Skull

The slightly deformed skull (ESK-6150; Figure 6) is incompletely preserved. It lacks the right premaxilla, maxilla, postorbital, pterygoid, jugal, palatine, quadrate and prootic, and both sides of the epipterygoids, squamosals, quadratojugals, and basisphenoid. It is long, narrow and arched, and about 63.1 mm long, 21.7 mm wide and 25.9 mm high. Preserved skull elements are damaged to varying degrees. On the dorsal surface, there are no visible sulci of scales on the head, as seen in most testudinids mentioned by Crumly (1984a).

Skull roof elements

Premaxilla.—The left premaxilla is incompletely preserved. It contacts the left maxilla laterally and the vomer posteriorly. On the triturating surface, the labial ridge and the posterior region abutting the vomer are extremely eroded.

The premaxillary ridge, which is well developed and reaches the junction of the premaxillae in Gopherus and Stylemys, is absent (Crumly, 1984a,b; Gaffney and Meylan, 1988; Meylan and Sterrer, 2000), and the premaxillary pit is deep and surrounded by labial and transverse ridges. The ridge is also absent in Mauremys and most testudinids, whereas it is incipiently developed in some Manouria emys and most M. impressa (Crumly, 1984a). The absence of the premaxillary ridge is a plesiomorphic condition (Crumly, 1984a; Meylan and Sterrer, 2000).

The foramen praepalatinum pierces the posterior region of the premaxilla on the triturating surface. It is distinct from the nutritive foramina seen in Mauremys and most testudinids. Meylan and Sterrer (2000) inferred the presence of the distinct foramen praepalatinum as a plesiomorphic character state.

Maxilla.—The left maxilla is missing the lingual part of the triturating surface. It contacts the left premaxilla anteriorly, the left prefrontal dorsally and the left jugal posteriorly. On the triturating surface, the transverse ridge (Meylan and Sterrer, 2000) extends onto the premaxillo-maxillary suture as seen in Astrochelys, some Chelonoidis, some Dipsochelys, Geochelone and Manouria emys. The presence of the transverse ridge in accordance with the premaxillo-maxillary suture is an intermediately derived condition (Meylan and Sterrer, 2000).

The median maxillary ridge extends between the labial and lingual ridges, and is interrupted by the transverse ridge at the anterior end of the maxilla. This condition of the median maxillary ridge is also seen in Astrochelys, Chelonoidis, some Dipsochelys, Geochelone, Indotestudo, Malacochersus, Manouria, Stylemys and some Testudo. The configuration of this ridge in the holotype is intermediately derived (Meylan and Sterrer, 2000), and is somewhat denticulate. A denticulated median maxillary ridge was reported in Chelonoidis denticulata and Manouria emys (Crumly, 1984a).

The lingual ridge is developed on the maxilla in most testudinids except for Chersina, Homopus, Kinixys, Psammobates and Pyxis, and extends from the maxilla to the premaxilla in some Geochelone, Gopherus, Hesperotestudo and Manouria (Meylan and Sterrer, 2000). However, it is not possible to confirm the presence or absence of the lingual ridge on the holotype of Manouria oyatnai sp. nov. Meylan and Sterrer (2000) observed that most testudinids, other than diminutive African forms (i.e., Chersina, Homopus, Kinixys, Psammobates and Pyxis) have the labial, median maxillary and lingual ridges, and the holotype of M. oyamai probably possessed these too.

The posterior maxillary process is absent from the holotype of Manouria oyamai sp. nov., although it is present in most testudinids. This process is also absent in Mauremys, some Astrochelys, some Chelonoidis, some Kinixys, Manouria and Stylemys. The absence of the posterior maxillary process is plesiomorphic (Meylan and Sterrer, 2000).

Prefrontal.—The prefrontals are incompletely preserved with most of the right prefrontal, and the posterior dorsal roof of the left prefrontal is missing. The left prefrontal contacts the left maxilla laterally, the left frontals posteriorly, and the vomer and palatine ventrally. The descending posterior walls of fossa nasalis are partly broken. However, they are well developed ventrally and a relatively wide ethmoid fissure is recognizable. The wide ethmoid fissure has been treated as a synapomorphy of Testudinidae (“ventral processes of the prefrontals” of Loveridge and Williams, 1957; Crumly, 1982; Crumly, 1984a,b; Gaffney and Meylan, 1988, mean processus frontalis circumolfactorius of Bour, 1981). Crumly (1982, 1984a) noted that the ethmoid fissure in Manouria and Psammobates is relatively narrower and less heart-shaped compared with other testudinids. However we observed that Manouria impressa (RH268) and Testudo graeca (ESK-6183 and 6224) have wide ethmoid fissures with undeveloped descending posterolateral wall of the prefrontal. The width of the ethmoid fissure is likely to relate the development of the descending posterolateral wall of fossa nasalis. The character state of the ethmoid fissure in the holotype is similar to that of Manouria emys and Psammobates.

Frontal.—The frontals are well preserved and contact the prefrontals anteriorly and the parietals posteriorly. The left frontal also contacts the left postorbital posterolaterally. Each frontal is trapezoidal dorsally, and is slightly wider than long.

Parietal.—The parietals are preserved, although the descending flange of the left one has been broken away. The parietals contact the frontals anteriorly and the supraoccipital posteriorly. The left parietal contacts the left postorbital laterally, and the right parietal contacts the vomer ventrally. Dorsally, no posterolateral processes extend from the postorbital to the supraoccipital spine.

On the lowest part of the left inferior parietal process, there is a comparatively large opening that runs from within the braincase to the anterolateral parietal flange (see f.c.l. in Figure 6). Bour (1981) identified and described this opening as the foramen caroticum laterale (anterior epipterygoid foramen of Crumly, 1982). The foramen caroticum laterale, in which the canalis caroticus lateralis from the canalis caroticus internus exits through the pterygoid to the sulcus cavernosus, is known in some emydids, geoemydids and testudinids (Albrecht, 1967, 1976). Furthermore, Crumly (1984a) determined that the canalis caroticus lateralis first runs to the sulcus cavernosus, and then pierces the braincase wall anteriorly, in Chelonoidis carbonaria, Homopus areolatus, Manouria and some Psammobates (e.g., P. oculiferus). He also clarified that the foramen caroticum laterale appears on the lowest anteromedial part of the parietal flange in G. carbonaria, and on its lowest anterolateral part in H. areolatus, Manouria and some Psammobates. Bour (1984) also observed this condition in Chelonoidis carbonaria, Cylindraspis vosmaeri, Cylindraspis peltastes, Manouria emys, Pyxis arachnoides, P. planicauda and Homopus areolatus. The foramen caroticum laterale in H. areolatus is located very posteriorly to the parietal flange (Crumly, 1984a). In living Manouria, the foramen is present in the juncture of the pterygoid or epipterygoid, palatine and parietal. In some Psammobates, however, the foramen is present above the epipterygoid, and the pterygoid and epipterygoid do not contribute to the formation of the foramen caroticum laterale (Crumly, 1984a). The character condition in the holotype is similar to living Manouria, and evidently is an intermediately derived condition.

Jugal.—The dorsal portion of the left jugal contacts the left postorbital dorsally and the left maxilla ventrally.

Palatal complex

Vomer.—The vomer is poorly preserved with its posterior part having crumbled away. It contacts the left premaxilla and the maxillae anteriorly, the right anterior palatine laterally, and the basal descending process of the right parietal posteriorly. On the ventral midline, there is a medial vomerine ridge (Crumly, 1982) and the vomer is bent sharply downward in the anterior region. The incomplete vomer in the holotype likely divided the palatine into two parts, as seen in some Mauremys and most testudinids. Meylan and Sterrer (2000) pointed out that the development of the posterior extension of the vomer is quite variable, both among the testudinids and their close relatives.

Pterygoid.—The left posterolateral part of the pterygoid is poorly preserved with its medial and posterior regions crushed and distorted. It contacts the quadrate laterally, but has been dislocated from the basioccipital posteriorly and the prootic dorsally. Ventrally, the left processus interfenestralis is covered by the left pterygoid as seen in Mauremys and most testudinids. The character condition in the holotype is plesiomorphic (Meylan and Sterrer, 2000).

Quadrate.—The left quadrate is incompletely preserved. Laterally, the greater part of the dorsal roof, the lateral wall and the stapedial notch are missing. The processus articularis is partially broken and the condylus mandibularis is damaged. In most testudinids, except for some Chersina, some Dipsochelys, Pyxis arachnoides, and Testudo kleinmanni, the incisura columellae auris, which traverses the stapes, is usually enclosed by the quadrate (Crumly, 1982, 1984a; Gaffney, 1979). Bour (1980) described this condition as “commisura quadrati”. The quadrate contacts the left prootic, opisthotic and pterygoid medially. The anterior region of the quadrate, along with the lateral part of the prootic, combines to form a relatively large trochlear process. An enlarged trochlear process is seen in some Astrochelys radiata, most Chelonoidis, some Geochelone pardalis, some G. sulcata and Manouria (Crumly, 1982, 1984a).

Braincase element

Prootic.—The left prootic is well preserved and contacts the left quadrate laterally and the left opisthotic posteriorly but is separated from the supraoccipital medially. The dorsal exposure of the prootic is triangular in outline, and wider anteriorly than long. A wide, anteriorly expanded prootic is present as seen in some Chelonoidis denticulata, some Indotestudo forstenii and Manouria, and was inferred to be a primitive condition by Crumly (1982; 1984a), although he also observed that the degree of dorsal exposure of the prootic varies considerably. Meylan and Sterrer (2000) did not use the “prootic shape” character state of Crumly (1984a), and redefined this state based on its overall width. Following Meylan and Sterrer (2000), the dorsal exposure of the prootic in the holotype is about equal to, or slightly wider than, its length. It is considered to be a plesiomorphic condition.

Dorsally, the foramen stapedio-temporale is situated in the middle of the lateral margin, contacting the left quadrate that contributes to form the lateral wall of the foramen stapedio-temporale.

Opisthotic.—The left opisthotic is completely preserved. It is separated from the left prootic anteriorly, the supraoccipital medially and the left quadrate laterally, whereas it contacts the left exoccipital medially. The right opisthotic is incompletely preserved, lacking its lateral portion, and is separated from the supraoccipital medially and the right exoccipital posteriorly.

Exoccipitals.—The left exoccipital is completely preserved, but the lateral part of the left one is missing. They contact the supraoccipital dorsally, the opisthotics laterally and the basioccipital ventrally, and form the longitudinally elongated foramen magnum and the relatively slender condylus occipitalis. On the posterolateral surface of the left exoccipital, there are two foramina nervi hypoglossi, a medial larger one and a lateral smaller one.

Basioccipital.—The basioccipital is incompletely preserved, with the majority of it having crumbled away. It contacts the exoccipitals dorsally, and also forms the condylus occipitalis posteriorly.

Supraoccipital.—The supraoccipital is relatively well preserved. It is separated from the left prootic and opisthotics laterally, and the exoccipitals ventrally. The supraoccipital spine is long, extending posteriorly beyond both quadrates. Ventrally, the flattened area of the supraoccipital spine rapidly reduces at a point that is one-fourth the distance from the entrance of the foramen magnum to the posterior end of the supraoccipital spine. As a result, approximately three-quarters of the posterior potion of supraoccipital spine forms a blade-like edge. This condition is also seen in Astrochelys yniphora, most Geochelone pardalis and G. sulcata, and some Manouria emys (Crumly, 1982).

Lower jaw

The lower jaw (Figure 7) is represented by the posterior half of the right side. The preserved lower jaw elements comprise the dentary, the coronoid, the surangular, the articular and the prearticular.

Dentary.—The right side of the dentary is missing its anterior half and the labial and lingual ridges on the triturating surface are slightly eroded. It contacts the coronoid dorsally, the surangular and the prearticular posteriorly, and the prearticular medially. Laterally, the foramen dentofaciale majus is located along with a few nutritive foramina. Medially, the sulcus cartilaginis meckelii is relatively well developed.

Coronoid.—The right coronoid is completely preserved. It is overlapped laterally by the dentary, and contacts the surangular ventrally and the prearticular medially.

Surangular.—The right surangular is preserved, except for the area around the foramen nervi auriculotemporalis, which is partially broken. It contacts the dentary anteriorly, the coronoid dorsally, and the articular posteriorly. The surangular does not interdigitate with the dentary, as seen in Gopherus, Hesperotestudo, Manouria, Psammobates, Pyxis and Stylemys (Crumly, 1982, 1984a; Meylan and Sterrer, 2000). The condition seen in the holotype is plesiomorphic. Moreover, the postdentary elements in the holotype show up as quite massive, as seen in Gopherus, Hesperotestudo and Manouria (Crumly, 1982; 1984a). The massive postdentaries may also be apo-morphic.

Articular.—The right articular is preserved. It contacts the surangular and the angular laterally, and prearticular medially.

Prearticular.—The right prearticular is incompletely preserved. It contacts the coronoid dorsally and the dentary anteriorly and the articular posteriorly.

Shell

The shell of the holotype, ESK-6150, is incompletely preserved (Figure 8). It is made up of four neurals (the fourth, fifth, seventh and eighth), twelve costals (the left first, the right second and third, the fourth and fifth pairs, the left sixth, the seventh and eighth pairs) and four peripherals (the right fourth to seventh). The shell is flattened and is covered by the first, third, fourth and fifth vertebral scutes, the left first, right second, paired third and fourth pleural scutes, and the fifth to eighth marginal scutes. Two nuchals (ESK-6121, 6132), three costals (ESK-6122-6124) and two peripherals (ESK-6125, 6152) are present in reference specimens. ESK-6121 is covered by the first vertebral and a pair of the first marginal scutes. ESK-6132 is covered by a cervical, the first vertebral and a pair of the first marginal scutes.

The specimens of the plastron are also incompletely preserved. ESK-6151 is the posterior half of the right side of the plastron, composed of the left hypoplastron and the xiphiplastron. The right posterior half of the plastron is covered by the right abdominal, femoral, inguinal, and anal scutes. The other plastral materials, ESK-6153 and 6154, are fragments in which the original arrangement is unclear.

As in living Manouria (Manouria emys, ESK-6170, RH68; M. impressa, RH268), no concentric growth lines of the scutes sculpture the exterior surface of the shell in the holotype. The scute sulci on the shell are also extremely wide (about 3.1–3.8 mm on the carapace and 3.9–6.0 mm on the plastron) and shallow without distinct edges along the sulci in comparison with other testudinids in which the scute sulci are usually well impressed with distinct edges except for Dipsochelys. By contrast, M. emys (ESK-6170, carapacial length; 500 mm) and M. impressa (RH268, carapacial length; 240 mm) have narrow scute sulci (about 1.0–2.6 mm and about 1.0 mm wide on the carapace, respectively) with distinct edges. All shell elements are extremely thin.

Carapace.—The nuchal, ESK-6121, from Hinigusuku, Kitanakagusuku village, Okinawa Island, is missing its anterior part, where the single cervical and a pair of the first marginal scutes covered dorsally. The preserved part of nuchal is flattened and covered by the first vertebral and a pair of the first pleural scutes. It is 66.9 mm in maximum length, 46.0 mm in maximum width, 9.9 mm thick in the anterior thickened zone and 1.6 mm thick in the posterior region.

Another nuchal, ESK-6132, from Kojima, Isen Town, Tokunoshima Island, is missing the left-lateral side and the posterior region. It is 33.6 mm in maximum length and 36.3 mm in maximum width, 5.7 mm thick in the anterior thickened zone and 1.0 mm thick in the posterior edge, and is emarginated anteriorly. Dorsally, it is overlapped by a cervical, the first vertebral and a pair of the first marginals. The cervical is long and narrow dorsally, and invisible ventrally. In primitive fashion, the dorsal cervical scute is wider than long, as also seen in Mauremys japonica, Astrochelys, Dipsochelys, Gopherus, some Homopus, some Kinixys, some Malacochersus, Manouria, and some Psammobates. Although this character is slightly variable, the elongated cervical scute seen in ESK-6132 is an inter-mediately derived condition (Crumly, 1984a; Meylan and Sterrer, 2000).

In all the specimens, the nuchal is extremely thin, even on the ventral thickened zone, compared with the other testudinids. The inferred (restored) shape of the nuchal, based on ESK-6121 and ESK-6132, is hexagonal in outline, with short lateral articulations directed anteriorly.

The neurals in the holotype, ESK-6150, are poorly preserved, and are very thin, about 0.7–2.4 mm thick. The fourth neural, which contacts the left fourth costal laterally and the fifth neural posteriorly, is lacking most of the anterior and right lateral regions. The fifth neural, which contacts the fourth neural and the left fifth costal, lacks most of its periphery. The dorsal vertebrae of the fourth and fifth neurals are missing. The seventh neural, which contacts the left seventh costal laterally, the eighth neural posteriorly and the eighth dorsal vertebra ventrally, is missing most of the anterior and right lateral areas. The eighth neural, which contacts the seventh neural anteriorly, and the left seventh and a pair of the eighth costals laterally, is missing the right lateral and posterior fringes and the sacral vertebrae. It is elongate-hexagonal, with short articulations directed anteriorly.

The left first, the right second and third, the pairs of fourth and fifth, the left sixth, and the pairs of seventh and eighth costals are preserved in ESK-6150. Most of them are about 0.7–.2 mm thick, but the medial edge of the first costal is 5.9 mm thick. The width of the fourth costal decreases from the proximal to the distal end. Conversely, the width of the fifth costal increases from the distal to the proximal end. This costal wedging pattern is seen in almost all testudinids (Auffenberg, 1974). Dorsally, the sulci between the pleurals are inscribed on the left side of the first, fifth, sixth, eighth, and the right side of the fourth and eighth costals. These even-numbered costals are slightly concave or flattened, whereas all odd-numbered costals are flattened. In Manouria emys (ESK-6170; RH68) and M. impressa (RH268), the even-numbered costals are convex, while the odd-numbered costals are flattened.

The left first costal has lost its fringe area, and its exterior surface is covered by the first vertebral and first pleural scutes. The sulcus on the first vertebral scute curves medially from the posterior to anterior.

The right second costal contacts the third costal posteriorly. The third costal contacts the second costal anteriorly and the fourth costal posteriorly. The left fourth costal contacts the fourth neural medially and the left fifth costal posteriorly. The right fourth costal contacts the third costal anteriorly and the fifth costal posteriorly. The width of the third costal decreases from the distal to the proximal end. The right fifth costal contacts the fourth costal anteriorly and the ascending process of the right inguinal buttress ventrally.

The rib ends of these costals disappear at the border of the fifth, sixth and seventh peripherals. On the other hand, the proximal processes of the fifth, sixth and seventh peripherals extend medially at the distal ends of the respective costals in dorsal view. The relationship between costals and peripherals is known for most testudinids (Auffenberg, 1974). The left seventh costal contacts the seventh and the eighth neurals medially, the sixth costal anteriorly and the eighth costal posteriorly. The right seventh costal contacts the eighth costal posteriorly. The eighth costals contact the seventh costals anteriorly and the eighth neural medially. Ventrally, the ilium-attachment scars are located on the medial part of the internal surface of the proximal eighth costals. In Hesperotestudo bermudae, and some Homopus areolatus, this scar is located on the proximal part of the seventh and the eighth costals (Meylan and Sterrer, 2000).

The right fourth, fifth, sixth and seventh peripherals forming the “bridge” are relatively well preserved, except for the distal parts. They are about 1.9–2.1 mm thick. These peripherals are extremely thin and are sutured in order. The right fifth and sixth peripherals contact the right third and the fourth costals, respectively, by the dorsally extended distal peripheral processes. These peripherals do not coalesce with the costals. The overlapping of the fifth, sixth, seventh and eighth marginal scutes on the peripherals do not reach the costals. The sulci between the pleural and marginal scutes coincide with the border of the costals and the peripherals. It is thin besides the thickened part and is not reverted.

Moreover, two peripherals, ESK-6125, from Hinigusuku, and ESK-6152 from Bise, Okinawa Island are preserved except for their proximal parts, so it is not possible to determine their original locations. They are uniformly thin (7.7 mm in ESK-6125 and 7.9 mm in ESK-6152 even on the thickened zone), and slightly reverted. Each peripheral is overlapped dorsally by two marginal scutes. The shallow and wide vertical scute sulci are also visible. The overlapping areas covered by the marginal scutes ventrally are quite narrow, and do not reach the thickened zone.

Plastron.—The plastral materials are sparse. The holotype, ESK-6150, includes the ascending process of the right inguinal buttress. ESK-6151, one of the referred specimens, consists of the right posterior half of a plastron.

The ventral part of the right inguinal buttress is missing on the holotype, ESK-6150. This buttress is very thin, slender and embedded in the ventral surface of the right fifth costal. The inguinal buttress is embedded in the distal part of the right posterior fifth costal, as also seen in Mauremys japonica, some Indotestudo, Kinixys, Psammobates and some Stylemys. Meylan and Sterrer (2000) suggested that the condition in which the inguinal buttress contacts the fifth costal is a primarily derived one.

ESK-6151, from Kamikurukubaru, is relatively well preserved, consisting of the incomplete right hypoplastron and the xiphiplastron (lacking their medial parts), which contact each other. These are very thin (about 10.4 mm in maximum thickness in the inguinal notch, and 2.3 mm in minimum thickness in the medial part of the right xiphiplastron), including the dorsolateral margins covered by the scutes. The right hypoplastron is covered by the right abdominal, femoral and inguinal scutes. The right xiphiplastron that is covered by the femoral and anal scutes is notched posteriorly, also covered by the femoral and anal scutes.

The abdominal scute contacts the femoral and inguinal scutes posteriorly. The right femoral scute is large, being about three times as long as the anal scute, and broadly contacts the inguinal scute anterolaterally on the plastral surface. In primitive fashion, the abdominal scute intervenes between the inguinal and femoral scutes, as also seen in Mauremys, Homopus, some Psammobates and Stylemys (Bramble, 1971; Meylan and Sterrer, 2000). The character condition seen in the holotype is apomorphic (Meylan and Sterrer, 2000).

The inguinal scute is vertically divided into two portions, a large anterior one and a small posterior one. The anterior one contacts the abdominal, and does not contact the femoral, whereas the posterior one contacts the abdominal and femoral. In emydids and geoemydids, the inguinal scute is single (Hirayama, 1984). In Manouria, the inguinal scute is obviously divided into two or three parts (Crumly, 1984a), and divided inguinal scutes are also seen in some Geochelone pardalis, G. sulcata, Gopherus agassizii, Homopus areolatus, H. femoralis, Malacochersus, some Psammobates tentorius, Testudo graeca, T. kleinmanni (Loveridge and Williams, 1957; Bramble, 1971). Divided inguinal scutes are considered to be an apomorphic condition (Crumly, 1984a).

Pectoral girdle

Scapula.—The entire scapular prong and dorsal half of the acromion process are missing from the right scapula of the holotype, ESK-6150. The glenoid fossa of the scapula is shallow and narrow. The angle between the acromion process and the body of the scapula is estimated at approximately 110°–120° as seen in Astrochelys, Dipsochelys spp., Geochelone, some Gopherus, some Hesperotestudo, and Psammobates (Bour, 1994; Meylan and Sterrer, 2000). Bour (1984) also observed this condition in Cylindraspis vosmaeri, Cylindraspis peltastes, Chelonoidis carbonaria, Manouria emys, Pyxis arachnoides, Pyxis planicauda, and Homopus areolatus. Gerlach and Canning (1998) recorded that the angle in Dipsochelys spp. also measures 100°–130° variably in lateral views, however the measurements used in Bour (1984) and Gerlach and Canning (1998) are incomparable to those of Meylan and Sterrer (2000) because they did not show the measuring criterion of the angle for the samples. Therefore, we follow the data shown by Meylan and Sterrer (2000) tentatively in the present study. The angle of the acromion process in the holotype is considered to be a primarily derived condition.

Coracoid.—Both coracoids are preserved in the holotype, ESK-6150. The right coracoid is about 42.9 mm long, but the coracoid blade is crushed and fragmented. The left coracoid is about 42.6 mm long and 23.0 mm wide, but lacks its central part of the blade. The preserved length of the left coracoid, ESK-6126, from Hinigusuku, is about 52.8 mm, but it is missing the proximal anterior blade.

The blades of these coracoids are fan-shaped as seen in all testudinids (Auffenberg, 1974). This character has also been considered as a synapomorphy in the Testudinidae (Auffenberg, 1974; Crumly, 1984a, b; Gaffney and Meylan, 1988).

Humerus

The left humerus of the holotype, ESK-6150, is about 64.9 mm in preserved length, but is missing the proximal part of the lateral process and the distal articular surface.

The humeral shaft is cylindrical, relatively slender, and somewhat curved ventrally from the proximal to the distal end. The caput humerus is long and narrow, and is oval in outline, as seen in Manouria. The medial process (i.e. the greater trochanter of Meylan and Sterrer, 2000) does not extend beyond the caput humerus, as seen in Mauremys, some Gopherus, Hesperotestudo bermudae, Malacochersus and Manouria. This character, as seen in the holotype, is plesiomorphic (Meylan and Sterrer, 2000). In ventral view, the intertubercular fossa between the medial and lateral processes is relatively narrow and long, as it is in Manouria.

There is a shallow scar lacking the distinct border on the dorsolateral surface of the proximal shaft of the humerus (Figure 9). This scar is regarded as the insertion of a tendon formed by the convergence of the latissimus dorsi and teres major muscles (Walker, 1973), and is usually invisible in emydids, geoemydids and most testudinids (Crumly, 1984a; Gaffney and Meylan, 1988; Meylan and Sterrer, 2000). Crumly (1984a) inferred this to be the latissimus dorsi scar, and described the latissimus dorsi muscle inserting onto the proximal shaft on the ulnar (medial) side of the humerus among testudinids, except for Manouria that usually lacks this scar. He also noted that emydids and geoemydids usually do not have a latissimus dorsi scar even in large-size groups, and that chelydrids have a visible scar on the lateral side of the humeral shaft. He concluded that the muscle scars seen in chelydrids and testudinids are not homologous. Meylan and Sterrer (2000) observed for Testudinidae that there is no obvious latissimus dorsi scar in Gopherus, Hesperotestudo, Manouria, and Stylemys. In the present study, Emys orbicularis, Mauremys and most of the testudinids were dissected and the location and development of the “latissimus dorsi scar” was checked. Our observations show that the latissimus dorsi and teres major muscles insert on the dorsal surface of the distal part of the lateral process at least in Emys, Mauremys, Dipsochelys, Manouria, and Testudo horsfieldii. In contrast, the muscles insert on the distal part of the medial process in Chelonoidis, Geochelone, Indotestudo, Kinixys, Malacochersus and Testudo except for T. horsfieldii (Figure 10). With regard to extinct testudinid species, the latissimus dorsi and teres major scar is located on the lateral process in Stylemys nebrascensis (Hay, 1908, fig. 491) and on the medial process in Testudo cubensis (Williams, 1950, pl. 5). Therefore, the latissimus dorsi and teres major scar seen in chelydrids, emydids, geoemydids and testudinids is likely to be homologous, and the character condition seen in the holotype is considered to be plesiomorphic.

Figure 10

Left pectoral muscles of Emys orbicularis (Emydidae: 1) and Geochelone pardalis (Testudinidae: 2). The latissimus dorsi and teres major insert on the lateral process in E. orbicularis, and on the medial process in G. pardalis.

Our dissections and observations also reveal that this muscle scar is undeveloped in Emys, Mauremys, some Chelonoidis, some Dipsochelys, Kinixys, Manouria and Testudo graeca. The muscle scar is present in other examined testudinids. Therefore, the presence of the developed latissimus dorsi and teres major scar is considered to be apomorphic.

There is a distinct ectepicondylar groove on the anterior part of the dorsal right distal humerus. Unfortunately, the presence or absence of the ectepicondylar foramen is not confirmed, because the distal articular surface of the humerus has been considerably eroded. In Mauremys, some Astrochelys, some Chelonoidis, some Hesperotestudo, some Indotestudo, some Kinixys, Malacochersus, some Manouria, Stylemys and Testudo horsfieldii, the ectepicondylar foramen opens with a groove. In Chersina, some Hesperotestudo, Homopus, some Indotestudo, some Kinixys, some Manouria, Psammobates, Pyxis and Testudo graeca, both ectepicondylar foramen and groove are absent (incorrectly described as an “entepicondylar foramen” by Crumly, 1984a and Meylan and Sterrer, 2000). Crumly (1984a) considered the presence of the ectepicondylar foramen to be primitive, whereas Meylan and Sterrer (2000) considered its absence to be primitive. The coding for the character in Meylan and Sterrer (2000) is probably incorrect, because the ectepicondylar foramen is usually present with a groove in emydids, geoemydids and other outgroup taxa such as trionychoids, chelonioids, cheridrids, etc. used for testudinids.

Ulna

The right ulna, ESK-6129, from Hinigusuku, is missing the proximal end. The biceps scar is poorly developed, and located about two-thirds of the distance from the distal to the proximal medial surface.

Femur

The right femur from Hinigusuku, ESK-6130, lacks the fibular condyle and the trochanter minor, while the caput femoris, trochanter major, and tibial condyle are considerably eroded. This femur is 80.7 mm in length, 26.5 mm in proximal width, 23.9 mm in maximum measurable distal width, and 9.9 mm in minimum shaft width. The trochanter major and minor are coalesced by a ridge, which has been considered as a synapomorphy for Testudinidae (Auffenberg, 1974). The intertrochanteric fossa is deep, and longer than wide.

The distal part of the right femur, ESK-6138, from Kamikurukubaru, has a ridge developed between the tibial and fibular condyles on the distal articulation surface, as seen in Mauremys and Manouria.

Tibia

The right tibia, ESK-6131, from Hinigusuku, is slightly eroded on both proximal and distal ends. It is nearly straight from the distal to proximal ends in outline, measuring 59.6 mm long, 13.6 mm wide proximally and 18.3 mm wide distally. The cnemial crest is not developed.