Cearachelyini, and Foxemydina.

Discussion: The open or reversed condition supports three groups in pleurodires (see above).

This character is used in Lapparent de Broin and Werner (1998) and Lapparent de Broin (2000b).

83. Exoccipital, recessus scalae tympani: not formed in bone ( Proganochelys ) 5 0; formed by bone, also forming fenestra perilymphatica ( Pelusios ) 5 1.

See Gaffney (1990), Gaffney and Kitching (1995), and Rougier et al. (1995).

84. Exoccipital, condylus occipitalis: basioccipital plus both exoccipitals ( Euraxemys ) 5 0; exoccipitals only ( Pelusios ) 5 1.

Morphology: The occipital condyle in the tribes Taphrosphyini and Bothremydini is unusual in being made up of only the two exoccipitals (fig. 203) rather than the three bones in nearly all cryptodires (except Carettochelys ) and most other pleurodires (except pelomedusids and Cearachelys ). The bothremydid condyle often has a vertical cleft down its center marking the suture.

Primitive condition: The outgroups Cryptodira , Chelidae , Araripemys , Euraxemydidae , Podocnemididae , and Kurmademys have three bones in the condylus occipitalis, the presumed primitive condition.

Homoplasy: The CI of this character is 0.33. The complete exclusion of the basioccipital from the condylus occipitalis neck must originate three times in the MPC: in the Pelomedusidae , Cearachelys , and the infrafamily Bothremydodda (the tribes Taphrosphyini and Bothremydini ). Galianemys has a wedge-shaped part of the basioccipital extending into the neck of the condylus occipitalis, almost to the articulating surface.

Discussion: This character is used in Gaffney and Meylan (1988), Meylan (1996), and Tong et al. (1998), Lapparent de Broin and Murelaga (1999), and Lapparent de Broin (2000b).

85. Exoccipital, exoccipital-quadrate contact: absent ( Pelusios ) 5 0; extensive, prootic absent ( Galianemys ) 5 1; narrow, prootic present ( Euraxemys ) 5 2; narrow, prootic absent ( Brasilemys ) 5 3.

Morphology: In most pleurodires the quadrate does not extend all the way to the occipital region so that the prootic and opisthotic prevent contact between the quadrate and exoccipital. In bothremydids the quadrate does extend medially to reach the occipital elements and has broad contact with the basioccipital (character 59) and the exoccipital (state 1, fig. 88). There is also contact between the quadrate and exoccipital in Euraxemydidae (state 2, figs. 44, 53), but the morphology differs from that in bothremydids in that the prootic is exposed ventrally, the exoccipital has a unique ventral process (see character 86), and the quadrate-exoccipital contact is very narrow. Brasilemys has a third condition (state 3) in which the contact is narrow and the prootic is absent ( Lapparent de Broin, 2000b).

Primitive condition: Quadrate-exoccipital contact is absent in Proganochelys ( Gaffney, 1990) and relevant outgroups, and thus this contact between these elements is considered derived.

Homoplasy: If the state for Euraxemydidae is considered distinct from that in the bothremydids, there is no known homoplasy for the states in this character (fig. 308).

Discussion: State 2 is synapomorphic for the Euraxemydidae and state 1 is synapomorphic for the Bothremydidae (fig. 308). Even though both states have ‘‘exoccipital-quadrate contact present’’, the differing morphologies suggest that these should not be ordered because the contact alone does not seem to be homologous.

A version of this character is used in Lapparent de Broin (2000b).

86. Exoccipital, ventral process: absent ( Pelusios ) 5 0; present ( Euraxemys ) 5 1.

Morphology: In most turtles the exoccipital forms the lateral margins of the foramen magnum, the dorsolateral part of the occipital condyle, and the posteroventral part of the cavum cranii. It does not normally form part of the ventral surface of the skull. In Euraxemys and Dirqadim there is a unique condition in which a process of the exoccipital extends ventrally between the opisthotic and basioccipital and is exposed on the ventral surface lateral to the basioccipital (figs. 44, 46, 53).

Primitive condition: The exoccipital of Proganochelys is excluded from participation in the ventral surface of the skull by prooticopisthotic contact ventral to the exoccipital. In most cryptodires, the basioccipital or pterygoid excludes the exoccipital from ventral exposure (but see Baena arenosa, Gaffney, 1979a : fig. 153), and in pleurodires some combination of contacts between the quadrate, basioccipital, prootic, and opisthotic prevents contribution of the exoccipital to the ventral surface. Thus, exclusion of the exoccipital from the ventral surface is clearly the primitive condition.

Homoplasy: Some baenids ( Gaffney, 1972a, 1979a) have a ventral exposure of the exoccipital, but its position and shape differ from euraxemydids. Brasilemys ( Lapparent de Broin, 2000b) has apparently independently acquired this character. It also occurs as a variation in some chelids ( Phrynops gibba PAM 2051 , FMNH 45669 in fig. 147; Gaffney, 1979c). Araripemys (fig. 26) has a widely exposed exoccipital on the ventral surface correlated with a very narrow basioccipital.

Discussion: This feature is a synapomorphy for Euraxemydidae .

87. Basioccipital, basioccipital short: long, length/width of 0.60 or higher ( Pelusios ) 5 0; short, length/width of 0.59 or lower ( Bothremys ) 5 1.

Morphology: The basioccipital of turtles is variable in shape but always makes up the posteriormost part of the ventral surface of the braincase. In most turtles it is approximately as long as wide. In bothremydids the basioccipital is comparatively short relative to its width (figs. 9–11), a clear difference from chelids, for example. However, this character could be judged as continuously variable with the distinction between ‘‘long’’ and ‘‘short’’ being subjective and perhaps arbitrary. A ratio of basioccipital length over width reveals a short gap so that ‘‘long’’ could be considered 0.60 and higher, with ‘‘short’’ being 0.59 and lower. Using this slight gap to differentiate the two states, podocnemidids and bothremydids are long. With one exception ( Taphrosphys sulcatus ), the Bothremydidae ratio varies from 0.31 ( Chedighaii ) to 0.57 ( Cearachelys , Phosphatochelys ). Podocnemidids (including Hamadachelys ) are 0.59 or lower. The chelids,which are variable (we have chosen Emydura ), are 0.64, but pelomedusids are higher at 0.70– 0.78. Euraxemydids are 0.73–0.77 and Araripemys is the highest at 1.2. Therefore, the closest ratios for the two states are podocnemidids at 0.59 and chelids at 0.64.

Primitive condition: The pleurodiran outgroups are long in the basioccipital.

Homoplasy: A simple character like this is difficult to test for homology, but the only reversal in the MPC is the relatively long basioccipital of Taphrosphys sulcatus at 0.68. Interestingly, the other two Taphrosphys species are 0.47 ( T. congolensis ) and 0.57 ( T. ippolitoi ), suggesting that this condition evolved within this genus.

Discussion: Although a continuous character like this one could be divided into a number of states, we have chosen only two to try to reflect the short basioccipital of bothremydids. As a result, it seems that this feature is shared with podocnemidids, although choosing another division point could change that.

This character or a version of it (‘‘telescoped occiput’’) is used in Antunes and Broin (1988), Lapparent de Broin and Werner (1998), and Lapparent de Broin (2000b).

88. Basioccipital, basioccipital thick: basioccipital and basisphenoid relatively thick in cross section ( Proganochelys ) 5 0; thinner ( Pelusios ) 5 1.

See Gaffney (1990) and Gaffney and Kitching (1995).

89. Basioccipital, basioccipital-opisthotic contact: absent ( Galianemys ) 5 0; present ( Pelomedusa ) 5 1.

Morphology: In most casichelydians, the opisthotic is excluded from contact with the basioccipital except in cases where the processus interfenestralis is well ossified ventrally ( Gaffney, 1979a: 135–136). In certain pleurodires, there is a ventral process of the opisthotic that is posterior to the processus interfenestralis that makes a strong, sutured contact with the basioccipital between the foramen jugulare posterius and the fenestra postotica ( Gaffney, 1979a: figs. 85, 86). Among pleurodires, this contact is present in most but not all Chelidae ( Gaffney, 1979a: fig. 88), but not in Phrynops gibba PAM 2051 or in Chelodina ( Gaffney, 1979a: 136) in which the exoccipital intervenes. It is present in the Pelomedusidae (fig. 21) and Podocnemididae as well as Hamadachelys . The contact is absent in all Bothremydidae .

Primitive condition: In Proganochelys the opisthotic makes a ventromedial contact with the basioccipital. Although the exact position of the suture is unclear ( Gaffney, 1990: 86), it is posterior to the processus interfenestralis and may be homologous to the contact seen in several pleurodiran groups. Whether the condition in Proganochelys is homologous (we have scored it as questionable), this sutured posterior contact in members of the Chelidae , Pelomedusidae , and Podocnemididae (plus Hamadachelys ) suggests that the presence of this contact may be primitive for pleurodires, although this is equivocal.

Homoplasy: Opisthotic–basioccipital contact between the foramen jugulare posterius and the fenestra postotica occurs in some cryptodires. It has arisen independently in podocnemidids. The CI is 0.5.

Discussion: The distribution of this character among the Pleurodira suggests that it has been gained twice or lost three times. In either case, it is uniformly absent in the Bothremydidae although it is present in the sister group, the Podocnemidinura.

A version of this character is used in Lapparent de Broin (2000b).

90. Prootic, hyomandibular branch of facial nerve lies in its own canal: no ( Chelydra ) 5 0; yes ( Podocnemis ) 5 1.

Morphology: Described and figured in Gaffney (1975 b, 1979a), determining this character requires access to the canalis cavernosus and the internal morphology of the prootic. A short canal contains the hyomandibular nerve, separating it from the canalis cavernosus.

Primitive condition: The absence of a canal is the condition in Proganochelys and Cryptodira .

Homoplasy: None known.

Discussion: This character is a synapomorphy for the Pleurodira . Despite the fact that it can only be seen in well-preserved and partially disarticulated skulls, it is included. It is determinable in some fossils; Galianemys and Dirqadim are known from CT scans.

This character is used by Gaffney and Meylan (1988) and Gaffney et al. (1991).

91. Prootic, foramen stapedio-temporale: not a canal ( Proganochelys ) 5 0; a distinct foramen and canal ( Pelusios ) 5 1.

See Gaffney et al. (1987), Gaffney et al. (1991), and Gaffney (1996).

92. Prootic, foramen stapedio-temporale opens anteriorly: no ( Pelusios ) 5 0; yes ( Galianemys ) 5 1.

Morphology: The foramen stapedio-temporale lies on the anterior surface of the otic chamber in most bothremydids. Compared with the outgroups, the foramen is placed more anteroventrally in the prootic-quadrate suture. When the skull is seen in dorsal view, there is either no sign of the foramen or only a narrow part of the margin (figs. 7,8). This depends to a certain extent on how the skull is oriented, but unless the skull is strongly tilted, the foramen margins usually cannot be seen. If an extensive skull roof covers the otic chamber, some rotation of the skull must be done to see the position of the foramen stapedio-temporale.

Primitive condition: In the pleurodire outgroups Chelidae , Pelomedusidae (fig. 21A), Euraxemydidae (fig. 45), and Podocnemididae , the foramen stapedio-temporale is visible in dorsal view because it is in a more posterodorsal position than in the Bothremydidae .

Homoplasy: The subfamily Bothremydinae (consisting of the tribes Cearachelyini , Taphrosphyini , and Bothremydini ) has this character (fig. 309). Kurmademys has the primitive condition, but Sankuchemys may have the derived condition. Unfortunately, Sankuchemys is badly crushed and the presence or absence of the foramen is ambiguous. However, the otic chamber is crushed directly dorsoventrally, and there is no sign of the foramen stapedio-temporale on either side. In the character set Sankuchemys is coded as missing for this character.

Discussion: Despite the ambiguity in Kurmademydini , this character is definitely found only in the subfamily Bothremydinae (fig. 309) and is a synapomorphy for it.

This character is used in Meylan (1996) and Tong et al. (1998).

93. Prootic, foramen stapedio-temporale and foramen nervi trigemini: separated by most of prootic ( Pelusios ) 5 0; separated by narrow bar of prootic ( Bothremys ) 5 1.

Morphology: In the bothremydid tribes of Bothremydini and Taphrosphyini , the foramen stapedio-temporale lies on the anterior surface of the otic chamber, separated from the foramen nervi trigemini by a thin bar of prootic (fig. 132F). In these taxa, the foramen stapedio-temporale is mostly formed by the prootic, with a narrow part of quadrate entering the margin. The foramen itself is usually close to the center of the prootic along its ventral edge. In many specimens the prootic bar is broken away, but the position of the foramen is still determinable.

Primitive condition: The pleurodiran outgroups show a foramen stapedio-temporale and foramen nervi trigemini separated by most of the prootic to be the primitive condition.

Homoplasy: None known.

Discussion: As this character requires the foramen stapedio-temporale to be on the anterior surface of the otic chamber, it could be coded as a third state in the preceding character: foramen stapedio-temporale not visible in dorsal view. When the character states are ordered, it produces the same MPC.

94. Prootic, ventral exposure: most of prootic exposed ventrally ( Emydura ) 5 0; prootic about half covered by quadrate and basisphenoid ( Euraxemys ) 5 1; nearly all of prootic covered by quadrate, basisphenoid, and pterygoid ( Taphrosphys ) 5 2.

Morphology: In euraxemydids the basisphenoid and quadrate cover most of the prootic, leaving only a narrow exposure of prootic (state 1, fig. 44). In podocnemidids and bothremydids the prootic is completely covered (except for a small exposure described as another character, Prootic, ventral exposure 2). This is state 2 (fig. 277).

Primitive condition: Proganochelys has a completely exposed prootic ( Gaffney, 1990). Chelids and pelomedusids (figs. 21, 276) also have this condition, although other bones encroach on the prootic.

Homoplasy: None. Cryptodires also cover the prootic ventrally, but it is done entirely by the pterygoid and appears to be nonhomologous ( Gaffney, 1975 b, 1979a).

Discussion: The partial covering of the prootic in euraxemydids is very similar to the complete covering in other pleurodires, as it involves a lateral extension of the basisphenoid and a medial extension of the quadrate. This character could be run ordered because state 1 is additive with respect to state 2 (fig. 310). All taxa that have state 2 also have state 1, so to run them as separate characters would lose this information. However, running this character unordered results in the same MPC (fig. 310).

This character is used in Antunes and Broin (1988), Meylan (1996), Lapparent de Broin and Werner (1998), and Tong et al. (1998).

95. Prootic, quadrate-pterygoid-basisphenoid exposure: complete exposure of prootic or no exposure of prootic ( Pelusios ) 5 0; exposure of prootic small, surrounded by pterygoid, basisphenoid, and quadrate, with foramen nervi facialis exposed ( Kurmademys ) 5 1.

Morphology: In the bothremydids Kurmademys , Sankuchemys , Cearachelys (not all specimens, see Cearachelys ), and Galianemys emringeri , there is a small, irregular exposure of the prootic at or near the junction of the pterygoid, basisphenoid, and quadrate (fig. 277E). This exposure surrounds or nearly surrounds the foramen nervi facialis, which is therefore visible in ventral view in these forms, in contrast to other bothremydids.

Primitive condition: The primitive condition for prootic exposure is widely exposed, as in chelids and pelomedusids, but in this character we identify as primitive all other prootic conditions that lack the character as specified.

Homoplasy: In the MPC (fig. 311), this character appears at least three and possibly four times independently, and it has a CI of 0.33. Elsewhere (character 68, fossa pterygoidea, fig. 305) we argue that at least in some cases the exposure of the prootic is the result of a deep fossa pterygoidea that has eroded the basisphenoid, pterygoid, and quadrate, exposing the underlying prootic.

Discussion: The high degree of homoplasy and the frequent (but not exclusive) association of the prootic exposure with a deep fossa pterygoidea suggest it may not be homologous everywhere it appears. However, Kurmademys has a moderate fossa and a prootic exposure, and Sankuchemys has the exposure but no evidence of a fossa (although the only known skull is crushed). In the MPC the small exposure is a synapomorphy for the tribe Kurmademydini (fig. 311).

96. Prootic, processus trochlearis oticum: absent ( Proganochelys ) 5 0; present ( Chelydra ) 5 1.

See Gaffney et al. (1987), Gaffney et al. (1991), Gaffney (1996), and Rougier et al. (1995).

97. Prootic, plane of fenestra ovalis: inclined from the vertical ( Proganochelys ) 5 0; very close to vertical ( Chelydra ) 5 1.

See Gaffney (1990), Gaffney and Kitching (1995), and Rougier et al. (1995).

98. Opisthotic, processus interfenestralis expanded ventrally: narrow ventrally ( Proganochelys ) 5 0; expanded ventrally ( Pelusios ) 51.

See Gaffney (1990).

99. Opisthotic, processus interfenestralis covered in ventral view: exposed ventrally ( Emydura ) 5 0; covered by bone ( Bothremys ) 51.

Morphology: The processus interfenestralis ( Gaffney, 1972 b, 1979a) of the opisthotic is visible in ventral view in chelids, pelomedusids, and Araripemys (fig. 276). This is correlated with portions of the cavum labyrinthicum, fenestra ovalis, and recessus scalae tympani also being open ventrally. In life, this area is filled with cartilage in chelids and pelomedusids ( Gaffney, 1979a). In euraxemydids, podocnemidids, and bothremydids, a medially large quadrate and posterolaterally large basisphenoid cover this area with bony ossification (figs. 276, 277).

Primitive condition: The exposed condition of chelids and pelomedusids is closest to the completely open condition in Proganochelys and is presumed to be primitive for pleurodires.

Homoplasy: None known, except for within Selmacryptodira.

Discussion: It might be possible to break the derived state into two states, including one for euraxemydids, which can be interpreted as having a smaller medial quadrate process than do podocnemidids and bothremydids. If ordered, no information would be lost. However, judging the difference in size of the medial quadrate process seems too subjective, even for me.

This character is used in Lapparent de Broin and Werner (1998), Lapparent de Broin and Murelaga (1999), and Lapparent de Broin (2000b).

100. Opisthotic, fenestra postotica closed medially: open medially ( Euraxemys ) 5 0; closed medially by opisthotic-quadrate contact ( Taphrosphys ) 5 1.

Morphology: The fenestra postotica ( Gaffney, 1972 b, 1979a) is limited laterally and ventrally by the quadrate and dorsally by the opisthotic. When the opisthotic and quadrate meet medial to the fenestra postotica, it is closed (figs. 132D, 218). Only the Cearachelyini (figs. 87, 88, 98–100) among bothremydids have a medially open fenestra postotica.

Primitive condition: All the pleurodiran outgroups have a medially open fenestra postotica.

Homoplasy: In the MPC, the closed condition evolved twice, once in Kurmademys ( Sankuchemys is unknown) and once in the infrafamily Bothremydodda (the tribes Taphrosphyini and Bothremydini ), or it was lost once in the tribe Cearachelyini .

Discussion: This character is independent of a laterally closed foramen jugulare posterius. Chelids and pelomedusids have a laterally closed foramen jugulare posterius with an open fenestra postotica.

This character or a version of it was used in Lapparent de Broin and Werner (1998) and Lapparent de Broin (2000b).

101. Opisthotic, fenestra postotica short horizontal slit: more open dorsoventrally ( Euraxemys ) 5 0; short slit ( Galianemys ) 5 1.

Morphology: The Cearachelyini have a uniquely shaped fenestra postotica. The fenestra postotica is open medially, and it is dorsoventrally compressed in comparison to other pleurodires. The result appears as a short slit (figs. 87, 88, 98–100), in contrast to the round or oval opening of other bothremydids.

Primitive condition: A more open condition is in all relevant outgroups and is presumed primitive.

Homoplasy: None known.

Discussion: This character is a synapomorphy for the tribe Cearachelyini . This character could be a state within the closed fenestra postotica character, but it would be confusing, at least to me.

102. Opisthotic, processus paroccipitalis: projects posteriorly beyond squamosal ( Euraxemys ) 5 0; smaller, does not project beyond squamosal ( Galianemys ) 5 1.

Morphology: The variation in shape of the opisthotic-squamosal area of the occiput in pleurodires is a source of characters, but some are hard to define. This character is an attempt to reflect the small size of the processus paroccipitalis ( Gaffney, 1972 b, 1979a) in bothremydids as a character. It is compared to the size of the squamosal, although this is not entirely satisfactory as some taxa (e.g., Labrostochelys ) have very long squamosals. The character is best seen in figures 9–11.

Primitive condition: The condition in Proganochelys ( Gaffney, 1990) is presumed to be primitive, with the processus paroccipitalis projecting posteriorly. However, Proganochelys has a unique morphology (among turtles), and the squamosal is very small. Cryptodires and chelids have a larger squamosal with a smaller processus paroccipitalis, and this may be primitive for pleurodires. Nonetheless, pelomedusids, Araripemys , euraxemydids, and podocnemidids have a long, projecting processus paroccipitalis, and it is possible that this is the primitive condition.

Homoplasy: The small processus paroccipitalis has originated twice in the MPC, once (possibly twice) within cryptodires and in chelids and once in bothremydids, being reversed in all other pleurodires. However, the reversed condition in pleurodires is morphologically distinct from the Proganochelys condition.

Discussion: Although it gets a little fuzzy around cryptodires, chelids, and Proganochelys , this character is a bothremydid synapomorphy.

103. Opisthotic, thin horizontal flange: absent ( Galianemys ) 5 0; thin horizontal flange present on posterior edge of opisthotic just dorsal to fenestra postotica ( Chedighaii hutchisoni ) 5 1.

Morphology: Best seen in Chedighaii hutchisoni (fig. 159), KUVP 14765, this thin flange lies ventral to the distinct curve or ‘‘step’’ as the opisthotic curves from a horizontal surface ventrally to a vertical surface. The flange is so distinct in Chedighaii that it has a shallow, horizontal groove above it. Below the flange is the fenestra postotica. In C. barberi , Alabama 2001.2, the flange is broken off on both sides but its base is clear and is nearly identical to that in KUVP 14765.

Primitive condition: All other pleurodires and Proganochelys lack this flange.

Homoplasy: None known, except that among the five skulls of Bothremys maghrebiana the flange is variable.

Discussion: Chelids and pelomedusids may have a narrow posterior margin to the processus paroccipitalis of the opisthotic, but this condition is distinct from the Chedighaii / Bothremys condition.

104. Basisphenoid, basisphenoid-quadrate contact: absent 5 0; present and wide, as in Galianemys 5 1; present and narrow, as in Azabbaremys 5 2.

Morphology: The basisphenoid-quadrate contact is a Podocnemidoidea ( Podocnemididae + Bothremydidae ) synapomorphy, known for some time ( Lapparent de Broin and Werner, 1998; Lapparent de Broin, 2000b). Examination of its distribution in the Bothremydidae (figs. 9–11) suggests that two states can be distinguished, a broader contact (as in Galianemys , fig. 81B) and a narrower contact (as in Azabbaremys , fig. 212B). As in all of these more subjective states, there is an arbitrary element to determining which state is present.

Primitive condition: The absence of a contact in nearly all of the relevant outgroups indicates that this is the primitive state. Some Pelusios may have the contact as an individual variation.

Homoplasy: None known for the presence of a contact (except some Pelusios ). However, the two states in the MPC (fig. 312) show two origins and a loss for state 2, the narrow contact, once in Bothremys + Chedighaii and once for Taphrosphyini , with a reversal in Taphrosphys + Ummulisani + Phosphatochelys . However, Phosphatochelys shows variation in this feature and it has been coded with both states. The CI is 0.5.

Discussion: There may be some ambiguity in discriminating a few taxa, such as Phosphatochelys , regarding the two derived states, but it still seems worthwhile to extract this information and reflect it in the datasets. The character can also be run as a simple presence or absence of the contact; this is a synapomorphy for Bothremydidae (fig. 312).

This character is used in Lapparent de Broin and Werner (1998) and Lapparent de Broin (2000b).

105. Basisphenoid, interpterygoid vacuity: large and open ( Proganochelys ) 5 0; small or absent ( Pelusios ) 5 1.

See Gaffney (1990), Gaffney et al. (1991), Gaffney and Kitching (1995), and Rougier et al. (1995).

106. Basisphenoid, ventral outline: elongate, basisphenoid not sutured to pterygoids ( Proganochelys ) 5 0; more triangular (baenids) 5 1; more pentagonal ( Taphrosphys ) 5 2; very elongate ( Araripemys ) 5 3; V-shaped pointing anteriorly ( Arenila ) 5 4.

Morphology: The shape of the basisphenoid as seen in ventral view varies, and this character set is an attempt to identify some of this variation. In most bothremydids (figs. 9– 11) the basisphenoid is triangular, with the base and the anterolateral contacts with the pterygoid being dominant (state 1, 9H). The shape can also be more pentagonal (state 2, fig. 11B), with the quadrate contact being much longer and the pterygoid contact shorter. The pentagonal (state 2) occurs in chelids, pelomedusids, podocnemidids, the Kurmademydini , Foxemys + Polysternon , and Taphrosphys . Araripemys is unique in having a very elongate basisphenoid (state 3, fig. 9C). In Nigeremys and Arenila , the basisphenoid is V-shaped or arrow-shaped (state 4, fig. 11I), with its posterior margin embayed, concave posteriorly.

Primitive condition: Although not exactly comparable to casichelydians due to the absence of a sutured pterygoid, Proganochelys has a roughly triangular shape, but it lacks the flat ventral surface and contacts of other turtles and is coded as a separate state. Cryptodires are scored triangular, and this is the outgroup for the pleurodires. However, the pentagonal shape (state 2) is in chelids and pelomedusids.

Homoplasy: The CI is 0.44 and there is a fair amount of homoplasy within bothremydids.

The V-shaped state 4 is unique to Nigeremys + Arenila , and state 3 is unique to Araripemys . The pentagonal state 1, however, is more complex. In the MPC, it is primitive for pleurodires and lost or reversed in Araripemys , euraxemydids, and Cearachelyini + Bothremydini + Taphrosphyini . It appears independently in Foxemys + Polysternon and the genus Taphrosphys .

Discussion: The pentagonal basisphenoid is a way of wresting information from the extent of quadrate-basisphenoid contact. Such a simple character is hard to homologize and is somewhat subjective. Phosphatochelys has a more irregularly shaped basisphenoid that could be considered pentagonal, but we have scored it triangular. Thus, there are some ambiguities in this state. The V-shaped basisphenoid of Arenila and Nigeremys is quite different from other taxa.

107. Basisphenoid, processus clinoideus: present, with abducens canal ( Pelusios ) 5 0; processus clinoideus absent, canal is a groove ( Bothremys ) 5 1.

Morphology: In Bothremys and Chedighaii barberi the small, spurlike processus clinoideus ( Gaffney, 1972 b, 1979a) is not ossified and the rim of the sella turcica and the dorsum sellae is a smooth edge (fig. 134). Just lateral to the dorsum sellae is a shallow groove that seems to be for the abducens nerve. This condition is visible in Bothremys cooki, AMNH 2521 ; B. maghrebiana, AMNH 30041; and in the endocast of a presumed Chedighaii barberi, YPM PU 12951.

Primitive condition: The processus clinoideus and foramen nervi abducentis are found in all other pleurodires and Proganochelys .

Homoplasy: None known at present, given the current MPC. However, the dorsal surface of the basisphenoid is determinable only in eight other bothremydid species.

Discussion: The dorsal surface of the basisphenoid is not visible in Bothremys arabicus , Chedighaii hutchisoni , and Araiochelys , which makes this character known in only about half of the important taxa, so there are many missing data.

108. Basisphenoid, skull akinetic: no, basipterygoid articulation present ( Proganochelys ) 5 0; yes, basipterygoid articulation sutured ( Australochelys ) 5 1.

See Gaffney (1990), Gaffney et al. (1991), Gaffney and Kitching (1995), and Rougier et al. (1995).

109. Basisphenoid, cultriform process: rodlike, thin ( Proganochelys ) 5 0; broad, flat, covered ventrally ( Australochelys ) 5 1.

See Gaffney (1990) and Gaffney and Kitching (1995).

110. Basisphenoid, sella turcicałdorsum sellae: deep, well-defined margins ( Bothremys ) 5 0; very shallow, low margins ( Taphrosphys sulcatus ) 5 1.

Morphology: Nearly all pleurodires have a well-defined and relatively deep sella turcica concavity with a distinct dorsum sellae at its posterior margin ( Gaffney, 1979a). In two species of Taphrosphys , T. sulcatus and T. ippolitoi , the sella is very low, being barely recognizable, and the dorsum sellae is not significantly above the floor of the sella turcica (fig. 174).

Primitive condition: The deep, well-defined sella turcica occurs in chelids, pelomedusids, and other outgroups and is the presumed primitive condition.

Homoplasy: None known.

Discussion: Although this character resolves the three species of Taphrosphys , the few bothremydid taxa for which it is known, particularly among the other Taphrosphyini , make its use questionable and it has been deleted in some of the analyses. When deleted, the three species are a multichotomy.

111. Basisphenoid, ventral tubercle: single tubercle formed by basisphenoid and basioccipital ( Proganochelys ) 5 0; paired tubercles ( Australochelys ) 5 1; no tubercle ( Pelusios ) 5 2.

See Gaffney (1990) and Rougier et al. (1995).

112. Columella auris, footplate: footplate absent, stapes thick ( Proganochelys ) 5 0; wide footplate present, stapes thin ( Pelusios ) 5 1.

See Gaffney (1990) and Gaffney and Kitching (1995).

113. Splenial: present ( Proganochelys ) 5 0; absent ( Pelusios ) 5 1.

Morphology: The splenial bone ( Gaffney, 1979a) lies on the medial surface of the lower jaw, anterior to and slightly separating the prearticular and angular. It is missing in all pleurodires except Chelidae . In these forms, the angular and prearticular tend to extend anteriorly, filling the space occupied by the splenial in chelids (fig. 241).

Primitive condition: A large splenial is present in Proganochelys ( Gaffney, 1990) , the presumed primitive condition.

Homoplasy: None known within pleurodires.

Discussion: Due to lack of lower jaws and the need of a well-preserved specimen to determine the absence of a splenial, the character can be scored in only four bothremydids. Nonetheless, it remains as a Pelomedusoides synapomorphy.

This character is used in Gaffney and Meylan (1988), Meylan (1996), and Lapparent de Broin (2000b).

114. Dentary, high lingual ridge: absent ( Pelusios ) 5 0; present ( Bothremys ) 5 1.

Morphology: The lingual ridge in bothremydids like Bothremys , Kurmademys , Cearachelys , and Foxemys is higher and deeper than the labial ridge (fig. 241). It is about the same height (or even lower) anteriorly, and it rises posteriorly to be wedge-shaped in lateral view.

Primitive condition: All the outgroups have both lingual and labial ridges of equal size or have a higher labial ridge. We have not distinguished these conditions.

Homoplasy: A high lingual ridge is present in the Cearachelyini , Kurmademydini , and Bothremydini , but it is reversed in Taphrosphyini . The living podocnemidids also have a larger lingual ridge, but fossil taxa vary. Hamadachelys has a lower jaw with both ridges nearly the same, but the lingual is slightly higher than the labial, an appropriate primitive condition. It is possible that a higher lingual ridge is primitive for Podocnemididae and could be argued as primitive for Podocnemididae + Bothremydidae , but at present, it is more likely to have appeared independently within Podocnemididae and Bothremydidae .

Discussion: The principal difficulty with this character is the few lower jaws available for bothremydids and other pleurodires. Nonetheless, the known jaws are representative of the bothremydid tribes.

115. Dentary, dentary pits: absent ( Euraxemys ) 5 0; present ( Bothremys ) 5 1.

Morphology: The pitted lower jaws were first described by Leidy (1865) in Bothremys cooki (fig. 19), and later by Hay (1908). Gaffney and Zangerl (1968) redescribed B. cooki and added the lower jaws of ‘‘ Bothremys ’’ barberi (here referred to Chedighaii barberi ). The conical pits (fig. 239) are concave anteriorly. The pits are known in Araiochelys , B. cooki (fig. 239), B. maghrebiana (fig. 241), and Chedighaii barberi .

Primitive condition: The absence of pits is primitive.

Homoplasy: None known.

Discussion: In Chedighaii barberi , the lower jaw has pits but the maxilla does not. In the other forms with lower jaw pits, there are also pits in the maxilla. This lack of complete correspondence is the rationale, however dubious, for recognizing separate characters in the lower jaw and the skull.

116. Dentary, U-shaped lingual ridges on symphysis: absent ( Euraxemys ) 5 0; present ( Bothremys ) 5 1.

Morphology: In bothremydids the lingual ridge of the dentary is higher than the labial ridge and is strongly developed. In some, the pair of lingual ridges join to form a U-shaped ridge on the dentary at the symphysis (figs. 241, 247). The jaw edges are not parallel to the labial ridges. The U-shaped ridge may be excavated medially, as in Bothremys maghrebiana and Araiochelys , or it may be filled with bone to form a wedge, as in Bothremys cooki (see also fig. 247, a Madagascar lower jaw described in Gaffney and Forster, 2003).

Primitive condition: The U-shaped symphyseal ridge is absent in Proganochelys , chelids, Araripemys , and Euraxemys . In the Podocnemididae there are strong lingual ridges that are taller than the labial ridges. but they form a V-shaped rather than a Ushaped ridge with a pointed apex.

Homoplasy: A U-shaped ridge like that in most bothremydids also occurs in some Pelusios . There is a reversal within the Taphrosphyini (see Discussion).

Discussion: A U-shaped ridge on the symphysis appears to be a synapomorphy for the Bothremydidae that is lost within the tribe Taphrosphyini . The distribution of lower jaws in the Taphrosphyini is too poorly known to be certain where this reversal occurs.

117. Dentary, symphysis: fused ( Bothremys ) 5 0; sutured ( Euraxemys ) 5 1.

Morphology: The symphysis between the two dentaries is normally fused in turtles with no evidence of a suture. In a few members of the Pleurodira the two dentaries are sutured in the midline instead of being fused. The symphysis is sutured in Araripemys , Euraxemys (figs. 231, 232), Pelomedusa , some Pelusios , and the chelid genera Platemys , Phrynops , Chelus , Chelodina , and Hydromedusa ( Gaffney, 1979a) .

Primitive condition: The fused condition seen in Proganochelys and all cryptodires is presumably the primitive condition.

Homoplasy: In the MPC, the character distribution is ambiguous due to polymorphism within chelids and pelomedusids. The CI is 0.5 but the character is reversed within the Chelidae and Pelomedusidae .

Discussion: Sutured rami are found in Araripemys and Euraxemys and were used by Meylan (1996) to unite them in a family Araripemydidae . However, in the MPC this character is polymorphic in chelids and pelomedusids. The chelid distribution suggests that if Gaffney (1977b) is correct, the fused condition is primitive for chelids. In pelomedusids it is fused only in some Pelusios , so the sutured condition may be primitive for this group plus Araripemys . If this interpretation is correct, the sutured condition evolved once within chelids and once for all remaining pleurodires, being reversed for the Podocnemidoidea (the families Podocnemididae + Bothremydidae ).

This character is used in Meylan (1996) and Lapparent de Broin (2000b).

118. Dentary, triturating surfaces: relatively narrow ( Euraxemys ) 5 0; wide posteriorly ( Cearachelys ) 5 1.

Morphology: Many bothremydids have a lower jaw with a triangular-shaped surface that is much wider posteriorly than anteriorly (fig. 241).

Primitive condition: Relatively narrow jaws, with labial and lingual ridges that are more nearly parallel to each other, are present in the outgroups.

Homoplasy: Podocnemidids evolve broad lower jaws, but these do not have the same morphology as bothremydids. The CI is 0.33, and the character is reversed in Araiochelys and the Taphrosphyini , which have narrower jaws. Araiochelys , however, is still very similar to Bothremys , just narrower. The Taphrosphyini , though, do have a very narrow triturating surface that is morphologically distinct form other bothremydids to the extent they are known.

Discussion: Although lower jaws are represented by many missing data, they are known for each major group (tribes). The triangular or wide shape is synapomorphic for the Bothremydidae , and the reversed, narrow shape is synapomorphic for the Taphrosphyini .

119. Dentary, widely exposed on lateral surface: yes, widely exposed posteriorly ( Euraxemys ) 5 0; no, covered posteriorly by surangular ( Bothremys ) 5 1.

Morphology: The surangular and coronoid in the tribes Bothremydini and Taphrosphyini are extensively exposed on the lateral surface of the jaw ramus (fig. 241), so that the dentary exposure is relatively small.

Primitive condition: The outgroups have broad exposure of the dentary posteriorly.

Homoplasy: A widely exposed dentary also occurs in the Podocnemididae (fig. 313) but not in Hamadachelys or Brasilemys .

Discussion: This character (fig. 313) is a synapomorphy for the infrafamily Bothremydodda (the tribes Bothremydini and Taphrosphyini ).

120. Surangular, foramen nervi auriculotemporalis: absent ( Proganochelys ) 5 0; present ( Podocnemis ) 5 1.

Morphology: The foramen nervi auriculotemporalis (fig. 232C, see also Gaffney, 1979a) lies at the posterolateral end of the lower jaw in the surangular. Its absence results in a smooth lateral surface on the surangular (fig. 241).

Primitive condition: The foramen is widely present in pleurodire outgroups and cryptodires, but it seems to be absent in Proganochelys .

Homoplasy: None, other than a reversal for the infrafamily Bothremydodda .

Discussion: The absence of the foramen nervi auriculotemporalis is a synapomorphy for the infrafamily Bothremydodda ( Bothremydini + Taphrosphyini ).

121. Coronoid, wide lateral exposure: no ( Euraxemys ) 5 0; yes ( Bothremys ) 5 1.

Morphology: In the Bothremydini , the coronoid is exposed ventrolaterally and anteroventrally on the lateral surface of the jaw (fig. 241C).

Primitive condition: The coronoid is typically not exposed widely on the lateral surface of the jaw (fig. 232C). This condition is in Proganochelys and most pleurodire outgroups.

Homoplasy: Araripemys independently acquires a wide coronoid in the MPC.

Discussion: Unfortunately, none of the Taphrosphyini lower jaws is well enough preserved to allow determination of this character. It only occurs in the tribe Bothremydini within the Bothremydidae .

122. Prearticular, fossa meckelii open anteriorly: no, closed by long angular-prearticular contact ( Euraxemys ) 5 0; yes, short prearticular-angular contact ( Bothremys ) 5 1.

Morphology: In the lower jaw of most turtles the sulcus cartilaginis meckelii is closed ventrally by the prearticular and angular, which meet in a long suture medial to the sulcus. In the tribes Bothremydini and Taphrosphyini , much of the fossa meckelii is open anteriorly, and the prearticular-angular contact is short (fig. 241D).

Primitive condition: The condition in Proganochelys , in which the prearticular-surangular suture begins below the fossa meckelii, is also found in chelids and baenids and is assumed to be the primitive condition for pleurodires.

Homoplasy: The Podocnemididae acquire the more open fossa meckelii independently of the infrafamily Bothremydodda ( Bothremydini + Taphrosphyini ).

Discussion: This is a synapomorphy for the infrafamily Bothremydodda ( Bothremydini + Taphrosphyini ). The problem with this character is the few lower jaws known, particularly for the Taphrosphyini .

123. Articular, processus retroarticularis: long and projecting posteriorly ( Bothremys ) 5 0; short or absent ( Pelusios ) 5 1; long and projecting posteroventrally ( Podocnemis ) 5 2.

Morphology: The lower jaw of bothremydids has a processus retroarticularis that projects straight posteriorly (state 0, fig. 241A). It is not an extremely long processus, as in some cryptodires like trionychids, but it is distinctly longer than in chelids, pelomedusids, and euraxemydids. Unfortunately, it is known for only 7 of 29 taxa of bothremydids. In podocnemidids, Brasilemys , and Hamadachelys , there is also a distinct processus retroarticularis, but it is directed posteroventrally (state 2) and lies below the axis of the jaw ramus ( Gaffney, 1979a: fig. 135).

Primitive condition: Proganochelys has a long processus retroarticularis ( Gaffney, 1990), but cryptodires generally lack one. It is also short or absent in pelomedusids, chelids, and euraxemydids. Although the presence of a processus seems to be primitive for turtles, its absence is primitive for pleurodires.

Homoplasy: The short or absent state of this character is present in Pelomedusoides primitively, but it is reversed for bothremydids. The Proganochelys condition is a process with a dorsally open pocket, not seen in bothremydids, supporting the idea that the bothremydid condition is a nonhomologous reversal.

This character is in Antunes and Broin (1988), Meylan (1996), and Lapparent de Broin (2000b).

124. Vertebrae, cervical ribs: present ( Proganochelys ) 5 0; absent ( Pelusios ) 5 1.

Morphology: Cervical vertebrae of pleurodires ( Williams, 1950; Hoffstetter and Gasc, 1969) lack ribs, even in the Jurassic Platychelys .

Primitive condition: The presence of cervical ribs is primitive for turtles as they occur in Proganochelys ( Gaffney, 1990) and primitive cryptodires ( Gaffney, 1996).

Homoplasy: None within pleurodires.

125. Vertebrae, cervical postzygapophyses fused: all separate ( Proganochelys ) 5 0; some fused ( Podocnemis ) 5 1.

Morphology: The postzygapophyses of cervicals in some pleurodires unite to form a combined, single curved surface for the articulation of the prezygapophyses ( Meylan, 1996: fig. 9).

Primitive condition: Separate postzygapophyses occur in Proganochelys and cryptodires, but within pleurodires the primitive condition is equivocal. The one cervical of Platychelys is unfused, as are all cervicals of Pelomedusidae . Within chelids, fusion is variable but absent in Emydura and Elseya . It is likely that the unfused condition is primitive for Pleurodira .

Homoplasy: A problem with this character is the recognition of the form of variability of postzygapophyseal fusion. All cervicals in Araripemys and Euraxemys show fusion, but in chelids and podocnemidids some cervicals in the column are fused and some are not, and this even varies among species. These are all treated as one character, but this may be too simplistic as it lumps a variety of morphologies that may not be homologous. However, to recognize more character states at this point will only make a series of autapomorphies.

Discussion: The cervicals of Euraxemys and Araripemys are known, but within Bothremydidae only a few cervicals of Taphrosphys sulcatus , Chedighaii barberi , and Cearachelys are known. Taphrosphys and Chedighaii show fused postzygapophyses, but the two probably posterior cervicals of Cearachelys do not. Because other cervicals of Cearachelys could be fused, it has been coded as ‘‘?’’, but all of the cervicals could lack fused zygapophyses.

This character is used in Meylan (1996) and Lapparent de Broin (2000b).

126. Vertebrae, cervical postzygapophyses elevated: separated and relatively low ( Proganochelys ) 5 0; extended posterodorsally on neural spine ( Podocnemis ) 5 1.

Morphology: The neural spine of the cervicals of living pleurodires ( Williams, 1950; Hoffstetter and Gasc, 1969) is extended posterodorsally and bears the postzygapophyses. This also occurs in bothremydids, Euraxemys , Araripemys , and Dortoka ( Lapparent de Broin and Murelaga, 1999) , but it is absent in Notoemys and Platychelys .

Primitive condition: The outgroups lack the extended neural spine.

Homoplasy: None known.

Discussion: Lapparent de Broin and Murelaga (1999) used this character in a dataset and figured a cervical for the skull-less taxon Dortoka . This character helps unite Dortoka with the Eupleurodira.

127. Vertebrae, cervical centra: amphicoelous, platycoelous ( Proganochelys ) 5 0; formed articulation is wider than high ( Platychelys ) 5 1; formed articulation is higher than wide ( Podocnemis ) 5 2.

Morphology: The wider than high condition (state 1) occurs in Platychelys and Notoemys ( Fernandez and Fuente, 1994: fig. 3). It also occurs within Selmacryptodira, but only within Eucryptodira. Unformed articular surfaces are primitive for Selmacryptodira, so the eucryptodiran condition is not reflected in the coding. Although there are few bothremydid cervicals known, they all have cervical centra that are higher than wide ( Cearachelys , Chedighaii , Taphrosphys ).

Primitive condition: Unformed central articulations are in Proganochelys , Palaeochersis , and Cryptodira , and this is the presumed primitive condition.

Homoplasy: None in the MPC, but eucryptodires evolve wide centra independent of Platychelys and Notoemys . The higher than wide condition only evolves once, and that is within the Pleurodira .

Discussion: State 1 is a synapomorphy for Notoemys + Platychelys . State 2 is a synapomorphy for Dortoka + Eupleurodira.

128. Vertebrae, cervical articulation pattern: amphicoelous, platycoelous ( Proganochelys ) 5 0; (2))3))4))5))6))7))8) ( Pelusios ) 5 1; (2((3((4((5))6))7((8) ( Emydura ) 5 2.

Morphology: The articulation patterns of procoelous, opithocoelous, biconvex, and biconcave cervical vertebrae developed particularly by Williams (1950) are based on the articulation surface of the cervical centrum. These are described in Vaillant (1881), Kasper (1903), Williams (1950), and Hoffstetter and Gasc (1969), who also provide an introduction to the turtle vertebral literature. The use of parentheses follows that of Williams (1950) and most authors describing vertebrae and symbolizes the shape of the central articulation.

Primitive condition: The unformed central surface ( Gaffney, 1990, 1996) is primitive for turtles, but Lapparent de Broin and Murelaga (1999) have argued that the chelid condition may be primitive for Pleurodira . According to those authors, Notoemys , Platychelys , and Dortoka all probably have state 2, the chelid articular pattern (see Lapparent de Broin and Murelaga, 1999: 153). This is based on only a few cervicals, but they do show that these taxa do not have the procoelous condition of cervicals 3–8. We have accepted the Lapparent de Broin and Murelaga (1999) hypothesis and coded Platychelys , Notoemys , and Dortoka as state 2, which then becomes the state primitive for Pleurodira , not the unformed (state 0) as used by Gaffney (1996). When the taxa lacking skull data are excluded, this issue becomes moot, as chelids are the only included taxon with the character.

Homoplasy: None. Note, however, that a complete vertebral series is as yet unknown for any bothremydid. Cearachelys has three cervicals of unknown position, all procoelous.

Discussion: The procoelous condition (state 1) is a synapomorphy for Pelomedusoides in the MPC, with or without shell-only taxa. It is also not affected by coding of Platychelys , Notoemys , and Dortoka with state 2 or with ‘‘?’’.

This character is used in Antunes and Broin (1988).

129. Vertebrae, caudal articulation pattern: platycoelous, amphicoelous ( Proganochelys ) 5 0; formed centra but articulations vary, opisthocoelous and procoelous ( Notoemys ) 5 1; all procoelous ( Pelusios ) 5 2.

Morphology: The central articulations of the caudals of pleurodires are figured in Tronc and Vuillemin (1974: pl. 13, figs. 4–6, Erymnochelys ), Gaffney (1990: fig. 130, Podocnemis ), Meylan (1996: fig. 2, Araripemys ), and Lapparent de Broin and Murelaga (1999: fig. 6, Dortoka ).

Primitive condition: As with cervicals, the presumed primitive condition is unformed: amphicoelous or platycoelous.

Homoplasy: None in the MPC, but the sparse distribution of even partially articulated tails may mask a more complex character distribution. Procoelous caudals, as well as many other patterns, appear with Cryptodira .

Discussion: Character state 2 is a synapomorphy for Chelidae + Pelomedusoides (Eupleurodira) and has been known for some time (it was used by Lapparent de Broin and Murelaga, 1999). Character state 1 is also from Lapparent de Broin and Murelaga (1999) and, although ambiguous, it may reflect a pattern of diverse central articulations that are at least not all procoelous. Nonetheless, this is a weak character state, known only from disarticulated caudals in Dortoka ( Lapparent de Broin and Murelaga, 1999) and two from Platychelys ( Bräm, 1965) that show opisthocoely as well as procoely.

130. Shoulder girdle, coracoid foramen: present ( Proganochelys ) 5 0; absent ( Podocnemis ) 5 1.

See Gaffney (1990) for description; also see Rougier et al. (1995).

131. Shoulder girdle, coracoid shape: coracoid a flat plate ( Proganochelys ) 5 0; coracoid columnar ( Podocnemis ) 5 1.

See Gaffney (1990) for description; also see Rougier et al. (1995).

132. Pelvic girdle, tenth thoracic centrum: not incorporated into sacrum ( Proganochelys ) 5 0; incorporated into sacrum ( Pelusios ) 5 1.

Morphology: Broin and Murelaga (1999: pl. 12, fig. 5: Peltocephalus ) figured the sacrum, showing this character, but this area is not well described in the literature (see also Fraas, 1913: Proterochersis ; Tronc and Vuillemin, 1974: Erymnochelys ; and Fernandez and Fuente, 1994: Notoemys ). The presence of the tenth thoracic rib in the sacrum can be determined by the suturing of the distal end to the ilium. The ilium itself is also sutured to the carapace.

Primitive condition: In Proganochelys ( Gaffney, 1990) the tenth thoracic rib is fused to the ninth costal, and this is presumably the primitive condition. In cryptodires, the tenth rib is usually free.

Homoplasy: None known, but the character is not known in most bothremydids and in many other extinct pleurodires.

Discussion: This character is a synapomorphy for Pleurodira and is presumably related to the sutured pelvis-carapace.

133. Pelvic girdle, pelvis sutured to shell: pelvis articulates only by ligaments, no sutures ( Proganochelys ) 5 0; all three bones of pelvis sutured to shell ( Pelusios ) 5 1.

Morphology: This character is figured in Rütimeyer (1873: pl. 8), Bräm (1965: pl. 1), Gaffney (1990: figs. 143, 144 for the pelvis alone of Podocnemis ), and Lapparent de Broin and Murelaga (1999: pl. 12). Considering how widely used this character is for Pleurodira , it is not well described in the literature.

Primitive condition: The outgroups Proganochelys and Cryptodira have a free pelvis, which is the primitive condition ( Gaffney, 1990). Rougier et al. (1995) described Palaeochersis as having a sutured pelvis. Examination of this material by the senior author shows that the pelvis in Palaeochersis is not sutured to the carapace or plastron, but rather it has been crushed dorsoventrally to produce a condition that can be mistaken for suturing. The pelvis and adjacent shell surfaces show no signs of sutures, rather they are very similar to the morphology in Proganochelys .

Homoplasy: None known.

Discussion: This is a long-used synapomorphy for Pleurodira .

Lapparent de Broin and Murelaga (1999: 150) added a second state, ‘‘ligamentous more firmly linked producing a true articulation scar’’, and coded it for Proterochersis in their dataset. Extensive study of the known material of Proterochersis by two of the authors (E.S.G. and P.A.M.) shows no difference in the type of attachment between pelvis and shell in Proterochersis and all other pleurodires, so we reject this second character state.

134. Pelvic girdle, dorsal part of ilium columnar: no, inclined with anterior and posterior processes ( Proganochelys ) 5 0; columnar, expanded mediolaterally ( Podocnemis ) 5 1.

Morphology: In pleurodires the ilium is a relatively thick columnar element, not flattened with anterior and/or posterior processes as in Proganochelys and cryptodires ( Gaffney, 1990: figs. 143, 144. It is also oriented more vertically in pleurodires than in other turtles.

Primitive condition: In Proganochelys the ilium is expanded anteriorly and posteriorly, forming a horizontally oriented, inclined structure. Cryptodires have a wide diversity of ilium shapes, but primitively they have an inclined, mediolaterally flattened shape, similar to Proganochelys .

Homoplasy: None known.

Discussion: This character attempts to reflect further information from the pelvic morphology of pleurodires. Perhaps it is correlated with the suturing of the pelvis to the shell.

135. Pelvic girdle, pelvis narrow: ilia and acetabula relatively far from midline ( Proganochelys ) 5 0; ilia and acetabula close to midline ( Podocnemis ) 5 1.

Morphology: Although this character is somewhat subjective, all pleurodires do have the ilial column and acetabulum closer to each other and to the midline than in other turtles (i.e., Proganochelys and cryptodires), producing a narrower pelvis. Presumably this is related to the sutured pelvis-shell. Gaffney (1990: fig 144) compared Proganochelys , a cryptodire, and a pleurodire pelvis, showing this condition.

Primitive condition: The wide pelvis of Proganochelys and cryptodires is primitive.

Homoplasy: None known; however, a complete pelvis is not preserved for most bothremydids and many other extinct taxa.

Discussion: This character is used by Lapparent de Broin and Murelaga (1999), it is another pleurodire synapomorphy.

136. Pelvic girdle, thyroid fenestra: separate ( Proganochelys ) 5 0; confluent ( Podocnemis ) 5 1.

Morphology: The thyroid fenestrae of Proganochelys , a pleurodire, and a cryptodire are figured in Gaffney (1990: fig. 144). In the advanced state the fenestra is not subdivided down the center by bone.

Lapparent de Broin and Murelaga (1999) divided this character in a series of states based on the fenestra size. Proterochersis has slightly larger fenestrae than does Proganochelys , but they are still relatively small and widely separated by bone. Rather than calling this a separate state, we identify both as primitive. Platychelys has fenestrae that are not separated by bone but are somewhat smaller than those in Eupleurodira. Rather than making this a separate state, we identify it as the advanced condition because the fenestrae are confluent.

Primitive condition: The small, widely separated fenestrae of Proganochelys are primitive.

Homoplasy: It is likely that the confluent thyroid fenestrae arose separately in cryptodires and pleurodires.

Discussion: This character is used in Rougier et al. (1995) and Lapparent de Broin and Murelaga (1999).

137. Humerus, shoulder on lateral side of head: present ( Proganochelys ) 5 0; absent ( Podocnemis ) 5 1.

Morphology: The humerus of pleurodires has a head that is hemispherical and slightly separated from the lateral and medial processes ( Zangerl, 1948: fig. 13, Chedighaii ; Gaffney, 1975a: fig. 12, Taphrosphys ; Gaffney, 1990: fig. 149, Podocnemis ; Fernandez and de la Fuente, 1994: fig. 6, Notoemys ). The curved shoulder found on the lateral side of the articular head in Proganochelys and cryptodires is absent in pleurodires.

Primitive condition: Proganochelys has a shoulder and an articular head that is not as spherical as in pleurodires. This is presumed to be primitive.

Homoplasy: None in the MPC; however, this character gets harder to identify when looking within various eucryptodire groups. Also, the humerus is not known for many bothremydids and other extinct pleurodires.

Discussion: This is a pleurodire synapomorphy. Proterochersis lacks a humerus, however.

138. Carapace, cervical scale: present ( Emydura ) 5 0; absent ( Podocnemis ) 5 1.

Morphology: The cervical or nuchal scale is an unpaired scale on the anterior margin of the nuchal bone of most turtles. Except in Hydromedusa (see Wood and Moody, 1976), it lies between the first marginal scales. It is absent in Pelomedusoides (fig. 265).

Primitive condition: A cervical scale is present in Proganochelys ( Gaffney, 1990) , in primitive cryptodires ( Gaffney, 1979b), in the primitive pleurodires Proterochersis , Platychelys , and Notoemys ( Fernandez and de la Fuente, 1994) , and in nearly all chelids. Presence of this scale is clearly the primitive condition.

Homoplasy: Loss of the cervical scale occurs independently within the Pleurodira (all members of the Pelomedusoides and one chelid, Elseya , fide Pritchard and Trebbau, 1984) and Cryptodira (some testudinids). It is variable within Dortoka ( Lapparent de Broin and Murelaga, 1999) .

Discussion: The absence of a cervical scale is a synapomorphy for the Pelomedusoides.

This character was used by Antunes and Broin (1988), Broin (1988), Gaffney (1988), Gaffney and Meylan (1988), Meylan (1996), Tong et al. (1998), and Lapparent de Broin and Murelaga (1999).

139. Carapace, nuchal bone width: nuchal bone two or more times wider than length ( Platychelys ) 5 0; nuchal bone width greater than length but less than two times ( Euraxemys ) 5 1; nuchal bone width approximately equals length ( Foxemys ) 5 2; nuchal width less than length ( Teneremys ) 5 3; nuchal bone greatly emarginated (unique to Araripemys ) 5 4.

Morphology: The width versus length of the nuchal bone in dorsal view (figs. 254– 274) is arbitrarily subdivided into four states. This character is used in Lapparent de Broin and Murelaga (1999); we have slightly modified it. Araripemys ( Meylan, 1996) has a uniquely emarginated nuchal, making it hard to compare with other turtles, so it is given state 4, unique to that taxon.

Primitive condition: Both Proganochelys and primitive cryptodires have a very wide (state 0) nuchal, and this occurs in Platychelys and Notoemys , but the character is indeterminate in Palaeochersis and Proterochersis .

Homoplasy: Quite a lot, within Selmacryptodira, Chelidae , Pelomedusidae , and Podocnemididae , but for this analysis the CI is 0.66. Nuchal width is often variable within a species. Nonetheless, some patterns are apparent in the MPC.

Discussion: The change from state 0 to state 1 is a synapomorphy for Dortoka + Eupleurodira, the parvorder Megapleurodira , and state 1 is primitive for nearly all groups in Eupleurodira. State 3 is unique to Teneremys and state 4 is unique to Araripemys . Foxemys and Polysternon are united by state 2.

This character is used in Lapparent de Broin (2000b), who agreed that the wide condition is primitive.

140. Carapace, pygal notch: present, wide and shallow ( Proganochelys ) 5 0; present, narrow and spherical (unique to Proterochersis ) 5 1; absent, margin smooth ( Podocnemis ) 5 2.

Morphology: The pygal notch in Proganochelys is figured in Gaffney (1990: figs. 69– 77) and in Palaeochersis by Rougier et al. (1995). The pygal notch of Proterochersis has not been figured. The absent condition can be seen in nearly all other turtles.

Primitive condition: Presumably, the presence of a pygal notch is primitive for turtles.

Homoplasy: None, although some cryptodires acquire an emarginated posterior carapace margin.

Discussion: The posterior carapace emargination or pygal notch in Proterochersis is so different from that in Proganochelys that a separate state is identified for it. In this we follow Lapparent de Broin and Murelaga (1999) who also used two states for the pygal notch. Rougier et al. (1995) used ‘‘marginals not separated by an anal notch’’, but we think that this obscures a more complex situation.

141. Carapace, neural series completeness: neurals reach suprapygal ( Euraxemys ) 5 0; to eighth costals ( Podocnemis ) 5 1; to seventh costals ( Foxemys ) 5 2; to sixth costals ( Chedighaii ) 5 3; neurals absent or discontinuous ( Araiochelys , Emydura ) 5 4.

Morphology: The neural bones of the carapace usually make up the midline of the shell between the nuchal and suprapygal. In most turtles they form a continuous series between the nuchal and suprapygal; in others, some or all costals meet on the midline. In some forms neurals are completely absent. The neurals reaching the suprapygal and forming a complete series is state 0 (fig. 254), reaching to the eighth costals is state 1 (fig. 272), reaching to the seventh costals is state 2 (fig. 259), reaching to the sixth costals is state 3 (fig. 264), and discontinuous or absent neurals is state 4 (fig. 263).

Primitive condition: The neural series of Proganochelys is not completely known. However, the neural series in such primitive cryptodires as Kayentachelys and Pleurosternon is complete (state 0), as are those of Platychelys ( Bräm, 1965) and Notoemys ( Fernandez and de la Fuente, 1994) . The neural series is complete among Pelomedusoides in Euraxemys , Cearachelys , and some Araripemys .

Homoplasy: Interruption of a complete neural series by costals meeting on the midline occurs within Podocnemididae and Bothremydidae . Among the Cryptodira , trionychids, kinosternids, and dermatemydids all have posterior costals meeting on the midline. Cearachelys is apparently a reversal in which the complete neural series is reacquired. The pelomedusids independently lose a full neural series. The CI is 0.42.

Discussion: State 1 is a synapomorphy for the epifamily Podocnemidinura, and state 2 has equivocal support for Bothremydidae , except for Cearachelys , which has a complete set of neurals.

A variant of this character is used in Antunes and Broin (1988), Meylan (1996), Tong et al. (1998), Lapparent de Broin and Murelaga (1999), and Lapparent de Broin (2000b).

142. Carapace, iliac scar position: iliac scar absent ( Chelydra ) 5 0; iliac scar restricted to costals 7 and 8 ( Pelusios ) 5 1; iliac scar on costals 7, 8, and suprapygal ( Taphrosphys ) 5 2.

Morphology: The ilium of pleurodires articulates with the bones of the carapace in a sutural contact. The elements with which the ilium articulates vary among species in the suborder ( Lapparent de Broin and Murelaga, 1999: pl. 12). In some the articulation is only with costals 7 and 8, and in others the suprapygal and or peripheral elements are also involved. For example, in Notoemys the ilium articulates with the eighth costals and suprapygal, while in Platychelys it articulates with the eighth costals, suprapygal, and eleventh peripherals ( Fernandez and de la Fuente, 1994). State 1 is the scar on costals 7 and 8 ( Lapparent de Broin and Murelaga, 1999: pl. 12, fig. 5), and state 2 is the scar on costals 7, 8, and the suprapygal (fig. 265).

Primitive condition: For turtles, the absence of a scar is primitive; the two states are derived. Contact of the ilium in Platychelys includes the eighth costals, suprapygal, and eleventh peripherals ( Fernandez and Fuente, 1994). In Araripemys , the ilium contacts the seventh and eighth costals and the suprapygal. Thus, it appears that contact to the suprapygal may be primitive for pleurodires. Antunes and Broin (1988) argued that the inclusion of the suprapygal in this contact in bothremydids is a reversal to the primitive condition. However, Lapparent de Broin (2000b: 45) considered suprapygal contact primitive for Pelomedusoides. We are uncertain of the evidence that contact to the suprapygal is lost and regained.

Homoplasy: There is individual variation in this contact within the specimens of Chedighaii barberi figured by Zangerl (1948, one figure shows suprapygal contact and one does not) and within recent specimens of Podocnemis expansa and P. unifilis seen by the senior author.

Discussion: Lapparent de Broin and Werner (1998) indicated that broad contact with costal 8 and narrow contact with neural 7 and the suprapygal are the general condition for bothremydids. However, contact in Kurmademys is on costals 7 and 8 only (ISIR 278), so the issue is still ambiguous. The area is known in only a few bothremydids.

A related character is used in Lapparent de Broin and Murelaga (1999), which reflects the scar shape. It is also used in Antunes and Broin (1988) and Lapparent de Broin (2000b).

143. Carapace, first costal length: costal 1 shorter or equal to twice the length of costal 2 ( Euraxemys ) 5 0; costal 1 more than twice the length of costal 2 ( Foxemys ) 5 1.

Morphology: The elements of the anterior portion of the shell of turtles include a first costal bone that is longer anteroposteriorly than the more posterior costals (state 1, fig. 259). In the most primitive turtles, the first costal is not longer than more posterior costals. However, in many forms the first costal is anteroposteriorly long, often two or more times longer than the second costal. This is a gradational character that is arbitrarily defined as the condition where the anteroposterior length of costal 1 is more than twice that of the second costal. This identifies the more extreme end of a gradational character and may be slightly more objective than dividing the character into more states.

Primitive condition: In Proganochelys and Platychelys the first costal is approximately as long as those of the more posterior costals. This is the best evidence for the primitive condition of this character in pleurodires.

Homplasy: The first costal is anteroposteriorly longer than the more posterior costals in most turtles. However, first costals that are more than twice as long as the second are not common, but they do occur in Dortoka , Kurmademys , Bothremydini , Taphrosphyini , Podocnemis , and some chelids. In the MPC, the CI is 0.33 and the character arises three times.

Discussion: In the MPC, this character is synapomorphic for the infrafamily Bothremydodda (consisting of the tribes Bothremydini and Taphrosphyini ), but its rather subjective nature is a problem.

144. Carapace, position of four-sided neural: neural 1 ( Euraxemys ) 5 0; neural 2 ( Cearachelys ) 5 1; neural 3 ( Araripemys ) 5 2; four-sided neural absent ( Platemys ) 5 3.

Morphology: Characterizing the neural morphology is a very dubious business. Multiple attempts have been made to characterize neural bone morphology, and none has been satisfactory. One way is to count the sides contacting surrounding elements, considering each contact a side ( Auffenberg, 1974). This produces a formula running from anterior to posterior, as in Euraxemys : 4-6-6- 6-6-6-6-6. In this case the four-sided or quadrangular neural is first (state 0, fig. 265). The four-sided neural may be the second, as in Cearachelys : 6-4-6-6-6-6-6-4 (state 1, fig. 258), or it may be the third neural, as in Araripemys : 6-6-4-6-6-6-6-4 (state 2; Meylan, 1996: fig. 1). In most Pleurodira , the posterior neurals are all six-sided, but the position of the reversed or four-sided neural varies. In some cases, including some pleurodires, reversal in neural orientation occurs via two consecutive five-sided neurals, as in Brasilemys ( Lapparent de Broin, 2000b: fig. 4).

Primitive condition: Only neurals 4–7 are known for Proganochelys , and these all seem to be six-sided. Meylan and Gaffney (1989) have argued that the primitive condition for Casichelydia is 4-6-6-6-6-6-6, and this is what we have chosen for state 0. However, Kayentachelys ( Gaffney et al., 1987) has the second neural four-sided (although there is individual variation), Platychelys and Notoemys have the second and fourth neurals four-sided, and Proterochersis is unknown, so the primitive neural pattern is still unclear.

Homoplasy: In the MPC, the four-sided first neural is primitive and pervasive throughout the pleurodires, with the other conditions mostly appearing independently, with a CI of 0.75. The neural 2 four-sided condition does unite Cearachelys with the two possible Galianemys shells, AMNH 30550 and AMNH 30551. In any case, these characters occur widely as individual variation among many species of turtles, and they can be highly variable within species in some families ( Auffenberg, 1976; Meylan, 1984, 1987).

Discussion: Lapparent de Broin and Murelaga (1999) used a variation of this character in their dataset. Lapparent de Broin (2000b: 45) implied that state 1 may be primitive for Pelomedusoides. In the MPC analyzed here, state 1 is synapomorphic for Cearachelyini (supporting the AMNH 30550 and AMNH 30551 shells as belonging to Galianemys ) and, independently, for Notoemys + Platychelys .

145. Carapace, neural series pattern: irregular, neurals 2 and 4 quadrangular, alternating in width ( Platychelys ) 5 0; irregular, width even ( Kayentachelys ) 5 1; regular, most hexagonal, coffin-shaped ( Podocnemis ) 5 2; neurals absent ( Platemys ) 5 3; neurals discontinuous ( Araiochelys ) 5 4.

Morphology: This is another feeble attempt to use neural diversity as a character. Lapparent de Broin and Murelaga (1999) and Fuente and Iturralde-Vinent (2001) used ‘‘regular’’ versus ‘‘irregular’’ to distinguish the common casichelydian pattern of one quadrangular neural (in the anterior part of the series) followed by hexagonal, coffin-shaped neurals, from the ‘‘irregular’’ neural pattern of alternating contacts and widths. Lapparent de Broin and Murelaga (1999) used the ‘‘irregular’’ as one state, but this is a diverse group, and we have subdivided it. The taxa placed in this group by Lapparent de Broin and Murelaga (1999) all have more than two quadrangular neurals in common, but the position and width of these neurals vary. They also have a high occurrence of asymmetry, with many five- and seven-sided neurals. Kayentachelys has the neural width the same for the series (state 1, Gaffney et al., 1987), while Notoemys ( Fernandez and Fuente, 1994) and Platychelys ( Lapparent de Broin, 2001) have an alternating width pattern (state 0). Dortoka ( Lapparent de Broin and Murelaga, 1999) , irregular but somewhat unique, is also coded as state 0. The common pattern, state 2, is seen in the Eupleurodira (fig. 265). Araiochelys has the discontinuous condition, state 4 (fig. 263).

Primitive condition: The primitive condition would be ‘‘irregular’’ as one state, but as we have subdivided them, there is no clear primitive pattern. We have arbitrarily chosen the Platychelys condition as ‘‘0’’. Note that the neurals known for Proganochelys (4–7) do show the ‘‘regular’’, coffin-shaped pattern, so state 2 could be argued as primitive. Because many neurals are missing in Proganochelys , it is coded as ‘‘?’’.

Homoplasy: The ‘‘regular’’, state 2 condition has evolved at least once within pleurodires and at least once within cryptodires.

Discussion: This character is from Lapparent de Broin and Murelaga (1999), but they basically recognized only two states outside of chelids. We have recognized more states in the ‘‘irregular’’ condition, because the absence of ‘‘regular’’ includes a series of states that are not the same. It is difficult to specifically define patterns in common because bilateral asymmetry is common among the early pleurodire shells. Nonetheless, the ‘‘regular’’ pattern (state 2) does support a group, the Eupleurodira.

146. Carapace, neural number: more than eight ( Kayentachelys ) 5 0; eight ( Cearachelys ) 5 1; seven ( Foxemys ) 5 2; six or fewer ( Kurmademys ) 5 3.

Morphology: The neural bones are easily counted, but some may not be fused to the underlying vertebral spines and might not be considered separate neurals (e.g., the posterior neurals of Kayentachelys and Platychelys , but this can also occur as an individual variation throughout chelonians). This is the argument for making this a separate character from character 141 (see below).

Primitive condition: This is difficult to assess because Proganochelys and Palaeochersis lack complete series. We have chosen the ‘‘more than eight’’ as primitive because of its occurrence in outgroups such as Kayentachelys and Platychelys .

Homoplasy: Even though the CI is 0.42, homoplasy is extensive for these character states due to variation within chelids, pelomedusids, cryptodires, Platychelys , and Araripemys . The loss of neurals within Bothremydidae could be more widespread than shown in the MPC, because the generically indeterminate shells from Tunisia (e.g., ‘‘ Eusarkia ’’, ‘‘ Gafsachelys ’’) show neural loss, and some of them seem to be Taphrosphyini .

Discussion: This is very similar to character 141, which might be considered just another way of counting neurals. Deleting either results in the same MPC, but deleting them does affect the placement of the poorly known shell-only taxa.

Although this is a widely variable character, the MPC shows that seven neurals (state 2) is a synapomorphy for podocnemidids + bothremydids + Teneremys . Similarly, despite independent acquisition in cryptodires and elsewhere, eight neurals (state 1) is a synapomorphy for pleurodires, with Proterochersis unknown.

A version of this character appears in Lapparent de Broin and Murelaga (1999).

147. Carapace, length of contact between peripheral 1 and costal 1 in dorsal view: no contact due to large nuchal ( Dortoka ) 5 0; wide contact so that anterior margin of peripheral 1 is less than twice the length of costal 1 contact ( Rosasia ) 5 1; narrow contact, so that anterior margin of peripheral 1 is twice or more the length of costal 1 contact ( Foxemys ) 5 2; no contact due to small nuchal (unique to Araripemys )5 3.

Morphology: This character is a modification of one used by Lapparent de Broin and Murelaga (1999). Their character states are: ‘‘wide contact (0), wide contact or no contact (1), wide or narrow contact (2)’’ ( Lapparent de Broin and Murelaga, 1999: 150). Apart from their not inconsiderable entertainment value, the state characterizations are useless. In an effort to squeeze something objective out of this mess, we have arbitrarily divided a fairly continuous character into wide (state 1, fig. 261) and narrow (state 2, fig. 259) by comparing the relative width of the anterior and posterior margins.

Primitive state: The character is not known for Proganochelys , Australochelys , Palaeochersis , and Proterochersis , and there is no obvious distribution pattern in the MPC for a primitive condition for turtles. The state 0 was chosen arbitrarily.

Homoplasy: Lots. The width of the contact between the first peripheral and the first costal is highly variable in turtles. However, the condition in Dortoka with anterolateral notches in the nuchal for the first peripherals seems to be autapomorphic. Although the CI is merely 0.33 due to rampant equivocality, state 2 appears six times independently in the MPC, and state 1 appears seven times.

Discussion: The narrow contact could be a synapomorphy for eupleurodires, although it is variable in pelomedusids (and probably in chelids, if enough shells are examined). The first peripheral shape could be characterized in other ways, comparing margin length with costal length, for example, but there does not seem to be any advantage in doing this.

This character is from Lapparent de Broin and Murelaga (1999) and, even as modified, we use it with reservations, but it reflects our effort to include as much of Lapparent de Broin’s work on shells as possible.

148. Carapace, axillary process contacts first costal: no contact ( Proganochelys ) 5 0; contact present, separated from posterior margin, may be placed in middle of costal ( Chedighaii ) 5 1; contact present, placed on posterior border close to second costal ( Dortoka ) 5 2.

Morphology: The dorsal process of the hyoplastron may contact the first costal and form a sutural surface. Lapparent de Broin and Murelaga (1999) used this character as two states, and we follow that here. State 1 is a contact area that is often curved and lies in the middle of the costal or at least separated from the posterior edge ( Zangerl, 1948: fig. 7, Chedighaii ). In state 2 ( Lapparent de Broin and Murelaga, 1999: pl. 3, fig. 12a), the axillary buttress lies on the posterior edge of costal 1, very close to the contact with costal 2.

Primitive condition: Proganochelys and Kayentachelys have no axillary buttress contact on the first costal, and this is presumably the primitive state.

Homoplasy: This character has reversals within chelids, pelomedusids, and podocnemidids, as well as within Taphrosphys sulcatus . Probably a more extensive study of recent pleurodire shells would show more withinspecies variation, although the CI is 0.5.

Discussion: Although the character shows a fair amount of homoplasy at the generic and specific levels, state 1 is synapomorphic for the Bothremydidae and state 2 pulls together Dortoka and eupleurodires.

Lapparent de Broin and Werner (1998) have an extensive discussion of the axillary buttress meeting the carapace in pleurodires. Unfortunately, most of the literature on pleurodire shells lacks figures and detailed morphology on the internal shell morphology, such as buttress attachments, limiting the value of this and other internal shell characters.

149. Carapace, axillary process extent on ventral surface of peripherals: reaches peripheral 2 ( Platychelys ) 5 0; reaches anterior margin of peripheral 3 ( Emydura ) 5 1; reaches onto main body or posterior edge of peripheral 3 or is restricted to peripheral 4 ( Chedighaii ) 5 2.

Morphology: The axillary process of the hyoplastron can extend anteriorly onto the peripherals, as well as medially, onto the first costal. The character states are continuous but can be divided because it is easy to see the sutural area separated by interperipheral sutures. State 0, the buttress reaching the second peripheral, can be seen in Proganochelys ( Gaffney, 1990: fig. 102). In state 1 the buttress reaches nearly to the peripheral 2–peripheral 3 contact, which can be seen in some chelids and some podocnemidids ( Tronc and Vuillemin, 1974: pl. 13, fig 1). In state 2 the buttress reaches the main body of peripheral 3 ( Zangerl, 1948: fig. 4).

Primitive condition: The anterior extension of the axillary buttress to peripheral 2 is found in Proganochelys , and it seems to be primitive for turtles.

Homoplasy: Except for homoplasy within chelids and podocnemidids, and a reversal in MNHN GDF 801, this character shows no reversals or homoplasy within pleurodires.

Discussion: The state 2 condition, restriction of the axillary contact to the middle or posterior part of peripheral 3, is a synapomorphy for Dortoka + Eupleurodira.

This character is used by Lapparent de Broin and Werner (1998) and Lapparent de Broin and Murelaga (1999).

150. Carapace, inguinal buttress: short or absent, not making extensive contact with costal 5 ( Proganochelys ) 5 0; inguinal buttress of hypoplastron contacts costal 5, extending medially onto it ( Chedighaii ) 5 1.

Morphology: This character is best seen in a ventral view of the fifth costal, which shows the articulation surface. When the plastron is present, the contact area is usually obscured. Proganochelys shows state 0 ( Gaffney, 1990: fig. 76) and Lapparent de Broin and Murelaga (1999) show state 1 for Dortoka ( Lapparent de Broin and Murelaga, 1999: fig. 1) and Polysternon ( Lapparent de Broin and Murelaga, 1999: pl. 5).

Primitive condition: Proganochelys has no inguinal buttress contact (state 0).

Homoplasy: State 1 has evolved at least three times in the MPC and is variable within the Pelomedusidae and the Podocnemididae .

Discussion: This character supports uniting Teneremys and the Podocnemidoidea . This character is used by Lapparent de Broin and Murelaga (1999) and Fuente and Iturralde-Vinent (2001).

151. Carapace, supramarginal scales: 12 or more pairs ( Proganochelys ) 5 0; 3 pairs ( Proterochersis ) 5 1; absent ( Podocnemis ) 5 2.

Morphology: Supramarginal scales showing states 0 and 1 are figured in Gaffney (1990: fig. 105). Their absence, state 2, can be seen in figure 264.

Primitive condition: Proganochelys and Palaeochersis have 12 or more supramarginals.

Homoplasy: The most parsimonious explanation for the distribution of this character in the MPC requires the reappearance of supramarginals in Platychelys and Proterochersis . The absence of supramarginals in nearly all cryptodires requires that state 1, three pairs of supramarginals in the earliest pleurodires, evolve from state 2, no supramarginals. Although this seems unlikely, it is apparently the case in Macrochelys (P.A.M., personal obs.), and the loss of supramarginals in cryptodires and pleurodires independently, the more likely hypothesis as far as supramarginals are concerned, is not supported by the MPC.

Discussion: Total loss of supramarginals, state 2, is a synapomorphy for Casichelydia in the MPC. State 1, three supramarginals, evolves twice (or is lost once) within that group. As discussed above, the reappearance of supramarginals seems unlikely, and it is only one step away from the independent loss of them within pleurodires and in cryptodires.

This character was used for Eupleurodira by Gaffney and Meylan (1988) and in datasets by Lapparent de Broin and Murelga (1999) and Fuente and Iturralde-Vinent (2001).

152. Carapace, vertebral scale width: equal to or wider than pleural scales ( Proterochersis ) 5 0; narrower than pleural scales ( Foxemys ) 5 1.

Morphology: State 0, wide vertebrals, is figured in Gaffney (1990: fig. 105); narrow vertebrals, state 1, is figured in figure 264. The wide vertebrals of Proterochersis , Platychelys , and Notoemys are distinct from Dortoka and Eupleurodira, with their narrow ones.

Primitive condition: Proganochelys has wide vertebrals, state 0.

Homoplasy: Narrow vertebrals have evolved twice, once within Selmacryptodira and once within Pleurodira .

Discussion: Some specimens of Platychelys seem to have vertebral scales that are about as wide as some of the pleural scales, but others have wider vertebrals.

This character is a synapomorphy for Dortoka + Eupleurodira.

153. Carapace, vertebral scale 1 reaches anterior margin of shell: no, first marginal scales meet on midline ( Podocnemis ), or cervical scale is present (chelids) 5 0; yes ( Araripemys ) 5 1.

Morphology: In most turtles the first marginal scales and the cervical scale or marginal scales alone (Pelomedusoides) intervene between the first vertebral scale and the shell margin. In MNHN GDF 801 (‘‘ Platycheloides cf. nyasae ’’ of Broin, 1980: pl. 1) and Araripemys , the first marginal scales do not meet on the midline, and the first vertebral reaches the anterior shell margin ( Meylan, 1996: fig. 1).

Primitive condition: First marginal scales meeting medially occurs in all Pelomedusoides outgroups.

Homoplasy: The character occurs twice, independently, in the MPC.

Discussion: This character does not form groups in the MPC, more evidence of staunch objectivity. Or poor judgment.

154. Carapace, nuchal embayment: absent ( Emydura ) 5 0; present ( Chedighaii ) 5 1.

Morphology: The anterior margin of the nuchal bone is curved, concave anteriorly, and centered on the midline in this character (fig. 264). When the margin is straight or convex anteriorly, the character is absent (fig. 265). A slight groove where the sulcus separates the first two marginals is not an embayed nuchal.

Primitive condition: Although the anterior margin of the nuchal is recessed from the peripherals in Proganochelys , and does not look like the condition in bothremydids, we have coded it as embayed. Kayentachelys also has a concave edge to the nuchal, but it has a different shape from the incised V-shape of Chedighaii .

Homoplasy: The embayed condition seems to have arisen at least six times, as reflected in the low CI of 0.2.

Discussion: The embayed nuchal occurs in three of the five Bothremydini species in which it is known, so it might be interpreted as a tribe Bothremydini synapomorphy. It supports uniting Teneremys and MNHN GDF 801.

This character is in Antunes and Broin (1988).

155. Carapace, first thoracic rib: large (close to size of first costal rib) and separated form first costal rib ( Platychelys ) 5 0; small (smaller than first costal rib) and closely attached to first costal rib ( Podocnemis ) 5 1.

Morphology: State 0 is clear in Proganochelys ( Gaffney, 1990: figs. 76, 77) in which the first thoracic rib is as large as the first costal (5 second thoracic) rib and is widely separated from it. Proterochersis , Notoemys , and Platychelys have smaller first thoracic ribs that are closer to the first costal rib than in Proganochelys , but they are not as small or as close as in other pleurodires. Furthermore, the degree of proximity of the two ribs and the relative size of thoracic rib 1 vary among Proterochersis , Platychelys , and Notoemys . Lapparent de Broin and Murelaga (1999) and Fuente and Iturralde-Vinent (2001) used four character states to reflect these differences. However, this only produces unique, autapomorphic character states for Platychelys and Notoemys , losing some potential information. We take a more inclusive view of first thoracic rib size and only use two states. State 1 is figured in Tronc and Vuillemin (1974: pl. 13, fig 1).

Each rib that is associated with a thoracic vertebra can be identified by the number of that vertebra. After the first thoracic rib, each rib is also fused distally to a costal bone and can be identified as a costal rib of a particular number. Because the first thoracic rib has no associated costal bone, the numbering of the following ribs is offset so that thoracic rib 2 is also costal rib 1, thoracic rib 3 is costal rib 2, and so on. There is no preferred identification of the rib heads as to thoracic or costal.

Primitive condition: Proganochelys shows the primitive state.

Homoplasy: None known, but Notoemys and Platychelys may be considered equivocal because the character is gradational.

Discussion: The eupleurodires + Dortoka have state 1; Notoemys may also have this state, however.

This character is used in Gaffney et al. (1991), Lapparent de Broin and Murelaga (1999), and Fuente and Iturralde-Vinent (2001).

156. Carapace, costovertebral tunnel: wide anteriorly and posteriorly only ( Proganochelys ) 5 0; wide for entire length ( Platychelys ) 5 1; narrow for entire length ( Chedighaii ) 5 2.

Morphology: The costovertebral tunnel is the space formed between the free part of the thoracic rib and the overlying costal bone; the thoracic centrum defines it medially. State 0 is figured in Gaffney (1990: fig. 78), and state 1 is figured in Bräm (1965: pl. 1, fig. 5).

Primitive condition: The Proganochelys condition, state 0, is unique to that taxon at present, so the primitive condition is ambiguous.

Homoplasy: The entire length being wide occurs only in Platychelys and Notoemys , although Chelus and some emydids approach this condition.

Discussion: The conditions of the costovertebral tunnel are not identical in Proganochelys and Proterochersis . Proterochersis has a wide tunnel, not reduced along costals 2–6 as in Proganochelys , but not as wide as in Notoemys and Platychelys .

This character is used in Lapparent de Broin and Murelaga (1999) and Fuente and Iturralde-Vinent (2001).

157. Carapace, articulation facet on thoracic rib 1: absent, anterior edge smooth ( Proganochelys ) 5 0; swollen facet or tubercle on anterior margin of first thoracic rib ( Platychelys ) 5 1.

Morphology: On the medial end of the first thoracic rib, there is a swollen tubercle that looks like an articulation facet ( Bräm, 1965: pl. 1, fig. 5). This occurs only in Platychelys and Notoemys . Although it is close to the centrum articulation, it does not seem to be for a cervical articulation. It is probably an articulation site for the dorsal process of the scapula.

Primitive condition: The absence of a facet, as in Proganochelys , is primitive.

Homoplasy: None.

Discussion: This character is a synapomorphy for Notoemys + Platychelys .

158. Plastron, mesoplastra: present and meet on midline ( Proganochelys ) 5 0; present, wider than long ( Platychelys ) 5 1; present, roughly equidimensional ( Chedighaii ) 5 2; absent ( Emydura ) 5 3.

Morphology: The medially meeting mesoplastra, state 0, are figured in Gaffney (1990: figs. 92, 105). State 1, transversely elongate mesoplastra, are figured in Bräm (1965: pl. 1, fig. 2) and Lapparent de Broin (2001: fig. 1). The equidimensional, that is, rounder, mesoplastra can be seen in figure 264.

Primitive condition: Medially meeting mesoplastra occur in Proganochelys and seem to be primitive for turtles. The two pairs of medially meeting mesoplastra seen in Proterochersis seem to be an autapomorphy for this taxon.

Homoplasy: Medially meeting mesoplastra have evolved twice, with one being a reversal in Pelusios . Among pleurodires, mesoplastra have been lost at least three times: in Dortoka , Araripemys , and chelids. In cryptodires, they may have been lost only once.

Discussion: Transverse mesoplastra not meeting on the midline are a synapomorphy for Platychelys + Notoemys , although there is another occurrence within extinct chelids and a similar condition exists in Pelusios broadleyi .

If the equidimensional shape of Pelomedusa is primitive for Pelomedusidae , then state 2 is synapomorphic for Pelomedusoides; if not, state 2 is synapomorphic for the magnafamily Podocnemidera.

This character is used in Gaffney et al. (1991), Rougier et al. (1995), Lapparent de Broin and Murelaga (1999), and Fuente and Iturralde-Vinent (2001).

159. Plastron, trapezoidal entoplastron: entoplastron arrow-shaped with posterolateral processes ( Proganochelys ) 5 0; entoplastron more trapezoidal ( Podocnemis ) 5 1.

Morphology: The entoplastron varies in shape in turtles, and this character is one aspect. The arrow-shaped entoplastron, state 0, is figured in Gaffney (1990: fig. 92), and the more trapezoidal shape is in figure 264.

Primitive condition: Proganochelys has state 0.

Homoplasy: Araripemys has a reversal to something similar to, but not identical to, the primitive condition.

Discussion: The trapezoidal entoplastron, even though it may be somewhat irregularly shaped (e.g., Kayentachelys , Proterochersis ), is a casichelydian synapomorphy.

This character is used in Lapparent de Broin and Murelaga (1999).

160. Plastron, epiplastra meet on midline: no, at least in ventral view ( Proganochelys ) 5 0; yes, meet on midline in ventral view ( Foxemys ) 5 1.

Morphology: The separated condition, state 0, is figured in Gaffney (1990: figs. 92, 105) and the midline meeting condition in figure 264. Lapparent de Broin and Murelaga (1999) differentiated meeting in ventral view versus meeting in dorsal view, but the absence of a dorsal contact would only be relevant for Proganochelys and Palaeochersis and the condition is only known for Palaeochersis (which has the epiplastra separated dorsally and ventrally).

Primitive condition: Presumably the complete separation of epiplastra by the entoplastron is primitive for turtles, as it is likely in Proganochelys and occurs in Palaeochersis and primitive cryptodires ( Kayentachelys ).

Homoplasy: Cryptodires evolve medially meeting epiplastra independently of pleurodires.

Discussion: This character is a synapomorphy for the parvorder Megapleurodira , consisting of all pleurodires minus Proterochersis . This character is used in Lapparent de Broin and Murelaga (1999) and Fuente and Iturralde-Vinent (2001).

161. Plastron, dorsal epiplastral process: large, reaching or nearly reaching carapace ( Proganochelys ) 5 0; small or absent ( Podocnemis ) 5 1.

Morphology: The dorsal process of the epiplastron is figured in Gaffney (1990: figs. 92–97). The absent condition can be seen in all the Pelomedusoides (e.g., Rütimeyer, 1873: pl. 8, fig. 6; Lapparent de Broin and Murelaga, 1999: pl. 2, fig. 5b).

Primitive condition: Proganochelys , Palaeochersis , and Proterochersis have the presumed primitive condition of a large dorsal process of the epiplastron.

Homoplasy: The dorsal process has been lost independently in cryptodires and within pleurodires.

Discussion: This character is a synapomorphy for the parvorder Megapleurodira (all Pleurodira minus Proterochersis ) and for the Cryptodira .

This character is used in Rougier et al. (1995) and Fuente and Iturralde-Vinent (2001).

162. Plastron, intergular scales: two (one pair) intergular scales ( Proterochersis ) 5 0; one intergular scale on midline ( Foxemys ) 51.

Morphology: A pair of intergulars (state 0) is figured in Gaffney (1990: fig. 105). A single intergular can be seen in figure 264.

Primitive condition: Proganochelys has one pair of intergulars.

Homoplasy: None.

Discussion: The single intergular is a synapomorphy for the parvorder Megapleurodira (all pleurodires minus Proterochersis ). This character is used in Lapparent de Broin and Murelaga (1999) and Fuente and Iturralde-Vinent (2001).

163. Plastron, gular projections: present ( Proganochelys ) 5 0; absent ( Podocnemis ) 5 1.

Morphology: The conical anterior processes on the epiplastra and entoplastron that are formed where the gular scales are attached can be seen in Gaffney (1990: fig. 105). Figure 264 , state 1, shows the smooth margin when the projections are absent.

Primitive condition: Proganochelys and Proterochersis have gular projections.

Homoplasy: The projections have been lost independently in cryptodires and within pleurodires. Lapparent de Broin (2000b) argued on the basis of morphology that the gular projections of Proterochersis and Proganochelys are not homologous, but this is ambiguous in the MPC.

Discussion: This character is synapomorphic for the Megapleurodira , which is all Pleurodira minus Proterochersis .

164. Plastron, anterior lobe of plastron short, wide at base, and semicircular in outline: width/length of 2.0 or less ( Podocnemis ) 5 0; width/length of more than 2.1 ( Chedighaii ) 5 1.

Morphology: The plastron in primitive turtles like Proganochelys , Proterochersis , and Kayentachelys includes a long and narrow anterior lobe ( Gaffney, 1990: fig. 105). However, the anterior lobe of the plastron in bothremydids and some other pleurodires is much shorter than its width across the base (state 1, fig. 264). Homology of this character is difficult to argue, howev- er, because of the simple nature of the character. This shape has been used as a character for the Bothremydidae ( Antunes and Broin, 1988) .

Primitive condition: Clearly a plastron in which the anterior lobe is longer than wide is the primitive condition for turtles. Within the Pleurodira , chelids, pelomedusids, Araripemys , Euraxemys , and most podocnemidids have longer anterior plastral lobes that are longer than half the width at the base.

Homoplasy: Bairdemys , a shweboemydine podocnemidid, shows a very similar, short, rounded, anterior plastral lobe that is broad across the base (fig. 275). In addition, a reversal to the longer, primitive condition occurs in the bothremydid Polysternon .

Discussion: The addition of Cearachelys and the possible Galianemys, AMNH 30550 and AMNH 30551, to the shells known for bothremydids indicates that a short anterior plastral lobe is not universal for the family (fig. 314). The anterior lobe of Kurmademys is not completely known, but the hyoplastra have parallel lateral sides, which suggests that this form also did not have the very short, rounded anterior lobe seen in other bothremydids. Thus it appears that the short anterior plastral lobe is more likely a synapomorphy for the infrafamily Bothremydodda ( Bothremydini and Taphrosphyini ).

This character is used in Antunes and Broin (1988) and Lapparent de Broin and Werner (1998).

165. Plastron, pectoral scales on entoplastron: absent ( Euraxemys ) 5 0; present ( Foxemys ) 5 1.

Morphology: The bony plastron of turtles is primitively covered by a set of seven plastral scales. The fourth set, the pectorals, covers most of the hyoplastron and, in the most primitive of turtles, the mesoplastron (state 0; Gaffney, 1990: fig. 105). In species in which the first three pairs of plastral scales are either reduced in size or absent, the pectorals may reach anteriorly to the entoplastron (state 1, fig. 259).

Primitive condition: In Proganochelys , Proterochersis , Platychelys , Notoemys ( Fernandez and Fuente, 1994) , Euraxemys , Araripemys , and the Pelomedusidae , the pectoral scales are located posterior to the entoplastron. This seems to be the primitive condition. The primitive condition also appears to be present in the bothremydids, Cearachelys , and the probable Galianemys, AMNH 30050 and AMNH 30551.

Homoplasy: Overlap of the pectoral scales onto the entoplastron may be a synapomorphy for the superfamily Podocnemidoidea , but if this is the case, then it is reversed at least in Cearachelys , Galianemys , and Rosasia . The pectoral scales also reach the entoplastron in some members of the Chelidae ( Emydura ), Trionychoidea (some Adocus , Basilemys , and Nanhsiungchelys ), and Emydidae ( Clemmys ). The CI is 0.33.

Discussion: Among the bothremydids for which we have data, only Cearachelys , Galianemys , and Rosasia lack this overlap.

A variant of this character is used in Lapparent de Broin and Murelaga (1999).

166. Plastron, pectoral scales on epiplastron: no, far behind epiplastra ( Cearachelys ) 5 0; yes, on epiplastra or epiplastronhyoplastron suture ( Foxemys ) 5 1.

Morphology: The bony plastron of turtles is primitively covered by a set of seven plastral scales. The fourth set, the pectorals, covers most of the hyoplastron and, in the most primitive of turtles, the mesoplastron (state 0; Gaffney, 1990: fig. 105). In species in which the first three scale pairs are reduced in size or absent, the pectorals may reach anteriorly to the epiplastron (state 1, fig. 259).

Primitive condition: Absence of pectoral scale-epiplastron contact seems to be the primitive condition. It appears to be absent in basal Pelomedusoides including Araripemys , Euraxemys , the probable Galianemys (AMNH 30050 and AMNH 30551), Cearachelys , and Rosasia .

Homoplasy: This contact occurs in most podocnemidids (but it is absent in at least some Podocnemis erythrocephala , P. sextuberculata , P. unifilis , and Peltocephalus ).

Discussion: The derived condition of this character is present in Kurmademys , Polysternon , Foxemys , and Elochelys (including Elochelys convenarum Laurent et al., 2002 ). This character supports monophyly of the subtribe Foxemydina ( Polysternon ,

Foxemys , Elochelys ) within the tribe Bothremydini .

167. Plastron, pectoral-abdominal scale sulcus crosses mesoplastron: yes ( Cearachelys )

5 0; no, anterior to mesoplastron ( Kurmademys ) 5 1.

Morphology: The plastron in primitive turtles like Proganochelys , Proterochersis , and Kayentachelys includes large medially meeting mesoplastra that are covered in part by the pectoral and abdominal scales. Thus, the pectoral-abdominal scale sulcus crosses the mesoplastron primitively (state 0; Gaffney, 1990: fig. 105) as well as in later turtles (state 0, figs. 258, 264). In more derived turtles in which the mesoplastra are present, this sulcus may be anterior to the mesoplastron (state 1, fig. 259).

Primitive condition: Contact of the pectoral-abdominal sulcus on the mesoplastron is clearly the primitive condition for turtles. The primitive condition is also seen in pleurodires such as Platychelys , Euraxemys , Bothremys , Rosasia , Pelomedusa , and some Podocnemis . This feature appears to show the primitive condition in most of the Bothremydidae .

Homoplasy: The known distribution of this relatively homoplastic character within the Pleurodira suggests that is has appeared at least four times within the Bothremydidae alone. The CI is 0.25.

Discussion: This character supports monophyly of the subtribe Foxemydina within the Bothremydini .

This character is used by Antunes and Broin (1988), and Meylan (1996).

168. Plastron, size and shape of ischial suture: attachment absent ( Proganochelys ) 5 0; large and linear or V-shaped ( Podocnemis ) 5 1; small and round or triangular ( Taphrosphys ) 5 2.

Morphology: In all pleurodires the pelvis sutures to the carapace dorsally and to the xiphiplastron ventrally. The pubic attachment is typically a narrow diagonal suture angled from posterolateral to anteromedial in the middle of the xiphiplastron and does not vary very much. The ischial scar is more variable in shape. In most forms it is a linear or V-shaped contact that is nearly as large as or larger than the pubic suture and is located along the posteromedial portion of the xiphiplastron (state 1, fig. 259, 260; Zangerl, 1948: fig. 11). In a few bothremydid taxa (state 2, fig. 265; Gaffney, 1975a: fig. 11) this suture is reduced to a small circle, much smaller than the pubic scar.

Primitive condition: The absent state is primitive for turtles; state 1 is primitive for pleurodires.

Homoplasy: The ischial suture of Araripemys is also quite small, but it is located more posteriorly on the posterior limits of the xiphiplastron.

Discussion: The derived condition of state 2, a relatively small, circular or triangular ischial suture on the xiphiplastron, is known only in Taphrosphys sulcatus , T. congolensis , and Ummulisani and is an equivocal Taphrosphyini synapomorphy. The presence of an ischial attachment scar is a pleurodire synapomorphy.

169. Plastron, posterior lobe wider than pelvis: no, pelvis visible in ventral view ( Proganochelys ) 5 0; yes, pelvis hidden by wide posterior lobe ( Podocnemis ) 5 1.

Morphology: The posterior lobe of the plastron is relatively narrow and V-shaped in Proganochelys and Palaeochersis , but wider and U-shaped in most Casichelydia (state 1, fig. 264). Although this character is somewhat subjective, the degree of pelvis exposure in ventral view helps make it more objective (compare Gaffney [1990: fig. 88] and Rougier et al. [1995: fig. 2] with Gaffney and Meylan [1991: 335]).

Primitive condition: Based on Proganochelys , the narrow posterior lobe is primitive.

Homoplasy: None in the MPC, but within Eucryptodira a narrow posterior plastron exposing the pelvis has evolved in chelydrids, trionychids, and a number of extinct forms.

Discussion: Although the wide posterior lobe is a synapomorphy for Casichelydia, it is such a simple character, with homoplasy within cryptodires, that its usefulness is limited.

This character is used in Lapparent de Broin and Murelaga (1999) as ‘‘pelvis width … pelvis wider than posterior lobe’’, but it is the plastron that is varying in size, not the pelvis.

170. Plastron, intergular scale: small ( Cearachelys ) 5 0; large intergular, separating gulars and humerals ( Taphrosphys ) 5 1; large intergular, very small gulars, and partial humeral separation ( Dortoka ) 5 2.

Morphology: The intergular scales are the anteriormost scales of the plastron. In Proganochelys , Proterochersis , and Kayentachelys they are relatively short scales that cover the anterior one-third of an anteriorly exposed entoplastron (state 0; Gaffney, 1990: fig. 105). In the genus Taphrosphys among the Pelomedusoides and in some chelids, the intergular coverage of the entoplastron is much expanded and it separates both the gular and the humeral scales (state 1, fig. 265). State two occurs only in Dortoka in our analysis.

Primitive condition: The small intergular (paired or single) widely occurs in outgroups.

Homoplasy: Hydromedusa is close to the Taphrosphys condition. Within the Bothremydidae , a large intergular occurs twice: in the Taphrosphyini and in Elochelys .

Discussion: Within the Bothremydidae , the derived condition is a synapomorphy for Ummulisani + Taphrosphys . This character has been used to argue for the close relation of Elochelys and Taphrosphys ( Lapparent de Broin and Werner, 1998: 163) ; however, our analysis suggests a close relationship among Elochelys , Foxemys , and Polysternon , requiring the large intergular to be acquired twice within the Bothremydidae .

The relative sizes of the anterior plastral scales vary considerably in turtles. However, most of this variation is relatively continuous and does not form readily recognizable groupings. This character is one attempt to use this variation but does not attempt to use all of it.

171. Plastron, axillaryłinguinal scales: present ( Kayentachelys ) 5 0; absent ( Emydura ) 5 1.

Morphology: The small bridge scales can be seen in cryptodire shells (state 0, Zangerl, 1969: fig. 1). State 1 is their absence (fig. 258).

Primitive condition: The absence of information about these scales in the outgroups ( Proganochelys , Australochelys , Palaeochersis ) makes it difficult to determine the primitive chelonian condition; presumably the presence of the scales is primitive.

Homoplasy: None known in pleurodires, although there may be multiple losses within Cryptodira .

Discussion: A problem with this character is the frequent damage to the bridge area of shells, making it difficult to determine the presence or absence of axillary or inguinal scales. In the MPC, Proterochersis is the only pleurodire definitely known to have these scales ( Gaffney, 1990: fig. 105), so their absence is a Megapleurodira synapomorphy.

172. Plastron, abdominal scale narrow: abdominal midline sulcus length equals or exceeds that of anal scale ( Podocnemis ) 5 0; abdominal midline length less than anal scales ( Taphrosphys ) 5 1; abdominal scales do not meet on the midline ( Araripemys ) 5 2.

Morphology: Most of the paired scales of the turtle plastron meet and share a common sulcus on the midline. The length of that midline sulcus for the abdominal scute is primitively one of the longest such midline sulci and is equal to or longer than the length of the anal scales at the midline (state 0, figs. 257–264). In some bothremydids, the abdominal scale is significantly narrowed on the midline (state 1, figs. 265, 267). In Araripemys the abdominal scales do not meet on the midline (state 2; Meylan, 1996: fig. 2).

Primitive condition: Abdominal scales in Proganochelys , Kayentachelys , and most Selmacryptodira are broad at the midline, suggesting that this is the primitive condition.

Homoplasy: The narrowed condition appears to occur several times within the Pleurodira . Among the Cryptodira , lateral displacement of the abdominals and loss occur in the Chelydridae and Kinosternidae ( Hutchison and Bramble, 1981) .

Discussion: Although this character shows homoplasy, it occurs only once within the Podocnemidoidea and, in the MPC, is a synapomorphy of a subset of the Taphrosphyini : Taphrosphys + Ummulisani .

173. Plastron, anterior lobe reaches carapace edge: anterior margin of plastron reaches same level or beyond anterior margin of carapace ( Cearachelys ) 5 0; anterior margin of plastron well posterior to the anterior margin of the carapace ( Foxemys ) 5 1.

Morphology: In most turtles the anterior lobe of the plastron is nearly as long as or longer than the anterior margin of the carapace. This can be seen in the primitive turtles Proganochelys , Proterochersis , and Kayentachelys (state 0; Gaffney, 1990: fig. 105). This is also the case in chelids, podocnemidids, pelomedusids, Euraxemys , Araripemys , and some bothremydids such as the probable Galianemys (AMNH 30550 and AMNH 30551) and Cearachelys . In the known Bothremydini the anterior limit of the plastron is well posterior to the anterior limit of the carapace (state 1, fig. 264).

Primitive condition: Approximate alignment of the anterior ends of the plastron and carapace is considered to be the primitive condition for turtles.

Homoplasy: In addition to those bothremydids with the derived condition described above, this condition occurs in some cryptodires with reduced plastra including Chelydra , Macrochelys , Staurotypus , Claudius , and Sternotherus .

Discussion: This character is a synapomorphy for the tribe Bothremydini . The character is gradational at best, and it can be ambiguous if the articulated carapace and plastron are held at varying angles of view.

174. Plastron, anal notch: absent ( Kayentachelys ) 5 0; present ( Taphrosphys ) 5 1.

Morphology: The posterior lobe of the plastron has a recessed or embayed posterior margin on the midline (state 1, fig. 264).

Primitive condition: The straight or posteriorly convex margin is primitive, as this occurs in Proganochelys ( Gaffney, 1990) and Palaeochersis ( Rougier et al., 1995) .

Homoplasy: None known in the MPC, but within Selmacryptodira many taxa may acquire at least a shallow anal notch, and within Chelidae some (e.g., some Chelodina ) may also have a nearly absent anal notch.

Discussion: The anal or xiphiplastral notch is a pleurodire synapomorphy, possibly related to the sutured pelvis.

The character is used in Antunes and Broin (1988), Lapparent de Broin and Murelaga (1999), Lapparent de Broin (2000b), and Fuente and Iturralde-Vinent (2001).

175. Shell, surface texture: surface rough with radiating ridges diverging from the posteromedial part (growth center) of the vertebral and pleural scale areas, as in Proganochelys 5 0; weakly granulated polygons, ‘‘pelomedusoid’’ pattern, also as in chelids 5 1; strongly granulated polygons, as in Taphrosphys 5 2; fine, striated ridges as in Polysternon 5 3; smooth, as in Cearachelys 5 4; numerous small pits, as in Araripemys 5 5; texture as in Dortoka 5 6.

Morphology: The surface of the shell of many turtles is textured or sculptured; in others the shell is quite smooth. Within the Pleurodira a few species have radiating patterns of raised ridges, including Platychelys and Notoemys (state 0). Others, such as Araripemys , have patterns of numerous small pits (state 5). In certain members of the Bothremydidae , Chelidae , and Podocnemididae , there is a pattern of reticulate and anastomosing furrows and/or long striations that do not radiate from growth centers (state 1). At least in the Podocnemididae , these may be limited to the bridge. Broin (1977) called this texture ‘‘décoration pélomédusidienne’’. She described it as fine vascular grooves more or less dichotomous and discontinuous.

Primitive condition: The shell surface of Proganochelys has a pattern of radiating raised ridges (state 0; Gaffney, 1990: figs. 73, 74). The same pattern appears to have been present in Proterochersis , Platychelys , and Notoemys . This may be the primitive condition for pleurodires. However, within the Pelomedusoides, neither the pelomedusids nor most podocnemidids have well-developed shell surface texture.

Homoplasy: Shell surface sculpturing similar to that observed within some bothremydines occurs also in some chelids (AMNH 103702) and some Podocnemis .

Discussion: The pattern of reticulate furrows and/or long striations has been used to recognize shell material of the Bothremydidae ( Antunes and Broin, 1988: character C-8; Lapparent de Broin 2000b: 67) or Taphrosphys ( Antunes and Broin, 1988; character H- 1; Broin, 1977; Lapparent de Broin and Werner, 1998: 41). Lapparent de Broin (2000b) argued that the carapace in bothremydids is more strongly decorated than in the other Pelomedusoides; that there are deeper short dichotomous sulci similar to marine cryptodires in Bothremys and sulci often united into salient polygons in Taphrosphys .

In our survey of bothremydid shells, we have found some inconsistencies in the distribution of these characters. In the primitive bothremydids, Kurmademys and Cearachelys , there is an extremely fine surface texture of very small raised areas that are only visible under magnification. Also, there are Taphrosphys -like polygons in Araiochelys , a bothremydine on the basis of skull morphology. Polysternon and Foxemys , also bothremydines, have surface texture, but these do not resemble those of other bothremydids. Polysternon is described ( Broin, 1977) as having grooves that are deeper and larger than those generally seen in ‘‘décoration pélomédusidienne’’. They are described as sinuous, longitudinal and continuous, and disposed in scale areas of the carapace anterior-posteriorly ( Broin, 1977).

The senior author thinks that this character should be dropped because the various states are difficult to identify objectively and consistently over a wide range of taxa. Making limited comparisons at the specific level for alpha taxonomic decisions should be done, but wide comparisons of all pleurodires becomes very subjective. Attempts to use surface texture in the highly ornamented Trionychidae have led to similar frustration. However, removing the character produces the same MPC. Nonetheless, state 1, if it can be recognized consistently, is a synapomorphy for Eupleurodira.

Innumerable authors have used surface texture as a character in studies of pleurodires, with some of the more explicit being Antunes and Broin (1988), Lapparent de Broin and Werner (1998), Lapparent de Broin and Murelaga (1999), and Lapparent de Broin (2000b). Lapparent de Broin and Murelaga (1999) use ‘‘decoration’’ states, similar to the ones used here.

SUMMARY

This study shows that pleurodires had a more extensive and more complex evolutionary history than has been realized. The discovery of new taxa of extinct groups has revealed a diversity of morphologies indicating a remarkable diversity of feeding and sensory strategies. The recognition of this new diversity is based on the discovery and description of many new skulls of pleurodires in this and other recent papers. Most previous work has been based on the shell, which in pleurodires in general, and in the Pelomedusoides in particular, is relatively conservative morphologically, masking the magnitude of pleurodire diversity.

The known history of the Pleurodira begins in the Late Triassic of Germany with the very high-domed, tortoiselike shells of Proterochersis . More than two dozen Proterochersis shells have been found in freshwater, stream-deposited sediments. Although they may have been aquatic, the high-domed shell is commonly assumed to indicate purely terrestrial turtles, however, it is also consistent with an adaptive response to large predators, like phytosaurs and metoposaurs. Although known only from the shell, Proterochersis is hypothesized as the sister taxon to all other pleurodires because it lacks the reduced scale arrangement of other pleurodires. Between the Late Triassic and Late Jurassic there is a long hiatus in the pleurodire record. The Late Jurassic yields a sparse pleurodire fauna of near-shore marine taxa of the family Platychelyidae , which is known from central Europe and the Caribbean. This family survives into the Early Cretaceous of South America. The Platychelyidae is the sister group to all remaining pleurodires, that is, the nanorder Eupleurodira consisting of the hyperfamilies Cheloides and Pelomedusoides.

It is in the latter part of the Early Cretaceous that pleurodire diversity sharply increases, supporting the idea that a significant amount of the earlier record is missing. The Cheloides, consisting of the family Chelidae , is the sister group to the Pelomedusoides. Chelids appear in the fossil record in the Albian at about the same time as the Pelomedusoides, but it is likely that this basal divergence took place earlier because of the high diversity of Pelomedusoides in the Early Cretaceous that we have been able to document with this study. Evidence for monophyly of the Chelidae is supported by the presence in all chelids of biconvex fifth and biconcave seventh cervical vertebrae (character 128) and the extreme cheek emargination extending into the temporal region (character 39).

At least four major clades of Pelomedusoides are known by the Albian. These are based on the genera Araripemys ( Araripemydidae ), Euraxemys ( Euraxemydidae ), Brasilemys (Podocnemidinura) , and Cearachelys ( Bothremydidae ) from the Albian Santana Formation of Brazil, which shows that the major groups of Pelomedusoides were already established. A possible fifth lineage is Teneremys , from the Early Cretaceous of northern Africa. Based primarily on shell characters, Teneremys is resolved in this study as the sister taxon to the superfamily Podocnemidoidea , but it is as yet too poorly known for a well-tested relationship hypothesis.

The family Araripemydidae consists of one taxon, Araripemys barretoi from the Aptian– Albian of Brazil. Description of new cranial material suggests a possible sister-group relationship to the extant family Pelomedusidae based on extensive temporal and cheek emargination (character 14), but this relationship is only weakly supported. Araripemys , although highly autapomorphic in many features, may represent the first record of the Pelomedusidae , a lineage still important in the recent African fauna. Regardless of whether they are sister taxa, they are outgroups to all the remaining Pelomedusoides because they lack the advanced feature of a partial or full covering of the prootic (character 94). Araripemys is characterized by very thin, narrow triturating surfaces and a shell that lacks mesoplastra and has the first costals reaching the shell margin. Araripemys was a near-shore, probably marine, possibly freshwater predator, very similar in habitus to recent turtles that prey on fish (trionychids, long-necked chelids).

The new family Euraxemydidae consists of two new genera: Euraxemys essweini , n. gen. et sp., from the Albian Santana Formation of Brazil, and Dirqadim schaefferi , n. gen. et sp., from the Cenomanian Kem Kem beds of Morocco. Members of the Euraxemydidae share the unique possession of a medial process of the quadrate partially covering the prootic (character 94) and narrowly contacting a ventral process of the exoccipital, in contrast to all other pleurodires, which have either complete exposure or complete covering of the prootic ventrally. Furthermore, they possess a ventral process of the exoccipital that is exposed at the lateral margin of the basioccipital in an elongate foot (character 86). The Euraxemydidae is hypothesized as the sister group to the families Podocnemididae and Bothremydidae , which together form the superfamily Podocnemidoidea . Members of the superfamily Podocnemidoidea have the processus interfenestralis of the opisthotic and most of the prootic covered ventrally by the quadrate and basisphenoid (characters 94 and 99). The Euraxemydidae provides a model for the primitive condition for many other characters in the superfamily Podocnemidoidea .

Brasilemys is the oldest member of the diverse group represented by the family Podocnemididae , which in the Recent fauna contains eight species divided into three genera currently living in South America and Madagascar. Brasilemys , from the Albian Santana Formation, the Cenomanian Hamadachelys from Morocco, and the Podocnemididae make up the epifamily Podocnemidinura. This clade is united by the possession of a cavum pterygoidei formed by the basisphenoid and pterygoid (character 69). The family Podocnemididae is known from the Cretaceous to the Recent and is known from all continents except Australia and Antarctica.

Cearachelys is the oldest known of the Bothremydidae , a now extinct family that appears to be the most diverse family among the Pleurodira , reaching its greatest diversity later in the Late Cretaceous and Paleogene. All pleurodires known from the Early Cretaceous Santana Formation, including Cearachelys , were probably near-shore marine species, although some may have been fresh-water. In the Santana fauna, Euraxemys and Brasilemys may have been more generalized predators, with Araripemys more specialized for fish-eating. However, it is Cearachelys that may have begun the bothremydid trend toward a crushing feeding apparatus with a habitus similar to that of the Recent emydid Malaclemys .

The Late Cretaceous reveals the persistence of euraxemydids in North Africa and a slight increase in diversity of podocnemidids in South America. The big event in pleurodire evolution at this time, however, is the explosive radiation of the Bothremydidae . This rapid expansion in morphological diversity can be observed in four clades, recognized as tribes, namely the Kurmademydini , Cearachelyini, Bothremydini , and Taphrosphyini . Although the oldest bothremydid is the Albian Cearachelys from Brazil, the most basal clade, and sister group to all other bothremydids, is the Maastrichtian tribe Kurmademydini from India. The two genera in the Kurmademydini have extensive temporal and cheek emargination. Kurmademys has a broad, triangular triturating surface, while that of Sankuchemys is narrower, with an accessory triturating ridge. The phylogenetic position of the Kurmademydini is based, among other characters, on the persistence of the fossa precolumellaris (character 56), which is absent in all other bothremydids. The Kurmademydini were most likely freshwater forms with lifestyles resembling those of the Recent pelomedusids Pelomedusa and Pelusios .

The tribe Cearachelyini consists of two genera, Cearachelys from the Albian of Brazil and Galianemys from the Cenomanian of Morocco. They have the triangular, posteriorly expanded triturating surfaces typical of bothremydids (character 34). Their monophyly is supported by the presence of a jugal that is retracted from the orbital margin (character 20). While the Kurmademydini are found in freshwater sediments, as is Galianemys , Cearachelys is in the presumably near-shore marine Santana Formation. The broad jaws of the Cearachelyini genera suggest that their lifestyles may have been comparable to broad-jawed emydids like Graptemys and Malaclemys .

The tribe Bothremydini is characterized by the very broad preorbital part of the skull with very wide triturating surfaces, often with conical pits on the skull and jaws (characters 34, 35). The function of the pits is unknown, but they may have allowed the animal to hold and crush slippery, ovoid prey, like gastropods. The Bothremydini are the most longranging bothremydids, extending from the Late Cretaceous Santonian to the Eocene Ypresian. Geographically, this clade is widespread, extending from its peak diversity area of North Africa to North America, Europe, and the Middle East. Although most are near-shore marine, the European subtribe Foxemydina and the North American species Chedghaii hutchisoni are known from freshwater sediments. This radiation of broad-jawed, probable molluscivores is unique among turtles. In the genus Graptemys , the females of four species approach the morphology seen in the Bothremydini , but they are much smaller, nonmarine turtles with a very restricted distribution. Among marine cryptodires many species have enlarged triturating surfaces, but none has produced the inflated face seen in members of this tribe.

The tribe Bothremydini consists of Foxemys mechinorum from the CampanianMaastrichtian of France; Polysternon provinciale from the Campanian of Europe; Zolhafah bella from the Maastrichtian Dakla Formation of Egypt; Rosasia soutoi from the Campanian–Maastrichtian of Portugal; Araiochelys hirayamai , n. gen. et sp. and Bothremys maghrebiana , n. sp. from the Danian phosphates of the Ouled Abdoun Basin, Morocco; Bothremys cooki from the Maastrichtian Navesink Formation of New Jersey; Bothremys kellyi , n. sp. from the Ypresian phosphates of the Ouled Abdoun Basin, Morocco; Bothremys arabicus from the Santonian of Jordan; Chedighaii hutchisoni , n. gen. et sp. from the Campanian Kirtland Formation of New Mexico; and Chedighaii barberi , n. gen. from the Campanian of Arkansas, Alabama, Kansas, and New Jersey.

The tribe Taphrosphyini has a variety of triturating surfaces but lacks the wide, triangular surfaces typical of the other bothremydids. Members of the Taphrosphyini are characterized by the presence of a jugal-quadrate contact (character 22), the absence of a maxilla-quadratojugal contact (character 38), and the absence of a supraoccipital-quadrate contact (character 79). Among the Bothremydidae , the Taphrosphyini is the most diverse morphologically. The feeding surfaces show surprising variation. The long, narrow skull of Labrostochelys is similar to the skull of some fish-eating trionychids, while the very short skull of Phosphatochelys is similar to some cheloniids. Other skulls, such as those of Azabbaremys and Arenila , are large and massive, but without broadly expanded triturating surfaces, while Ummulisani has very narrow and deep labial ridges. The nasal regions of Taphrosphyini also show wide diversity. Rhothonemys has nasal openings and cavities that are more than twice the size of the orbits, in contrast to the nasal openings in Labrostochelys , which are smaller than the relatively small orbits. This diversity in the skull morphology of the Taphrosphyini is mostly evident in the Paleogene of North Africa, but the group is also known from the East Coast of North America, central Africa, and (based on shells) Europe and South America. All the Taphrosphyini are near-shore marine.

The tribe Taphrosphyini consists of Taphrosphys sulcatus from the Danian Hornerstown Formation of New Jersey; Taphrosphys congolensis from the Paleocene of Cabinda, west Africa; Taphrosphys ippolitoi , n. sp. and Labrostochelys galkini , n. gen. et sp. from the Danian phosphates of the Ouled Abdoun Basin, Morocco; Phosphatochelys tedfordi and Ummulisani rutgersensis , n. gen. et sp. from the Ypresian phosphates of the Ouled Abdoun Basin of Morocco; Rhothonemys brinkmani , n. gen. et sp. from the Paleogene phosphates of the Ouled Abdoun Basin of Morocco; Azabbaremys moragjonesi from the Paleocene Teberemt Formation of Mali; Nigeremys gigantea from the Maastrichtian of Niger; and Arenila krebsi from the Maastrichtian Dakla Formation of Egypt.

When the skull morphology of the members of the Taphrosphyini is considered alongside that of the other three tribes, it becomes apparent that the family Bothremydidae has the greatest range of skull forms of any turtle family yet known. In fact, the skull morphologies of many turtle families seem remarkably uniform in comparison (e.g., Testudinidae , Kinosternidae , Pelomedusidae , Trionychidae , Carettochelyidae ). Even in families with relatively diverse skull morphologies (e.g., Geoemydidae , Chelidae ), variation generally occurs in one or two major aspects of their structure (triturating surface and cheek completeness in geoemydids; cheek emargination in chelids). There are also families with bizarre skull morphologies (e.g. Nanhsiungchelyidae , Protostegidae ), but these are not diverse, at least as they are now known. In no other family do we see the extremes exemplified by the skulls of Cearachelys , Bothremys , Labrostochelys , Azzabaremys , Rhothonemys , and Phosphatochleys. It is this remarkable variation in skull morphology that has allowed us to formulate a strong hypothesis of bothremydid relationships in spite of the presence in Pelomedusoides of remarkably uniform shells.

A phylogenetic analysis of the core dataset of 41 taxa, 122 cranial characters, and 52 postcranial characters relies on comparative descriptions of these taxa. The analysis using PAUP results in one most parsimonious cladogram of 382 steps and a consistency index of 0.6. A Bremer decay analysis shows that the family Bothremydidae is strongly supported at five steps, the tribes Cearachelyini and Kurmademydini have an index of 2, and the tribe Taphrosphyini has an index of 3. The tribe Bothremydini becomes unresolved at one step and is the most weakly supported of these groups. The addition of selected, shell-only taxa with low missing data to the core dataset results in one equally parsimonious cladogram, which is resolved as ( Proterochersis ( Platychelyidae ( Dortoka ( Chelidae ( Pelomedusidae + Araripemys ) ( Euraxemydidae ( Teneremys ( Podocnemididae + Hamadachelys + Brasilemys ( Bothremydidae )))))))). A partitioned dataset consisting only of cranial characters (excluding all shell-only taxa) results in one equally parsimonious cladogram identical to the most parsimonious cladogram resulting from the whole dataset; however, a partitioned dataset consisting only of postcranial characters (excluding all skull-only taxa) resulted in 2704 trees, the consensus of which lacks resolution for nearly all of the Pelomedusoides, but which does resolve more basal pleurodires.

Figures

Fig. 264. Bothremys or Chedighaii barberi (Schmidt, 1940). Partially restored shell based mostly on FMNH P26055, the holotype of ‘‘Podocnemis’’ barberi Schmidt, 1940. Upper left, dorsal view; upper right, ventral view; lower, right lateral view. See Note Added in Proof. [J. Kane, del.]