A Review of the Fossil Record of Nonbaenid Turtles of the Clade Paracryptodira

Abstract 
 The fossil record of nonbaenid paracryptodires ranges from the Late Jurassic (Kimmeridgian) to the Paleocene of North America and Europe only. Earlier remains may be present as early as the Middle Jurassic (Bathonian). Only a single dispersal event is documented between the two continents after their breakup during the Cretaceous in the form of the appearance of the Compsemys lineage in the Paleocene of France. Nonbaenid paracryptodires were restricted to freshwater aquatic environments but display adaptations to diverse feeding strategies consistent with generalist, gape-and-suction, and hypercarnivorous feeding. Current phylogenies recognize two species-rich subclades within Paracryptodira, Baenidae and Pleurosternidae, which jointly form the clade Baenoidea. A taxonomic review of nonbaenid paracryptodires concludes that of 34 named taxa, 11 are nomina valida, 15 are nomina invalida, and 8 are nomina dubia.


Introduction
The term Paracryptodira is herein used to refer to the most inclusive clade of turtles that includes the Eocene baenid Baena arenosa Leidy, 1870 and the Late Jurassic pleurosternid Pleurosternon bullockii Owen, 1842, but no extant turtle . The name was originally introduced by Gaffney (1975) to unite all turtles that possess a foramen posterius canalis carotici interni located midway along the contact of the basisphenoid with the pterygoid. Although Evans and Kemp (1976) soon after questioned the utility of this character, as it reasonably may be interpreted as a symplesiomorphy, more recent analyses of turtle relationships that densely sample at the species level typically retrieve the group of turtles with the paracryptodiran condition as monophyletic (e.g., Joyce 2007;Joyce et al. 2016).
As it is somewhat difficult to rigorously differentiate paracryptodires from sympatric helochelydrids (see Joyce et al. [2011] compared with ), future work will need to clarify if all or parts of the latter, at least as currently understood, are nested within the other. The fossil record of Baenidae, one of the primary clades of Paracryptodira, was previously summarized by Joyce and Lyson (2015). We therefore here summarize the fossil record of the remainder of the group.
The idea that European pleurosternids and North American baenids form a group had already been proposed by Dollo (1886), who united these turtles in his classification. This arrangement was soon after supported by Lydekker (1889a), who proposed the redundant names Amphichelydia and Pleurosternidae based on a series of characters that now must be viewed as plesiomorphic (Gaffney 1975). The grouping of Lydekker (1889a) does not fully correspond to the current understanding of Paracryptodira, however, as it included the stem pleurodire Platychelys oberndorferi Wagner, 1853 and the helochelydrid Helochelys danubina Meyer, 1854. Over the course of the succeeding decades, the meanings of the names Amphichelydia and Pleurosternidae were further removed from the currently hypothesized content of Paracryptodira, as increasing numbers of stem turtles were united with the former (e.g., Williston 1925;Nopcsa 1928), while the latter was increasingly circumscribed to the exclusion of baenids (e.g., Baur 1891;Hay 1908b;Gilmore 1919;Williams 1950). Although Gaffney (1975) could have used either name for the clade of turtles he recognized, we here conform to his choice of the new name Paracryptodira, as this has been firmly entrenched in the literature in the last decades.
For institutional abbreviations, see Appendix 1. Named nonbaenid paracryptodiran genera are listed in Appendix 2.

Skeletal Morphology
Cranium At present, the skull is only known for five nonbaenid paracryptodires: Compsemys victa from the Paleocene of Colorado, USA (see ; Dorsetochelys typocardium from the Early Cretaceous of the United Kingdom (see Dorsetochelys delairi of Evans and Kemp 1976); Pleurosternon bullockii from the Early Cretaceous of the United Kingdom (see Mesochelys durlstonensis of Evans and Kemp 1975 and Pleurosternon bullockii of Sterli et al. 2010); Glyptops ornatus from the Late Jurassic of Wyoming, USA (see Glyptops plicatulus of Gaffney 1979); and Uluops uluops from the Late Jurassic of Wyoming, USA (Carpenter and Bakker 1990). A well-preserved skull represents the holotype of Dorsetochelys buzzops (Bakker 1998), but as this specimen was never figured (only idealized sketches are available) and is now lost, it is disregarded herein. Although the available nonbaenid paracryptodire skulls are relatively complete, most are either heavily crushed (e.g., Dorsetochelys typocardium, Glyptops ornatus, and Mesochelys durlstonensis) or poorly documented (e.g., Dorsetochelys typocardium and Uluops uluops). The only remaining skull, that of Compsemys victa, is highly apomorphic and not representative for the group. We therefore do not figure any material herein and keep our summary to a minimum based on the sources cited previously.
The skulls of nonbaenid paracryptodires generally show poorly developed lower and upper temporal fenestrae (fully absent in Compsemys victa), but they range from flat and elongate with dorsally oriented orbits (e.g., Glyptops ornatus), to short and high with dorsolaterally oriented orbits (e.g., Uluops uluops), to triangular and massive with laterally oriented orbits (e.g., Compsemys victa). All forms possess relatively large nasals with a midline contact, relatively small prefrontals that lack a midline contact, frontals that contribute to the orbits (absent in Compsemys victa), parietals that contact the squamosals (unclear in Compsemys victa), jugals that contribute to the orbits (absent in Glyptops ornatus), and open incisurae columella auris (superficially enclosed in Compsemys victa). The pterygoids have been reported as lacking a midline contact in Glyptops ornatus and Pleurosternon bullockii, but this is probably a preservational artifact, especially as a midline contact is clearly present in all taxa where this region is sufficiently preserved. The foramen posterius canalis carotici interni enters the skull midway along the contact of the pterygoid with the basisphenoid, but it is unclear if the palatal branch of the internal carotid is present, considering that it has recently been shown to be absent in some baenids (Lipka et al. 2006;Rollot et al. 2018) and that the internal anatomy of no skull has of yet been studied. The basicranial region resembles that of most turtles by having an elongate basisphenoid and broad pterygoids, with the exception of Compsemys victa, which displays an extremely shortened basisphenoid and narrow pterygoids.
Much variation is apparent in the palate of nonbaenid paracryptodires, which highlights likely differences in feeding ecology. In Glyptops ornatus and Pleurosternon bullockii, the labial ridges are low, and the triturating surfaces narrow. The labial ridges are higher and the triturating surfaces are expanded in Dorsetochelys typocardium and Uluops uluops, particularly through contributions from the palatines. In contrast, the labial ridges of Compsemys victa are marked by a toothlike median projection, and the broad secondary palate elaborated by low accessory ridges is formed by the maxillae, palatines, and vomer. These morphologies are broadly consistent with gapeand-suction feeding, generalist feeding, and macrocarnivory, respectively (see "Paleoecology").
The plastron of most nonbaenid paracryptodires consists of an entoplastron and pairs of epi-, hyo-, meso-, hypo-, and xiphiplastra covered by paired gulars, extragulars, humeral, pectorals, abdominals, femorals, and anals and four pairs of inframarginals. All bones and scutes have a straight midline contact with the exception of the extragulars and inframarginals ( Figure 1). Compsemys victa ( Figure 1C) and Selenemys lusitanica deviate from this pattern by having highly sinuous midline scute contacts; Pleurosternon bullockii, by possessing a single, median gular; and Riodevemys inumbragigas, by lacking a midline contact of the mesoplastra. The development of the plastral buttresses is only poorly documented.

Postcranium
Only small numbers of nonshell postcranials have been described for Glyptops ornatus (Glyptops plicatulus of Hay 1908b), Riodevemys inumbragigas (Pérez-García, , and Toremys cassiopeia , but these remains are too fragmentary to allow drawing rigorous conclusions about the postcranial anatomy of these turtles.

Phylogenetic Relationships
The phylogenetic relationships of nonbaenid paracryptodires have only been explored in the more recent past. The global analysis of Joyce (2007) includes 11 paracryptodires, of which 6 are baenids and 5 are nonbaenids. The more narrowly focused analysis of paracryptodire relationships of , on the other hand, densely samples baenids but still only includes six nonbaenid species. The sample of nonbaenid species of the latter analysis was increased stepwise by Pérez-García and Ortega (2011), Pérez-García (2012), Pérez-García, Espílez, et al. (2015), and Pérez-García,  through the addition of European forms. The trees presented herein (Figures 2 and 3) adhere to the topology of Pérez-García, . All trees agree with each other by not supporting the historical dichotomy of paracryptodires into baenids and pleurosternids, but results are still too preliminary to allow drawing affirmative conclusions.

Paleoecology
Most nonbaenid paracryptodires have been collected from riverine and lacustrine sediments, and the group can therefore be reconstructed to have been freshwater aquatic. This conclusion is broadly consistent with the overall hydrodynamic shape of their shells, splayed femoral trochanters, at least as documented for Glyptops ornatus (Glyptops plicatulus of Hay 1908b) and Toremys cassiopeia , and the aquatic feeding specializations apparent in the jaws (see following paragraph).
The significant amount of morphological variation apparent in the small set of available skulls (see "Skeletal Morphology") attests to much ecological plasticity within the group. The elongate skull with low labial margins of Glyptops ornatus and Pleurosternon bullockii is broadly consistent with gape-and-suction feeding (Pritchard 1984;Foth et al. 2017). The higher labial ridges and expanded triturating surfaces of Dorsetochelys typocardium and Uluops uluops, on the other hand, are consistent with dietary generalists (Foth et al. 2017). The massive skull of Compsemys victa, by contrast, is unique among turtles by being extremely thick boned, but its overall shape, in particular the toothlike median projection and reduced pterygoids, is highly reminiscent of the extant bigheaded turtle Platysternon megacephalum (Hutchison and Holroyd 2003) and suggests the lifestyle of a macrocarnivorous snapping turtle Foth et al. 2017).

Paleobiogeography
All know paracryptodires have been collected from continental sediments across North America and western Europe, supporting the notion that these landmasses once formed a faunal province to the exclusion with the rest of the world (Hay 1908b;Hirayama et al. 2000;Pérez-García, Royo-Torres, et al. 2015;Joyce et al. 2016;Figures 4 and 5). The oldest remains were historically known from the Late Jurassic of North America and Europe (e.g., Lydekker 1889a; Marsh 1890), but histological analysis of fragments from the United Kingdom suggests that the group may reach back to the Middle Jurassic (Scheyer and Anquetin, 2008).
A significant gap in the fossil record on both continents obscures the history of the group during the post-Berriasian Early Cretaceous after the separation of Europe and North America through the formation of the North Atlantic. Only fragments document the persistent presence of the group in the Early Cretaceous (Hauterivian-Barremian) of France and Spain, and later by remains of the pleurosternid Toremys cassiopeia from the late Early Cretaceous (Albian) of Spain (Pérez-García, Espílez, et al. 2015). In North America, the record continues in the late Early Cretaceous (Aptian/Albian [slashes used herein connote "or"]) in the form of the diverse clade Baenidae, which persists until the Eocene (see Joyce and Lyson [2015] for summary).
The Compsemys lineage is documented in the form of Compsemys victa from the Santonian to late Paleocene of Laramidia (see  for recent summary) and Compsemys russelli from the late Paleocene of France (Pérez-García 2012). As the lineage is more extensive in North America and otherwise absent from Europe, it seems that the Compsemys russelli lineage dispersed from North America to Europe during the early Paleocene via Greenland (Pérez-García 2012; Joyce et al. 2016). Current phylogenies place the Compsemys lineage outside of the clade formed by Baenidae and Pleurosternidae Pérez-García, Royo-Torres, et al. 2015). The group is therefore predicted to possess a ghost lineage that reaches back to the Late Jurassic.

Valid Taxa
See Appendix 4 for the hierarchical taxonomy of nonbaenid paracryptodires as described in this work.
Paracryptodira Gaffney, 1975 Phylogenetic definition. In accordance with , the name Paracryptodira is herein referred to the most inclusive clade containing Pleurosternon bullockii Owen, 1842 and Baena arenosa Leidy, 1870, but no species of extant turtle.
Diagnosis. Paracryptodires can be differentiated from all other turtles by symplesiomorphically showing a midline contact of the nasals and modest lower and upper temporal emarginations, lacking a midline contact of the prefrontals, and possessing large mesoplastra with a midline contact, and by apomorphically showing a dense surface texture that covers the skull and shell, a foramen posterius canalis carotici interni located halfway along contact between pterygoid and basisphenoid, and thickenings of the plastron medial to the bridges. Compsemys Leidy, 1856 Type species. Compsemys victa Leidy, 1856. Diagnosis. Compsemys can be diagnosed as a paracryptodire by the full list of characters provided for that clade above. Compsemys can be differentiated from all other paracryptodires by possessing greatly thickened skull bones; reduced temporal emarginations; expanded, rectangular quadratojugals; a postorbital contribution to the small, fully enclosed cavum tympani; a median, hooklike tomial process; an expanded secondary palate with a broad contribution from the vomer and palatines; a midline contact of peripheral I anterior to the nuchal combined with a loss of contact between costal I and peripheral I (also present in Selenemys lusitanica); large mesoplastra with an offset midline contact; a deeply sinuous midline sulcus; and a deep anal notch. Type material. USNM 960 (holotype), shell fragments, including a neural and a costal (Leidy 1860, pl. 11.5-7;Hay 1908b Comments. Compsemys victa is based on an assortment of shell fragments that were collected near Long Lake (Leidy 1856) in what is now the state of North Dakota (Gaffney 1972). The Late Cretaceous (Maastrichtian) Hell Creek Formation is exposed in this area, and it is therefore highly likely that this fossil was collected from that stratum , not from the Jurassic, as stated by Cope (1870). Hay (1910) established two species from New Mexico based on fragmentary material: Compsemys parva and Compsemys vafer. Gilmore (1919) referred more complete shell material to the two species of Hay (1910) and furthermore established two additional species based on more complete material, Compsemys puercensis and Compsemys torrejonensis, which he diagnosed relative to one another by differences in the size of the eighth costal and in the shape of the anal notch. It is important to note that the shells reported by Gilmore (1919) are composed of eroded fragments that were assembled with plaster into three-dimensional models. Various purported differences therefore seem to be a result of imperfect reconstructions, such as the unnaturally broad and square anal notch of the shell Gilmore (1919) referred to Compsemys vafer. The apparent difference in the shape of the anal notch is otherwise consistent with sexual dimorphism as seen in various extant turtles as well. We therefore agree with the original assessment of Gaffney (1972) that all named taxa from New Mexico should be synonymized with Compsemys victa but disagree with  that Emys obscurus and Glyptops depressus should be added to the list of synonyms as well (see Emys obscurus and Glyptops depressus below). We furthermore agree with Gaffney (1972) and  that it is prudent for the moment to refer all fragmentary material from the Campanian to Paleocene to Compsemys victa.  recently provided the description of a skull of Compsemys victa based on material first reported by Hutchison and Holroyd (2003). The skull of this turtle is unique by being triangular, lacking emarginations, and showing a strong median premaxillary hook, small, medially oriented eyes and ears, a small brain case, and unusually thick bones. An overall shape resemblance of the skull and shell with that of the extant bigheaded turtle Platysternon megacephalum suggests that Compsemys victa was a hypercarnivorous snapping turtle that likely could not retract its head within the shell (Hutchison and Holroyd 2003;. The preponderance of Compsemys victa material in muddy overbank deposits suggests that this aquatic turtle preferred ponded habitats (Hutchison and Holroyd 2003;.

Compsemys victa
Given the highly characteristic shell texture of Compsemys victa, we readily accept all reports of this taxon that are either associated with figures or specimen numbers. The earliest record is a fragment from the Santonian of Alberta (Brinkman 2003). Russell (1935) also reported a fragment from the same time and region, but this clearly represents the shell of the solemydid Naomichelys. Compsemys victa has been broadly reported from Late Cretaceous (Campanian) to Paleocene localities across the central portions of Laramidia (see "Referred material and range" above). We cannot confirm reports made by Winchester et al. (1916) and Robison (1986) for the Late Cretaceous (Maastrichtian) of South Dakota and early Paleocene of Utah, respectively, although we find it highly plausible that these regions yield this taxon as well. Compsemys victa occurs consistently in all major Late Cretaceous (Campanian) to Paleocene localities in the southern Rocky Mountains but is notably absent from the northern record during the Campanian and seems to be a rare occurrence in general in the southern provinces of Canada. The northern distribution of this species may therefore have been controlled by climate (Brinkman 2003).
The phylogenetic relationships of Compsemys victa were historically unclear, because informative shell and skull material was lacking. Hay (1908b) originally thought Compsemys victa to be referable to Dermatemydidae, as he presumed it to lack mesoplastra, but the discovery of more complete remains with well-developed mesoplastra allowed him to soon after change his mind in favor of Baenidae (Hay 1910). Gilmore (1919) noted similarities in the shell sculpturing of Compsemys victa with Glyptops spp. and Neurankylus spp. and therefore suggested affinities with Pleurosternidae. Gaffney (1972) once again favored affiliations with Baenidae, whereas Hutchison and Holroyd (2003) favored those with Pleurosternidae. The phylogenetic analysis of  concluded that Compsemys victa is a paracryptodire closely affiliated with neither pleurosternids nor baenids. However, given that relationships among nonbaenid paracryptodires are far from resolved, this conclusion must be viewed with caution. Diagnosis. Compsemys russelli can be diagnosed as a paracryptodire and a representative of Compsemys by the full list of shell characters provided for those clades above. Compsemys russelli can most easily be differentiated from Compsemys victa by possessing two suprapygals that lack lateral contact with peripheral X.
Comments. Although Compsemys russelli is typified by a single nuchal from the late Paleocene (Thanetian) of Department of Marne, France, the type locality yielded a relatively rich fragment fauna that documents most parts of the shell (Pérez-García 2012). Additional fragments are available from similarly dated sediments from the Department of Puy-de-Dôme farther to the south. The possible presence of a derived paracryptodire reminiscent of Compsemys victa had already been hinted at by Broin (1977) and Lapparent de Broin (2001), but the formal description of Compsemys russelli only occurred recently (Pérez-García 2012). Although the French material is characterized by a texture that is different from that in Compsemys victa by being more variable and subdued, many morphological details, in particular the inset nuchal, well-developed mesoplastra, highly sinuous plastral sulcus, and deep nuchal notch, confirm close relationship. Indeed, the two forms are nearly identical in all important aspects. Given that it seems all but certain that the European form derives from its North American relatives (Pérez-García 2012), we find it prudent to unite the French and North American species into the genus Compsemys to avoid rendering the North American taxon paraphyletic. In addition to the trionychid Axestemys vittata, the presence of Compsemys russelli in the Paleocene of Europe is evidence for faunal exchange between North America and Europe during the early Paleogene (Georgalis and Joyce 2017).
Diagnosis. Dinochelys whitei can be diagnosed as a paracryptodire by the full list of shell characters provided for that clade above with exception of a lack of surficial shell sculpturing. Dinochelys whitei can be differentiated from all other paracryptodires by symplesiomorphically lacking a midline contact of peripheral I (in contrast to Selenemys lusitanica and Compsemys spp.) and a nuchal notch (in contrast to Dorsetochelys typocardium), by possessing a cervical (in contrast to Pleurosternon bullockii, Selenemys lusitanica, and Toremys cassiopeia) and a midline contact of the mesoplastra (in contrast to Riodevemys inumbragigas and most baenids), and by apomorphically possessing wide vertebrals and lacking surficial shell sculpturing.
Comments. Dinochelys whitei is based on a nearly complete shell from the Late Jurassic (Tithonian) Dinosaur Park Visitor Center located near Jepsen, Utah (Gaffney 1979). The shell differs substantially from that of Glyptops ornatus (Glyptops plicatulus of Gaffney 1979), and we therefore agree that it represents a valid taxon. Dinochelys whitei is notably smooth shelled relative to Glyptops ornatus, and we therefore reluctantly maintain its placement as a separate genus. All three skull morphotypes currently documented from the Morrison Formation (e.g., "Dorsetochelys buzzops, " Glyptops ornatus, and Uluops uluops) are associated with textured shells, which further supports the validity of this taxon. We herein disregard all juvenile specimens referred to this taxon (Gaffney 1979;Brinkman et al. 2000), as we do not believe that the broad vertebral scutes they show should be used to support taxonomic decisions. Diagnosis. Dorsetochelys typocardium can be diagnosed as a paracryptodire by the full list of characters provided for that clade above. Dorsetochelys typocardium can be differentiated from all other paracryptodires by symplesiomorphically lacking a midline contact of peripheral I (in contrast to Selenemys lusitanica and Compsemys spp.), by possessing a cervical (in contrast to Pleurosternon bullockii, Selenemys lusitanica, and Toremys cassiopeia) and a midline contact of the mesoplastra (in contrast to Riodevemys inumbragigas and most baenids), and by apomorphically possessing a deep nuchal notch and slightly expanded triturating surfaces with a elongate contribution from the palatines.
Comments. The taxonomic history of Dorsetochelys typocardium is somewhat tortuous. The holotype, a carapace from the Purbeck Limestone Group, was initially described by Seeley (1869). As Seeley (1869) did not provide any figures, Lydekker and Boulenger (1887) and Lydekker (1889a) were unable to assess its taxonomic standing. Watson (1910a) noted that Seeley' s (1869) Pleurosternon typocardium seemed to be the same as Glyptops ruetimeyeri, a species described by Lydekker (1889a), but he nevertheless continued usage of the name Glyptops ruetimeyeri, even though Pleurosternon typocardium has priority. Nearly a century later, Milner (2004) described the specimen that Seeley (1869) had originally named Pleurosternon typocardium, confirmed the synonym of Watson (1910a), and concluded that the name "Glyptops" typocardium should be given priority. Milner (2004) furthermore tentatively referred the species to "Glyptops, " as he saw similarities with this North American turtle. Karl, Gröning, et al. (2012) instead highlighted differences and proposed the new genus name Ballerstedtia. Karl, Gröning, et al. (2012) furthermore recognized a new species, Ballerstedtia bueckebergensis, based on several specimens from the late Berriasian of northwestern Germany ). Pérez-García (2014) soon after highlighted that the characters used by Karl, Gröning, et al. (2012) to distinguish their new species of Ballerstedtia are present in both species and that Ballerstedtia bueckebergensis therefore does not represent a distinct species. Evans and Kemp (1976) described an isolated cranium from the Purbeck Limestone Group as the new taxon Dorsetochelys delairi. Milner (2004) suggested that this cranium probably belonged to the shell-based species Pleurosternon typocardium, but this proposal lacked morphological support. Pérez-García (2014) reached the same conclusion by noting that the ornamentation on the dorsal surface of the skull roof of the holotype of Dorsetochelys delairi matches that of the shell of Pleurosternon typocardium. This conclusion seems reasonable because other nonbaenid paracryptodires are also characterized by the presence of a similar ornamentation on the external surface of the bones of the shell and the skull roof (e.g., Pleurosternon bullockii and Glyptops ornatus). Pérez-García (2014) therefore proposed the new combination Dorsetochelys typocardium. We concur with this conclusion herein as well.
The shell anatomy of Dorsetochelys typocardium long remained poorly understood because of an unfortunate combination of factors, including the absence of illustration of Seeley's (1869) Pleurosternon typocardium and the apparent loss of the holotype of Thalassemys ruetimeyeri. Milner (2004) was able to clarify the distinctive carapace anatomy of Dorsetochelys typocardium. He furthermore speculated that other specimens of this species likely existed in historical collections, but that they were probably misidentified as representatives of Pleurosternon bullockii. This was recently confirmed by the referral of several historical specimens housed in different British museums (Pérez-García 2014). The shell of Dorsetochelys typocardium is notably characterized by a deep nuchal emargination, an inverted pentagonal shape, relatively narrow vertebral scutes, a moderately developed anal notch, and a shell bone surface ornamentation consisting of a low vermiculate pattern forming a poorly defined or coarse pitting with striations perpendicular to the plate margins. The skull of Dorsetochelys typocardium is relatively long and broad posteriorly. A second, undescribed cranium (DORCM G.10715) from the Purbeck Limestone Group was tentatively referred to Dorsetochelys delairi by Milner (2004), but our preliminary study of this specimen indicates that it is not a paracryptodire.
The phylogenetic analysis of Joyce (2007) concluded based on the skull of Dorsetochelys typocardium only (his Dorsetochelys delairi) that this taxon is a paracryptodire placed just outside the clade formed by Baenidae and Pleurosternidae. The more recent analysis of Pérez-García, , however, concluded based on the combined skull and shell anatomy of this taxon that Dorsetochelys typocardium forms a polytomy with Glyptops ornatus (their Glyptops plicatulus) at the base of Pleurosternidae. This conclusion further supports the notion that all known Mesozoic European paracryptodires are pleurosternids.
Diagnosis. Glyptops ornatus can be diagnosed as a paracryptodire by the full list of characters provided for that clade above. Glyptops ornatus can be differentiated from all other paracryptodires by symplesiomorphically lacking a midline contact of peripheral I (in contrast to Selenemys lusitanica and Compsemys spp.); by possessing a cervical (in contrast to Pleurosternon bullockii, Selenemys lusitanica, and Toremys cassiopeia), paired gulars (in contrast to Pleurosternon bullockii), and a midline contact of the mesoplastra (in contrast to Riodevemys inumbragigas and most baenids); and by apomorphically possessing an elongate skull with low labial ridges and narrow triturating surfaces.
Comments. Turtles are reported to be regular occurrences in the Morrison Formation (Foster and McMullen 2017), but surprisingly little material of better quality has been described, and the taxonomy of the group is therefore wanting. Glyptops (orig. Compsemys) plicatulus is the first turtle named from the Morrison Formation. It is based on a small assortment of shell fragments collected from outcrops near Garden Park, Colorado (Cope 1877). The holotype does not bear any characters of particular significance beyond its unusual surface texture consisting of densely packed crenulations and tubercles. Marsh (1890) soon after described Glyptops ornatus based on a partial skull from Como Bluff, Wyoming. Marsh (1890) also reported a partial carapace from the type locality with a similar surface to that of Compsemys plicatulus, but he was hesitant to refer the shell to the skull, as they were not found in direct association. Baur (1891) felt less hesitant and referred all turtle material from Como Bluff to Glyptops plicatulus as he believed it to be plausible that the locality only yielded a single species. Hay (1908aHay ( , 1908b conformed to the synonymy of Baur (1891) and referred additional material to Glyptops plicatulus, including a carapace from an undisclosed quarry at Como (Hay 1908a) and a nearly complete skeleton from the Bone Cabin Quarry at Como (Hay 1908b), which is located in the Late Jurassic (Kimmeridgian) Salt Wash Member of the Morrison Formation (Turner and Peterson 1999). Gilmore (1916b), on the other hand, broke with the taxonomic tradition set by Baur (1891) by naming a new species, Compsemys utahensis, this time from the Carnegie Quarry in Utah, which is located in the Kimmeridgian Brushy Basin Member of the Morrison Formation (Turner and Peterson 1999). As the Morrison Formation seemed to have yielded only a single skull morphotype associated with shell material reminiscent of Glyptops plicatulus, Gaffney (1979) much later united all textured shells and their associated skulls into Glyptops plicatulus, including the types of Glyptops ornatus and Glyptops utahensis, as opposed to Dinochelys whitei (see above), which is characterized by the absence of surface sculpturing.
A series of more recent finds has revealed that two additional skull morphotypes are associated with crenulated shell material: Uluops uluops (Carpenter and Bakker 1990; see below) and the enigmatic "Dorsetochelys buzzops" (Bakker 1998; see below). Unfortunately, even though these new skull morphotypes are associated with postcranial material, their shells remain undescribed. Although additional study will need to clarify if both new taxa are valid, we agree that they are systematically different from the skull morphotype preserved in the type of Glyptops ornatus and the referred specimen from the Bone Cabin Quarry (Hay 1908b).
We here consider the type of Glyptops plicatulus to be a nomen dubium, because the texture of its shell is not unique among basal paracryptodires. As a result, Glyptops ornatus is reinstated as valid. Although the type skull of Glyptops ornatus and the skull from the Bone Cabin Quarry are of poor quality, we agree that they are referable to the same taxon, as both are notably narrow and have narrow triturating surfaces and low maxillae. The Bone Cabin Quarry specimen therefore documents the shell morphology of Glyptops ornatus. We agree with Gaffney (1979) that the shell morphology of the type of Glyptops utahensis fully corresponds to that of the Bone Cabin Quarry specimen, and we here therefore synonymize Glyptops utahensis with Glyptops ornatus. As the morphology of all other specimens from the Late Jurassic of North America remains undescribed, we finally refer all other crenulated shell material to Glyptops sp. Our taxonomic concept of Glyptops ornatus greatly resembles that of the taxonomic concept of Gaffney (1979) for Glyptops plicatulus by including the types of Glyptops ornatus, Glyptops utahensis, and the Bone Cabin Quarry speci-men, but it differs by excluding all other referred specimens with crenulated shell material. All herein referred specimens of Glyptops ornatus were extensively described by Gaffney (1979) under the name Glyptops plicatulus.
Glyptops ornatus has been included under the name Glyptops plicatulus in all phylogenetic analyses that are based on the character-taxon matrix of Joyce (2007). Most analysis concludes that this turtle is a basal pleurosternid (e.g., Joyce 2007; Pérez-García, Royo-Torres, et al. 2015). The narrow skull and jaws suggest that this aquatic turtle was a generalist.
The name Glyptops has been used in the last 100 years as a wastebasket taxon for all Late Jurassic to Early Cretaceous turtles with a fine sculpted ("pleurosternid") shell surface texture. As this texture persists among basal baenids, such as Arundelemys dardeni, Trinitichelys hiatti, and Neurankylus spp. (Joyce and Lyson 2015), it is clear that surface texture is not sufficient to diagnose a turtle as belonging to Glyptops. We here refer all isolated shell material with textured surface from the Late Jurassic Morrison Formation to Paracryptodira indet., in particular from Colorado (Cope 1877;Brinkman et al. 2000), New Mexico (Lucas et al. 2006), Utah (Gaffney 1979), and Wyoming (Marsh 1890;Hay 1908aHay , 1908bGaffney 1972). Additional remains have been reported from the Morrison Formation of Montana, South Dakota, and Oklahoma (e.g., Foster and McMullen 2017), but we cannot reproduce these reasonable claims, as specimen numbers or figures are lacking. We herein do not summarize the fossil record of textured shell fragments from the Early Cretaceous of North America, as these likely represent baenids (see Joyce and Lyson [2015] for a summary of more complete material). Type material. NHMUK R911 (holotype), a large, complete plastron (Owen and Bell 1849, pl. 21;Milner 2004, fig. 3).
Type locality. Purbeck, probably near Swanage, Dorset, United Kingdom (Lyddeker and Boulenger 1887; Figure 4); Purbeck Limestone Group, Berriasian, Early Cretaceous (Allen and Wimbledon 1991;Feist et al. 1995;Milner 2004 The syntype series of the latter two species are now synonymized with Hylaeochelys belli (in part for Pleurosternon emarginatum; see Lydekker and Boulenger 1887), a pancryptodiran turtle (see Pérez-García 2012) that will be treated elsewhere. Pleurosternon concinnum and Pleurosternon ovatum were initially distinguished based on minor differences (outline of the carapace and width of the first vertebral scute) but are unanimously regarded as junior synonyms of Pleurosternon bullockii since Lydekker and Boulenger (1887) and Lydekker (1889a, 1889b). Seeley (1869) proposed four new names of Pleurosternon based on material from the Purbeck Limestone Group in Swanage: Pleurosternon oweni, Pleurosternon sedgwicki, Pleurosternon typocardium, and Pleurosternon vansittarti. Lydekker (1889a) noted that he was unable to assess these species and their possible affinities with Pleurosternon bullockii without seeing the material. Watson (1910a) apparently saw Seeley's (1869) material because he concluded in a note that Pleurosternon typocardium was identical to Thalassemys ruetimeyeri (see above), and that Pleurosternon oweni, Pleurosternon sedgwicki, and Pleurosternon vansittarti seemed to be typical examples of Pleurosternon bullockii. Delair (1958), on the other hand, synonymized all of Seeley's (1869) material with Pleurosternon bullockii. More recently, Milner (2004) confirmed Watson's (1910a) observations by setting Pleurosternon typocardium apart from the rest (see Dorsetochelys typocardium above). Although the holotypes of Pleurosternon oweni, Pleurosternon sedgwicki, and Pleurosternon vansittarti still remain to be described, we here conform to the synonymy of Milner (2004) as the morphology of Pleurosternon bullockii is strikingly different from that of Pleurosternon (now Dorsetochelys) typocardium.
Megasternon koenigii has repeatedly been cited by previous authors as a synonym of Pleurosternon bullockii, but variably attributed to different publications by Gray (see Lydekker 1889a; Woodward and Sherborn 1890; Kuhn 1964). The genus name Megasternon first appeared in isolation in Gray (1842). Without species name and proper description, Megasternon must be considered unavailable at that time. The species name Megasternon koenigii first appeared in Gray (1844), again without description or indication, so this name must be considered a nomen nudum. Megasternon koenigii finally became available when Gray (1870) mentioned it again with a clear indication to both Gray (1844) and Owen and Bell's (1849) illustration of the holotype of Platemys bullockii. Megasternon koenigii Gray, 1870 is therefore a junior objective synonym of Pleurosternon bullockii (Owen, 1842).
Pleurosternon bullockii is a large species with a carapace length reaching up to 560 mm in the largest individuals (Milner 2004). The carapace is relatively flat, oval in shape, and almost without nuchal emargination. A small posterior notch is present on the pygal bone. The plastron is characterized by a broad, rounded anterior plastral lobe; broad mesoplastra meeting in the midline; and a deep, V-shaped xiphiplastral notch. The shell bone surface ornamentation consists of small, regular, and clearly defined pits with fine striations perpendicular to the plate margins.
A well-preserved skull from the Purbeck Limestone Group in Durlston Bay was described by Evans and Kemp (1975) as a distinct species, Mesochelys durlstonensis. A fragmentary postcranium was found associated with this skull. Since Gaffney and Meylan (1988), Mesochelys durlstonensis is treated as a junior synonym of Pleurosternum bullockii, and this unique specimen is used in all subsequent phylogenetic studies to score the cranial anatomy of the latter species. However, Milner (2004) was the first to properly assess this synonymy by comparing the shell remains associated with the skull with the turtle material known from the Purbeck Limestone Group. Based on several morphological arguments (similar surface ornamentation, posterior marginal scutes covering the distal part of costals, and presence of a deep, V-shaped xiphiplastral notch), he concluded that Mesochelys durlstonensis was indeed a junior synonym of Pleurosternon bullockii (Milner 2004).
Pleurosternon bullockii was reported from the Tithonian of northern France (Sauvage 1894(Sauvage , 1900(Sauvage , 1912(Sauvage , 1921, but this material must be regarded as indeterminate (Lapparent de Lapparent de Broin 2001). The species was also reported from the Early Cretaceous (Berriasian) of Germany ), but we identify this material as Pleurosternidae indet. as it consists of steinkerns only. Referred material and range. No specimens have been formally referred to date.
Diagnosis. "Pleurosternon" portlandicum can be diagnosed as a paracryptodire by the presence of expanded mesoplastra with a midline contact, a textured shell surface, and overall similarity with Pleurosternon bullockii. "Pleurosternon" portlandicum can be differentiated from Pleurosternon bullockii in its finer surface texture and inframarginals that do not broadly overlap the peripherals.
Comments. "Pleurosternon" portlandicum is based on an imperfect plastron from the Tithonian of the Isle of Portland, Dorset, United Kingdom (Lydekker 1889a). It is therefore only slightly older than the earliest Cretaceous (Berriasian) Pleurosternon bullockii from the same region. The surface ornamentation on the plastral bones consists of small, regular, and clearly defined pits with striations perpendicular to plate margins. This is very similar to Pleurosternon bullockii, but pits are sparser. The entoplastron is about as wide as long in "Pleurosternon" portlandicum, but usually wider than long in Pleurosternon bullockii. The intergular scute is described as pyriform and narrow posteriorly (Lydekker 1889a), but personal observation of the holotype reveals this area to be abnormal in the development of three asymmetric gulars. Finally, the inframarginals do not extend laterally onto the ventral part of the peripherals. Although more specimens are needed to improve our understanding of this taxon, we here provisionally accept the validity of "Pleurosternon" portlandicum based on these differences combined with its stratigraphic age. However, as all characters that would unambiguously link this taxon with Pleurosternon bullockii are not preserved in the holotype, in particular the single median gular, deep anal notch, and lacking cervical scute, we here highlight phylogenetic uncertainty by placing Pleurosternon in quotes.  fig. 4).
Referred material and range. No specimens have been formally referred to date.
Diagnosis. Toremys cassiopeia can be diagnosed as a paracryptodire by the full list of shell characters provided for that clade above. Toremys cassiopeia can be differentiated from all other paracryptodires by symplesiomorphically lacking a midline contact of peripheral I (in contrast to Selenemys lusitanica and Compsemys spp.), by possessing paired gulars (in contrast to Pleurosternon bullockii), and by apomorphically lacking a cervical (as in Pleurosternon bullockii and Selenemys lusitanica).
Comments. Toremys cassiopeia is based on three partial shells from the lower Albian of Teruel Province, Spain (Pérez-García, Espílez, et al. 2015). As in most basal paracryptodires, the external surface of the shell bones of this turtle is ornamented with a pattern of low tubercles often joining to form vermiculate ridges. Well-developed striations are also present perpendicular to the sutural margins of the shell plates. Toremys cassiopeia notably differs from other basal paracryptodires by its young stratigraphic age, small size (carapace length about 15 cm), nuchal bone with reduced anterior margin, and anteriorly particularly wide vertebral .
Referred material and range. No specimens have been referred to date.
Diagnosis. Uluops uluops can be diagnosed as a paracryptodire by the full list of cranial characters provided for that clade above. Uluops uluops can readily be differentiated from Dorsetochelys typocardium, Glyptops ornatus, and Pleurosternon bullockii by having a notably short and high skull; from Compsemys victa by lacking an expanded secondary palate and a median tomial hook; and from baenids by lacking accessory triturating ridges and expanded prefrontal lappets.
Comments. Uluops uluops is based on a well-preserved partial skull from the Late Jurassic Morrison Formation of Colorado (Carpenter and Bakker 1990). The type skull has not yet been figured, but a reconstructive illustration highlights many differences with the skull of the coeval Glyptops ornatus, in particular by being significantly higher and broader. We can confirm these differences based on personal observations of the holotype. The validity of this taxon is therefore unproblematic. Preliminary analyses place Uluops uluops in a polytomy with pleurosternids and baenids Pérez-García, Royo-Torres, et al. 2015), but this is likely to change in the future through the additional study of the skull and the discovery of shell material. We specifically note that many fragments herein referred to "Glyptops" indet. based on their sculpturing may well represent Uluops uluops.
The name Uluops uluops is somewhat unusual for a recently proposed name by being an absolute tautonym (i.e., a binomen where the genus is spelled identically to the associated species epithet). However, this choice of name is explicitly endorsed by the International Code on Zoological Nomenclature (ICZN 1999).

Chelys blakii
Comments. Chelys? blakii is based on an isolated right coracoid from the Middle Jurassic of Stonesfield, United Kingdom (Mackie 1863). Lydekker (1889a) tentatively synonymized this taxon with Protochelys stricklandi (Phillips, 1871), which is based on a collection of isolated shell scutes from the same locality. Anquetin and Claude (2008) more recently concluded that all available turtle material from Stonesfield represents a single taxon, Protochelys blakii, which they considered to be a nomen dubium. There is no definitive evidence that the coracoid and the isolated scutes belong to a single taxon, apart from the fact that they were found at the same locality. Therefore, we consider them separately in the present work. Although chelonian in nature, the isolated coracoid from Stonesfield is undiagnostic at the species level, and Chelys? blakii must be regarded as a nomen dubium. The turtle material from Stonesfield, notably the scutes referred to Testudo stricklandi (see below), has been considered by some authors to belong to pleurosternids (Bergounioux 1955;Romer 1956Romer , 1966. However, there is little support for this conclusion, and the present material should be regarded as Testudinata indet. (Anquetin and Claude 2008 Type material. USNM 6548 (holotype), shell fragments (Hay 1910, fig. 1, pl. 10.1-3).
Comments. Desmemys bertelsmanni is based on a relatively complete shell and associated limb and girdle remains that were collected from a clay pit in Gronau, Germany, and described and figured in detail by Wegner (1911). The specimen can confidentially be interpreted as a paracryptodire, as it possesses welldeveloped mesoplastra, in contrast to coeval thalassochelydians (Anquetin et al. 2017), and a finely textured shell, in contrast to coeval helochelydrids . The specimen is otherwise characterized by being relatively small (carapace length approximately 18 cm), by having well-developed fontanelles in the carapace and plastron, and by having radiating scute patterns. It therefore seems all but certain that this is a juvenile individual. Gaffney (1979) noted that Desmemys bertelsmanni is similar to Dinochelys whitei by possessing these radiating scute patterns, but we do not believe this to be particularly meaningful, as Dinochelys whitei is mostly known from juvenile material as well.
To our knowledge, Desmemys bertelsmanni has not been included in a phylogenetic analysis to date. Karl, Nyhuis, et al. (2012) more recently designated a neotype for Desmemys bertelsmanni as they concluded that the holotype had been destroyed during World War II. We here reject this designation, as it disregards nearly all rules set forth by the ICZN (1999) in regard to the designation of a neotype. In particular, the authors do not sufficiently show loss of the holotype, they do not outline why the purported loss of the holotype creates taxonomic instability that needs to be addressed, and they do not discuss why they believe the neotype to be consistent with the morphology of Desmemys bertelsmanni. In addition, the proposed neotype, a fragmentary hyoplastron from the type locality, does not create stability, as it is undiagnostic beyond Testudinata indet. and therefore cannot serve as a meaningful name bearer. We therefore here maintain the status of the purportedly lost holotype. We agree with Jansen and Klein (2014) that the juvenile specimen from Oker, Germany, which had been referred to Desmemys bertelsmanni by Karl, Nyhuis, et al. (2012), is not diagnosable as a paracryptodire and rather represents a eucryptodire instead.
In contrast to all previous authors, we here conclude Desmemys bertelsmanni to be a nomen dubium, as the type specimen represents a juvenile individual. As a result, most bones are poorly ossified, and most scute sulci are not preserved, making it impossible to rigorously compare this taxon with roughly coeval paracryptodires such as Dorsetochelys typocardium, Pleurosternon bullockii, and Riodevemys inumbragigas.
Comments. Dorsetochelys buzzops is based on a fossil turtle from the Late Jurassic of Wyoming (Bakker 1998). Although the type specimen supposedly includes much of the skeleton, it was only documented in the form of a single figure, which consists of three reconstructive illustrations. As the type specimen is now considered lost (J.-P. Cavigelli, pers. comm., 2018), we conclude this taxon to be a nomen dubium, as its morphology cannot be replicated. Comments. Emys obscurus is based on an isolated costal that was likely collected from Late Cretaceous (Maastrichtian) sediments exposed in the vicinity of the type locality in North Dakota, USA (Leidy 1856;. Hay (1908b) referred fragmentary remains from the Late Cretaceous (Maastrichtian) of Montana to Emys obscurus but provided no explicit rationale for that decision, beyond similarities in sculpturing. , on the other hand, synonymized Emys obscurus with Compsemys victa but did not provide an explicit justification for that taxonomic decision. Other paleontologists have ignored this species. The holotypes of Emys obscurus and Compsemys victa are figured on the same plate (Leidy 1860). Whereas the holotype of Emys obscurus is smooth, that of Compsemys victa is characterized by the finely crenulated surface texture noted as diagnostic for that species. We presume that  accidentally viewed the wrong specimen when proposing that Emys obscurus is a junior synonym of Compsemys victa, as the former taxon shows no apparent similarities with the latter. As the holotype is now lost and lacks apparent similarities with other taxa, we here declare Emys obscurus to be a nomen dubium instead. Hay, 1908b nomen dubium Type material. USNM 5731 (holotype), a small, heavily weathered shell lacking all peripherals (Hay 1908b, figs. 33 and 34;Gaffney 1979, figs. 3-5).
Comments. Glyptops depressus is based on a small shell that lacks all peripheral elements (Hay 1908b). Although it seems all but certain that the specimen originates from Colorado, only a note associated with the specimen hints at its possible origin from the Denver Basin (Hay 1908b), which would imply a Late Cretaceous (Maastrichtian) to Paleocene age. The surface of the type is completely worn, and it is therefore possible to establish neither the surface texture of the specimen nor even the placement of sulci. Gaffney (1979) treated Glyptops depressus as a nomen dubium but noted that the large mesoplastra are consistent with a referral to Paracryptodira (his Baenidae). , on the other hand, synonymized Glyptops depressus with Compsemys victa but did not provide an explicit rationale. Given that the morphology apparent in Glyptops depressus is fully consistent with Compsemys victa by having a large entoplastron that is broader than long, large mesoplastra that contact one another along the midline, broad, regular-shaped neurals with anterior short sides, and an anteriorly protruding costal I that forms a V-shaped space for the inset nuchal, we here agree with Lyson and Joyce (2011) that it seems highly plausible that Glyptops depressus is indeed synonymous with Compsemys victa. However, given that the locality data are indeed unclear, we here agree with Gaffney (1979) that it is more prudent to consider this taxon a nomen dubium.

Mesochelys durlstonensis
Type locality. Durlston Bay, Dorset, United Kingdom (Evans and Kemp 1975); Purbeck Limestone Group, possibly Durlston Formation, Berriasian, Early Cretaceous (Allen and Wimbledon 1991;Feist et al. 1995;Milner 2004 We were under the initial impression that none of these names are available in accordance with the rules of the ICZN (1999, art. 11.5, 15), as Seeley (1869, p. xv) specifically noted in the introduction that they were only intended "for convenience" and are "not necessarily to take rank as names of described species. " However, the tentative nature of Seeley's (1869) statements does not allow considering his newly proposed names as unavailable on this ground alone (M. Kottelat, pers. comm., 2019). We therefore concur with previous authors in recognizing the availability of these three names but agree that they are junior synonyms of Pleurosternon bullockii (e.g., Watson, 1910a;Delair 1958;Milner 2004 (Allen and Wimbledon 1991;Feist et al. 1995;Milner 2004).
Comments. Probaena sculpta is based on a near-complete shell from the Morrison Formation of Colorado (Hay 1903). The type locality is now known as the Marsh-Felch Quarry and to be of Late Jurassic (Tithonian) age (Turner and Peterson 1999). The holotype is only 105 mm long and therefore clearly represents a juvenile. We therefore agree with Gaffney (1979) that this taxon should be regarded as a nomen dubium, as juveniles rarely display enough characters to allow diagnosing a valid taxon. We instead refer this fragment to Paracryptodira indet.  (Torrens 1980;Boneham and Wyatt 1993).
Comments. Testudo stricklandi is based on a series of isolated shell scutes from the Middle Jurassic of the Stonesfield Slate housed at the OUMNH (Phillips 1871). Only one of these syntypes can be identified with confidence based on illustrations that accompany the type description (see above). Four additional fossil scutes from Stonesfield are still present at the OUMNH, but it remains uncertain whether these are part of the syntype series (Anquetin and Claude 2008). Lydekker (1889a) proposed the new combination Protochelys stricklandi for this material and referred a collection of similar fossil scutes from Stonesfield housed at the NHMUK. He also tentatively proposed to include an isolated coracoid, the holotype of Chelys? blakii Mackie, 1863(Lydekker 1889a. In accordance with this author, Anquetin and Claude (2008) reassessed all of the available turtle material from Stonesfield as a single taxon, Protochelys blakii, which they considered to be a nomen dubium. However, because there is no evidence that the coracoid and scutes belong to a single taxon (see Chelys? blakii, above), we treat Chelys? blakii and Testudo stricklandi as two separate taxa in the present review.
Testudo stricklandi is known only from isolated carapacial scutes, which represents a rare occurrence in the fossil record (Anquetin and Claude 2008). These vertebral scutes are wider than long, and the pleural scutes are slightly longer than wide.
The configuration is relatively common in stem turtles and basal crown-group turtles and provides few characters of diagnostic value. Therefore, in accordance with Anquetin and Claude (2008), we consider Testudo stricklandi to be a nomen dubium herein.
The turtle material from Stonesfield has been considered by some authors to belong to pleurosternids (Bergounioux 1955;Romer 1956Romer , 1966, but there is no support for this conclusion. This material should therefore be regarded as Testudinata indet. (Anquetin and Claude 2008).