Osteology and taxonomic revision of Hyphalosaurus (Diapsida: Choristodera) from the Lower Cretaceous of Liaoning, China

Abstract

Although the long-necked choristodere Hyphalosaurus is the most abundant tetrapod fossil in the renowned Yixian Formation fossil beds of Liaoning Province, China, the genus has only been briefly described from largely unprepared specimens. This paper provides a thorough osteological description of the type species Hyphalosaurus lingyuanensis and the con-generic species Hyphalosaurus baitaigouensis based on the study of fossils from several research institutions in China. The diagnoses for these two species are revised based on comparison of a large sample of specimens from the type area and horizon of each of the two species. The skull, better known in H. baitaigouensis, exhibits key choristodere synapomorphies including an elongate contact between the prefrontals and posteriorly expanded supratemporal fenestrae that strongly support the placement of the highly derived hyphalosaurids within Choristodera. Both species of Hyphalosaurus share a proportionally small head, an elongate neck, a relatively unspecialized appendicular skeleton and a long, dorsoventrally heightened tail. Soft tissue preservation in several specimens provides rare insight into the integument of an extinct group. The integument of Hyphalosaurus is made up of small polygonal scales with several parasagittal rows of large, keeled, ovoid scutes. These possibly ornamental scutes bear a strong resemblance to the rows of large scutes in the monjurosuchid choristodere Monjurosuchus splendens. Observations from a variety of growth stages reveal that significant ontogenetic change in the proportions of the body and limb bones occurred in both species of Hyphalosaurus.

Introduction

The genus Hyphalosaurus includes two species of highly specialized aquatic choristoderan reptiles with unusually long necks. Fossils of this genus are known only from the Early Cretaceous Yixian Formation of western Liaoning Province, China. The type species, Hyphalosaurus lingyuanensis, is known from multiple specimens from a single locality in the Lingyuan area (Fig. 1), where the fossil beds of the lower Yixian Formation have yielded a radiometric date of 123–126 Ma (Smith et al. 1995; Ji et al. 2004a). The fossil beds exposed at the Lingyuan area are also known for producing superbly preserved fossils including feathered dromaeosaurs (Ji et al. 2001), a juvenile enantiornithine bird (Hou & Chen, 1999; Zhang et al. 2003), and multituberculate and eutherian mammals (Hu & Wang, 2002; Ji et al. 2002). The second species, Hyphalosaurus baitaigouensis, is known from far more abundant material from the Yizhou area (Fig. 1), where multiple localities exposing strata from the upper part of the Yixian Formation (Wang et al. 2004; but see also Ji et al. 2004a) have yielded at least several thousand specimens of this species, including soft-shelled fossil eggs (Ji et al. 2004b). These two species, along with Shokawa ikoi from the Early Cretaceous Okurodani Formation of Japan (Evans & Manabe, 1999), form the family Hyphalosauridae (Gao & Fox, 2005). In addition to features such as dorsally positioned orbits and pachyostotic ribs that are widely distributed within Choristodera, hyphalosaurids are characterized by a small head, a long neck consisting of more than 16 cervical vertebrae, and a greatly elongated and heightened tail. This combination of traits suggests that hyphalosaurids were the most completely aquatic choristodere clade. Taphonomy also supports this inference, as Hyphalosaurus is common in deep-water lacustrine facies bearing the large chondrostean fish Protopsephurus in the Lingyuan area and co-occurs with the abundant large shrimp Liaoningogriphus in the Yizhou area (Shen, 2003). Interestingly, Hyphalosaurus is absent from the Chiufotang Formation exposed near the city of Chaoyang (Fig. 1). These deposits appear to represent a shallow-water swampy environment and contain the blunt-snouted monjurosuchid choristodere Philydrosaurus proseilus (Gao & Fox, 2005; Gao et al. 2007) and gavialiform neochoristoderes (unpublished specimens LPMC R00070, 00071), as well as birds and non-avian dinosaurs.

Fig. 1.

Map showing distribution pattern of Hyphalosaurus fossils in western Liaoning. Note the absence of Hyphalosaurus from the Chaoyang area (satellite image of the area from TAGEO.com).

Thousands of Hyphalosaurus specimens are currently deposited in research collections or in private hands. Despite the surfeit of material, the anatomical details of these animals have scarcely been touched upon. The first account of Hyphalosaurus briefly described the type species H. lingyuanensis, at that time considered Diapsida incertae sedis (Gao et al. 1999). Following the recognition of the choristoderan affinities of the genus (Gao et al. 2000), a second species, H. baitaigouensis, was described (Ji et al. 2004b). However, the original publication of the second species lacks essential anatomical treatment of the skull, and, more unfortunately, several skull elements are misidentified and mislabeled in the published figure (see discussion below). The two species are most easily distinguished by the number of cervical vertebrae, but several other features (listed below) are also diagnostic, as recognized in this study. In this paper, we present a detailed description of the osteology of Hyphalosaurus, primarily from newly prepared specimens from several institutions including the BMNHC, IVPP, LPMC, and Peking University Paleontological Collections. The holotype specimen of H. lingyuanensis(IVPP V11075) was incompletely exposed when the original description of the species was completed. Further preparation of the holotype and examination of additional previously undescribed specimens of H. lingyuanensis allow us to provide new observations on the osteology of the type species, including previously unknown information on the skull roof and integument of this species. Based on examination of a large number of well-preserved specimens in several institutions, we also provide a revision of H. baitaigouensis, including an emendation of the diagnosis and clarification of its cranial and postcranial anatomy. This study provides essential information for understanding the evolution of Hyphalosauridae in Asia as a highly specialized family of aquatic diapsids, and the phylogeny of the Choristodera, now recognized as an important clade in Mesozoic terrestrial ecosystems throughout the Northern Hemisphere.

Abbreviations

Institutional: AMNH – American Museum of Natural History, New York, NY, USA; BMNHC – Beijing Museum of Natural History, Beijing, China; CAGS – Chinese Academy of Geological Sciences, Beijing, China; GMC – Geological Museum of China, Beijing, China; IVPP – Institute of Vertebrate Paleontology and Paleoanthropology, Beijing, China; LPMC– Liaoning Paleontological Museum of China, Shenyang, China; PKUP – Peking University Paleontological Collections, Beijing, China.

Anatomical: ar, articular; as, astragalus; bo, basioccipital; ca, calcaneum; ch, choana; cl, clavicle; co, coronoid; cor, coracoid; cv, cervical vertebra; den, dentary; dc, distal carpal; dt, distal tarsal; dv, dorsal vertebra; ect, ectopterygoid; ectf, ectepicondylar foramen; fem, femur; fi, fibula; fr, frontal; hu, humerus; icl, interclavicle; il, ilium; intv, interpterygoid vacuity; isc, ischium; j, jugal; lac, lacrimal; m, maxilla; mc, metacarpal; mt, metatarsal; na, nasal; neo, neomorph; pa, parietal; pal, palatine; palf, palatal foramen; par, parasphenoid; pm, premaxilla; porf, postorbitofrontal; prf, prefrontal; ptg, pterygoid; pub, pubis; qj, quadratojugal; qu, quadrate; rad, radius; sa, surangular; sc, scapula; sv, sacral vertebra; sq, squamosal; ti, tibia; ul, ulna; vo, vomer.

Systematic palaeontology

Class Reptilia Laurenti, 1768

Subclass Diapsida Osborn, 1903

Order Choristodera Cope, 1876

Family Hyphalosauridae Gao and Fox, 2005

Type genus: HyphalosaurusGao et al. 1999

Family diagnosis (revised from Gao & Fox, 2005): Distinct choristoderan family characterized by having proportionally small skull, infratemporal fenestra closed in adult individuals, greatly elongated neck consisting of 16–24 cervical vertebrae, and development of triangular and spike-like neural spine of the caudal vertebrae.

Constituent genera: The family is currently known from two genera, including the type genus Hyphalosaurus from western Liaoning Province of China and the closely related genus Shokawa from Japan (see remarks below).

Known distribution: Early Cretaceous lacustrine deposits in East Asia, including western Liaoning Province of China and central Japan.

Remarks: Although highly specialized for an aquatic ecology, Hyphalosauridae share many character states that support the inclusion of the family group in the order Choristodera (Evans & Manabe, 1999; Gao & Fox, 2005). Within the Choristodera, the family occupies a position outside the Neochoristodera and may represent the sister group of the Monjurosuchidae or a more basal clade (Gao & Fox, 2005). New evidence revealed in this study will help to resolve the relationships of the group within the Choristodera in future cladistic analyses.

The Early Cretaceous Shokawa ikoi from Japan shares many postcranial features with Hyphalosaurus. Because the skull is almost completely unknown and important postcranial characters such as the cervical vertebral count also remain uncertain for the Japanese taxon, it is not possible to provide a coherent diagnosis separating the two genera at present. We retain the name Shokawa here, pending the discovery of additional materials that can help clarify the anatomy and taxonomic status of this genus.

Genus HyphalosaurusGao et al. 1999

Type species: Hyphalosaurus lingyuanensisGao et al. 1999.

Constituent species: The type species and H. baitaigouensis (see discussion below).

Generic diagnosis: The genus is diagnosed by derived character states including an elongated neck consisting of 19–24 cervical vertebrae and subequal length of the third and fourth metatarsals. Because the skull of Shokawa remains almost entirely unknown, it is uncertain which features of the skull diagnose Hyphalosaurus and which are shared between the two taxa.

Known distribution: Known only from the Early Cretaceous Yixian Formation, western Liaoning Province, China.

Species Hyphalosaurus lingyuanensisGao et al. 1999

Synonym: Sinohydrosaurus lingyuanensisLi et al. 1999 (see Smith & Harris, 2001 for discussion).

Holotype: IVPP V11075, nearly complete skull, mandibles, and well-preserved postcranial skeleton; BMNHC V398 (BPV-398), counterpart impressions of IVPP V11075. The part and counterpart of the same specimen were acquired independently by the two institutions; however, both parts are here treated as the holotype, as they are of the same individual.

Type locality and horizon: Fanzhangzi, approximately 20 km southwest of the city of Lingyuan, Liaoning Province, China; Early Cretaceous Yixian Formation, radiometrically dated to 123–126 Ma (Smith et al. 1995; Ji et al. 2004a).

Referred specimens: PKUP V1052, GMC juvenile (Fig. 3); both topotypic specimens from the type locality near Lingyuan as described above.

Fig. 3.

Hyphalosaurus lingyuanensis: part and counterpart of an unnumbered juvenile specimen in GMC vertebrate fossil collections from Lingyuan; Early Cretaceous Yixian Formation.

Known distribution: Known only from the type locality and horizon.

Diagnosis (revised from Gao et al. 1999): Hyphalosaurid choristodere differing from the congeneric species H. baitaigouensis by possession of the following autapomorphies: postorbital part of skull substantially shorter than anterior part including snout and orbit; frontal-parietal suture located at level slightly posterior to the anterior border of supratemporal fenestrae; postorbitofrontal not contacting parietal; elongated neck consisting of 19 cervical vertebrae; dorsal vertebrae 16 in number; interclavicle essentially T-shaped, with great reduction of anteromedial process; iliac blade oriented horizontally.

Remarks: Hyphalosaurus lingyuanensis is known from several specimens (Figs 2, 3) from the type Lingyuan area. Since the original publication by Gao et al. (1999), little progress has been made towards understanding the anatomical details of this morphologically peculiar animal. With further preparation of the holotype and the discovery of new specimens, it is now possible to provide a thorough anatomical description of the species. A comparison with the congeneric species H. baitaigouensis in this study reveals new information to distinguish the two species from one another, leading to the revision of the diagnosis of the type species presented above. As a result of this revision, two features have been removed from the species diagnosis. One feature was ‘pachyostotic dorsal ribs strongly sigmoid, imbricated distally with posteroventral extension’, but it is now clear that the ventral imbrication of the ribs is preservational. Similar distortion is seen in a few specimens of H. baitaigouensis (e.g. LPMC R-00065 and R-00066). The second feature, ‘third and fourth metatarsals subequal in length’ (Gao et al. 1999), is now also known from specimens of H. baitaigouensis (e.g. BMNHC V050) and thus no longer diagnostic at the species level.

Fig. 2.

Hyphalosaurus lingyuanensis: A) IVPP V11075 (holotype), nearly complete skeleton preserved in volcanic shale slab; B) PKUP V1052, topotypic specimen from Lingyuan (Early Cretaceous Yixian Formation). Both specimens with posterior tail omitted.

Osteology of Hyphalosaurus lingyuanensisGao et al. 1999

Cranial skeleton

Comparative study of several specimens (IVPP V11075, uncataloged juvenile specimen from the GMC, PKUP V1052) consistently shows a roughly pear-shaped skull configuration, with a pointed snout, laterally bulging orbital region and short posterior expansion of the supratemporal fenestra. This skull configuration is consistent through ontogenetic development, as both the smallest specimen examined (GMC-juvenile specimen) and the largest specimen (IVPP V11075) show essentially the same configuration (Figs 2–4). The external narial openings are small, paired and anteriorly placed. The orbits are proportionally large, and dorsally oriented. The supratemporal openings are small (less than one-half size of the orbit), while the infratemporal opening is closed. The postorbital portion of the skull is significantly shorter than the anterior part of the skull, including the snout plus the orbit. Such proportions of the skull are different from those in H. baitaigouensis, in which the postorbital skull accounts for roughly one-half of the skull length (see below).

Fig. 4.

Skull of Hyphalosaurus lingyuanensis: A) palatal view of the holotype skull IVPP V11075; B) dorsal view of the skull of a juvenile specimen in GMC collections.

Skull roof: The morphology of the skull roof was hitherto unknown for H. lingyuanensis, as the holotype skull is exposed in palatal view (Fig. 4A) and the skull of one other specimen (PKUP V1052) was split along the horizontal plane as the slab encasing it was excavated. Fortunately, a new GMC juvenile specimen from the type locality shows the relationships of the many elements of the skull roof with clarity (Fig. 4B).

The premaxillae and nasals are not well exposed in any of the three available specimens, and thus the detailed morphologies of these skull elements cannot be ascertained at this stage. However, the nasals were necessarily short and narrow in keeping with the short and pointed snout.

The paired prefrontals are elongated with an extensive midline sutural contact. This feature is considered a key synapomorphy of Choristodera (Gao & Fox, 1998). The prefrontal-frontal suture is located slightly posterior to the anterior borders of the orbits. The frontals are also paired, and strongly narrowed anteriorly to form a narrow interorbital bar as seen in H. baitaigouensis. The frontals, however, are posterolaterally widened with an extensive, slanted sutural contact with the fused postorbitofrontal. Posteriorly, the frontal has a transverse suture with the parietal at a level slightly posterior to the anterior border of the supratemporal fenestra, and in this aspect it is different from H. baitaigouensis (see below). The parietals are paired, forming a roughly squared table with very short supratemporal processes. The lateral border of the parietal is slightly concave, forming the medial border of the small supratemporal fenestra. A parietal foramen is absent as in other choristoderes.

The maxilla (best shown on the right side of GMC juvenile) lacks a clearly defined dorsal process, and instead the medial portion takes the form of a low, inwardly scrolled flange. The anterior margin of the flange forms the posterior border of the external narial opening, and the posterodorsal margin of the flange meets the prefrontal and lacrimal. The jugal has an anterior process that forms a large part of the lateral rim of the orbit, and a well-defined dorsal process that contacts the fused postorbitofrontal. The extent of the posterior process cannot be ascertained. The postorbital and postfrontal are fused into a postorbitofrontal. The postorbitofrontal forms the anterior border of the small supratemporal fenestra, but appears not to have a contact with the parietal. This condition differs from that in H. baitaigouensis, which has a small contact between the postorbitofrontal and parietal.

Palate: The holotype of H. lingyuanensis (IVPP 11075) exposes the palate, but this region was unprepared at the time the species was named and described. The holotype has since been further prepared (Fig. 4) and restudy of this specimen together with observations from PKUP V1052 (a topotypic specimen of the same species) provide a better understanding of the palate of hyphalosaurids.

The paired vomers share a long midline contact. The vomer forms the entire medial and anterior border of the choana, and sends out a short process to form a portion of the lateral border as well. In PKUP V1052, a single row of small palatal teeth is preserved on the right vomer. The choanae are only slightly posteriorly displaced relative to the nares. Although the contact between the vomer and pterygoid cannot be deciphered with absolute certainty, it appears to occur at the posterior border of the choana.

The palatines form the major portion of the lateral and posterior borders of the choanae and meet the pterygoids in a long, straight sutural contact medially. The right palatine preserves a few palatal teeth near its anterior tip in IVPP 11075 (Fig. 4A). Some parts of the palatal dentition have clearly been lost to damage, so the scarcity of palatal teeth in this specimen and in PKUP V1052 is most likely due to preservational effects. Damage caused by the splitting of the slab makes it difficult to discern whether a nasopalatal trough was present, and also precludes identifying the palatine foramen.

The pterygoids are identifiable in the holotype and the left element is exposed in dorsal view in one slab of PKUP V1052. In the holotype, the pterygoids are separated from one another over most of their lengths, and show only a limited anterior contact. However, the asymmetry of the opening between the pterygoids suggests the two elements were displaced from one another by deformation. If so, the pterygoids may have shared a more extensive midline contact as in H. baitaigouensis. It is unclear from either specimen whether the pterygoids or the parasphenoid complete the posterior border of the vacuity.

The suborbital fenestra is roughly subtriangular and slightly elongated. Location of the craniomandibular joint is roughly at the same level as the occipital condyle, as seen in both the holotype and PKUP V1052 as well as in the congeneric species H. baitaigouensis (LPMC unnumbered specimen). Unfortunately, much of the posterior half of the skull remains obscured by damage caused by the splitting of the slabs in these specimens.

Mandible and dentition: The holotype and all other available specimens of the species have the mandibles preserved in articulation with the skull, making it difficult to discriminate the structural details of the jaw. The lower jaw is slenderly built, in accordance with the lightly built skull. The dentary and splenial can be identified, but the boundaries of the more posterior elements of the lower jaw are difficult to discern. The mandibular symphysis is limited to the anterior extremity (IVPP V11075). The retroarticular process is weakly developed but clearly defined. Monjurosuchidae and some neochoristoderes lack a definite retroarticular process, although other neochoristoderes possess a short, broad retroarticular process (Gao & Fox, 1998, 2005).

In the original publication, the marginal teeth of the species were described as likely ‘slender, simple, and needle-like’ (Gao et al. 1999) based on the exposed tips of the maxillary teeth. Further preparation of the holotype has shown that the upper teeth of this animal are generally peg-like. Posterior teeth are substantially smaller than those anterior ones, and the tooth crowns are weakly wrinkled with vertical ridges on the medial surface. All teeth are unicuspid, and broken tooth bases show no indication of developing plicidentine infoldings.

Postcranial skeleton

Vertebral column and ribs: The vertebral column in the holotype consists of a total of 35 presacral, three sacral, and more than 55 caudal vertebrae. All vertebrae have platycoelous centra, with essentially flat anterior and posterior articular surfaces.

As described in the original publication, the elongate neck of H. lingyuanensis consists of 19 cervical vertebrae (Gao et al. 1999). This number is confirmed by observation of another adult specimen (PKUP V1052; Fig. 2B) and the GMC juvenile specimen (Fig. 3) from the same locality and horizon. This number of the cervical vertebrae clearly distinguishes the type species from H. baitaigouensis, which possesses 24 cervical vertebrae. The holotype shows that the cervical vertebrae are narrow and elongate, with the 3rd through the 19th vertebrae varying between 10 and 12 mm in length. Most of the cervical centra in the holotype were damaged by the splitting from a horizontal plane, but the last two cervical vertebrae are well preserved and reveal a well-defined ventral crest. All but the first two cervical vertebrae bear ribs. These ribs are double-headed and become progressively longer and more spatulate posteriorly (Fig. 5).

Fig. 5.

Hyphalosaurus lingyuanensis (holotype IVPP V11075): photograph and line drawings of the pectoral girdle and fore limbs.

The dorsal series of the type species was described as having 16–17 vertebrae (Gao et al. 1999). Re-examination of the newly prepared holotype and other topotypic specimens (PKUP V1052, GMC juvenile specimen) shows a count of 16 dorsal vertebrae is correct. This number indicates that the trunk of the H. lingyuanensis is shorter than that of H. baitaigouensis by three vertebrae (see description below). The dorsal vertebrae are equal in length or slightly shorter than the cervical vertebrae. The first four anterior dorsal vertebrae are completely preserved and have a smooth ventral surface with no clearly defined crest. The next six dorsal vertebrae exhibit slight damage to the centrum, leaving the morphology of their ventral surface uncertain. The posteriormost six dorsal vertebrae are completely preserved, but are largely obscured by a thick sheath of gastralia. All dorsal vertebrae are preserved in articulation with their ribs, although the distal ends of the ribs have been displaced posteriorly by compression of the specimen. The dorsal ribs are unicapitate and strongly expanded distally. The high degree of pachyostosis seen in the ribs is similar to that in Neochoristodera.

The original description of H. lingyuanensis estimated the number of sacral vertebrae as either three or four. Comparison of the holotype with other specimens (PKUP V1052, GMC juvenile) confirms that all specimens possess three sacral vertebrae. This agrees with other choristoderes, though Lazarussuchus inexpectens was described as having four sacral vertebrae (Hecht, 1992). The sacral vertebrae are partially exposed in ventral view in the holotype, revealing that the robust sacral ribs are unfused to the vertebrae. The strongly expanded lateral ends of the sacral ribs do contact one another (Fig. 6).

Fig. 6.

Hyphalosaurus lingyuanensis (Holotype IVPP V11075): photograph and line drawings of the pelvic girdle and hind limbs.

As described by Gao et al. (1999), the holotype preserves 55 caudal vertebrae (PKUP V1052 shows a similar number). All but the first caudal have a shallow groove on the ventral surface of the centrum, and the groove is laterally flanked by well-developed longitudinal ridges (Fig. 6). The holotype also shows that at least the anterior caudal ribs are unfused to the centrum, though the rest seem tightly articulated if not fused. Starting from the 4th caudal, the centra become more or less hour-glass-shaped with a constricted middle portion and expanded anterior and posterior ends (Fig. 6). The caudal ribs are laterally directed and become progressively shorter toward the end of the tail.

Gastralia: As in other choristoderes, a mass of gastralia is developed ventrally at the abdominal region (Figs 2, 5). The gastral elements are arranged in longitudinal and transverse rows to form a sheet of bony protection between the pectoral and pelvic girdles. The extensive gastral sheet starts anteriorly at the level of the 7th dorsal vertebra, and posteriorly ends at a level below the 15th or the 16th dorsal vertebra. Individual elements are stout and spindle-like, with a cylindrical shaft and pointed ends. Each body segment is associated with two to three horizontal rows. Thus, there are more than 20 horizontal rows of the gastralia, each consisting of a central element in articulation with two lateral elements.

Pectoral girdle and forelimb: The pectoral girdle and the forelimbs are well preserved in the holotype (Fig. 5) and all the referred specimens. The interclavicle was mistakenly listed as unpreserved in the original publication (Gao et al. 1999). Additional preparation of the holotype revealed that the interclavicle was in fact embedded in the shale matrix. Sharply distinguished from that in the congeneric species H. baitaigouensis, the interclavicle of the type species is essentially T-shaped and has a poorly defined anteromedial process (Fig. 5). The anterior two-thirds of the stem is slightly widened giving it a sword-like shape, while the posterior one-third tapers to a point.

As shown on the left side in the holotype, the scapula has a short blade with a strongly expanded dorsal end. This expansion creates a well-developed neck at the lower part of the blade near the glenoid cavity. An acromion process is absent. The dorsal border of the scapular blade is convex dorsally, with a thickened rim for muscle attachment. The lower part of the scapula is a greatly widened plate, which has its curved ventral border in articulation with the coracoid plate. At the posterior border, the scapula contributes slightly over half of the glenoid cavity, with the remaining portion contributed by the coracoid. No supraglenoid foramen is identified on the scapula, and thus such a foramen is probably absent in Hyphalosaurus as in Monjurosuchidae (Gao & Fox, 2005). A cleithrum is absent, a derived condition within Diapsida (Evans, 1988). There is also no trace of ossified sternal plates.

The coracoid is a large suboval plate, with a rounded medial border. At the dorsomedial border, the coracoid is thickened and projects dorsally to form part of the glenoid cavity, but the actual shape of the cavity is hard to determine due to slight disarticulation on both sides of the specimen. Anteroventral to the glenoid cavity, a coracoid (supracoracoid) foramen penetrates the coracoid plate below the scapula-coracoid suture. Such a foramen is commonly seen in other diapsids and serves as the passage of the supracoracoid nerve and its associated blood vessels (Romer, 1956).

The clavicles are preserved in tight articulation with the scapulae and the interclavicle. The clavicles are only weakly angled at midpoint, giving the bone a boomerang-like shape. The dorsal half of the clavicle is attached to the anteromedial border of the scapula. The ventral half of the bone extends more or less horizontally and interlocks with a groove on the anterior border of the interclavicle crossbar. Medially, the two clavicles may contact one another at the ventral midline (Fig. 5).

The holotype includes all the forelimb elements, well-preserved and in articulation, except for the probable loss of a few carpals. The humerus has a short shaft and moderately expanded proximal and distal ends. The distal end is expanded and twisted at an approximately 45° angle in relation to the proximal expansion. The better-preserved right humerus shows that the proximal articular surface is well ossified, indicating the maturity of the individual at death. The proximal ventral surface is concave, with a well-defined subtriangular depression tapering toward the shaft. The anterior border of this depression marks the presumably well-developed deltopectoral crest, which is broken off on both humeri as a result of splitting the slab. The dorsal surface of the distal end is smoothly convex as preserved on the right element. The status of the ectepicondylar and entepicondylar foramina is indeterminate due to breakage.

The epipodial elements and the forefoot were largely unknown at the time of the original publication as they were embedded in the matrix. Subsequent preparation of the holotype thereafter shows that the epipodial elements are preserved crossing one another on both sides of the specimen. The radius is slightly longer than the more stoutly built ulna. The proximal end is expanded and slightly wider than the distal end in both elements. As in other choristoderes, no olecranon process or sigmoid notch is developed proximally on the ulna. Loss of the olecranon process and its associated sigmoid notch has been recognized as a derived morphology at the neodiapsid level (de Braga & Rieppel, 1997; Evans, 1988).

At least nine carpal elements can be identified in the left manus of the holotype (Fig. 5). The same number and pattern are also seen in PKUP V1052. The two large proximal elements are evidently the radiale and ulnare. A much smaller element distal to the radiale and ulnare can be identified as the intermedium. Two or three smaller elements positioned near metacarpals I–II are tentatively identified as the medial centrale and lateral centrale. There are four distal carpals. Distal carpal 5 is absent. No known member of the neodiapsid Choristodera has shown the presence of distal carpal 5, regardless of whether the absence is interpreted as the result of fusion with metacarpal IV or a simple loss of the element (see Romer, 1956). Of the five metacarpals, metacarpal I is the shortest but the most stoutly built, and metacarpal III is the longest, in keeping with the relative length of the digits. The forefoot has a phalangeal formula of 2–3–4–4–3 as shown on the left and right side of the holotype.

Pelvic girdle and hind limb: All three pelvic elements in the holotype are well ossified and preserved in articulation on the left side (Fig. 6), but on the right side these elements are slightly damaged by splitting and incomplete. On both sides of the specimen the ilium is exposed in lateral view, with weakly developed surface rugosities along the dorsal border of the iliac blade for attachment of axial muscles (Romer, 1956). The iliac blade lacks an anterior extension but has a short and deep posterior elongation beyond the acetabulum, a morphology that is primitive within Diapsida. The iliac blade is oriented horizontally. As consistently shown in both juvenile and adult specimens, this orientation of the blade is a taxonomically significant difference from the more vertical orientation in H. baitaigouensis(see below). Between the iliac blade and the acetabulum is a constricted neck. Constriction is variable among different taxa of choristoderes and also shows ontogenetic variation in Champsosaurus(Erickson, 1972). Ventrally, a large part of the base of the ilium is occupied by the acetabulum, the dorsal rim of which is formed by the weakly developed supra-acetabular buttress.

The puboischiadic plate is formed by a large pubis and a slightly smaller but more stoutly built ischium. The two elements share a tight sutural articulation, leaving no space for a thyroid fenestra. The pubis bears a well-developed pectineal tubercle, the thickness of which is proportionally comparable to the expanded condition in Champsosaurus (Romer, 1956). An obturator foramen perforates the thickened pubic plate near the distolateral margin as in other choristoderes. The ischium contributes a small part of the acetabulum, while the pubis forms only the anteroventral border of the pelvic cavity. A marked rugose angle projects from the posterior border of the ischium. This tubercle is similar to that seen in most neochoristoderes, but sharply different from the spike-like process in monjurosuchids (Gao & Fox, 2005; Gao & Li, 2007).

The femur in the holotype is nearly completely preserved on the left side, but is damaged by breakage on the right. The femur is roughly 136% of the length of the tibia in the holotype individual, the largest studied here (see Table 1). However, the proportions of the limb bones change markedly over ontogeny. In a juvenile individual (Fig. 3) the femur is relatively longer at 168% the length of the tibia, and in another specimen (PKUP V1052, slightly smaller than the holotype), the femur is 158% of the length of the tibia. The femur is basically a straight rod with weakly expanded proximal and distal ends. The expanded distal end is twisted at a 40–45° angle with respect to the expanded proximal end. As shown on the left element, the intertrochanteric fossa on the ventral aspect of the femur is greatly reduced to a shallow and short depression, extending just slightly more than one-fifth of the femoral length. There is neither any sign of the development of a fourth trochanter near the distal end of the intertrochanteric fossa, nor any evidence of the adductor ridge that is normally associated with the fourth trochanter.

Table 1.

Lengths of major elements from Hyphalosaurus lingyuanensis and Hyphalosaurus baitaigouensis in millimeters. Measurements for LPMC specimens were taken with dial calipers and rounded to the nearest 1 mm, other measurements were taken with digital calipers and rounded to the nearest 0.1 mm

	Hyphalosaurus lingyuanensis					Hyphalosaurus baitaigouensis
Specimen	PKUP V1052	IVPP V11075	PKUP V1057	PKUP V1058	BMNHC V050	PKUP V1056	LPMC R00052	BMNHC V053	BMNHC V060	LPMC R00066	LPM C R00065
Skull	54.1	57.8	13.6	16.1	27.0	33.0	35.0	36.0	33.0	66.0	65.0
Humerus	39.6	40.0	4.5	4.7	13.5	24.0	20.0	27.3	27.6	45.0	46.0
Ulna	25.5	26.0	3.1	3.5	9.0	15.8	13.0	15.3	17.9	–	28.0
Radius	25.5	27.0	2.9	3.1	9.5	15.9	13.0	15.3	19.3	–	28.0
Femur	50.5	43.0	4.9	5.1	16.5	27.6	21.0	29.6	31.6	52.0	52.0
Tibia	31.9	31.5	3.2	3.7	11.0	17.0	14.0	18.1	19.1	30.0	31.0
Fibula	30.5	31.0	3.0	3.3	10.0	16.4	13.0	17.5	18.9	28.0	29.0

The shafts of both the fibula and the tibia have a straight external border, but a concave internal border, creating a spindle-shaped gap between the articulated bones. The slenderly built fibula is distally expanded, in contrast to the proximally expanded tibia (Fig. 6). The distal end of the fibula articulates with both the astragalus (tibiale) and calcaneum (fibulare), while the tibia articulates only with the astragalus. At the proximal end of the tibia, a weakly developed cnemial crest can be identified extending from the head toward the middle portion of the shaft.

The tarsus of the holotype includes six ossified elements. Six tarsals have also been reported for adults of Monjurosuchus (Gao et al. 2000) and for simoedosaurids (Liu, 2004). The astragalus and calcaneum remain separate and form the proximal portion of the tarsus. Four smaller elements represent distal tarsals 1–4, and distal tarsal 5 is absent. The same number and pattern are also present in PKUP 1052. Metatarsals III and IV are subequal in length and are substantially longer than the remaining metatarsals. Metatarsal V is best preserved in the left hind foot, where it shows a strong proximal expansion but no development of plantar tubercles. The phalangeal formula for the hind foot is 2–3–4–4–3, as clearly shown on the right side (Fig. 6).

Integument

Impressions of soft tissue are preserved in two specimens of Hyphalosaurus examined in this study. PKUP V1052 (Fig. 7C–E) is a specimen of H. lingyuanensis. The other specimen, GMC V351 (Fig. 7A–B), cannot be identified to the species level because autapomorphy-bearing areas of the anatomy are obscured by artificial modification of the specimen. The skull and part of the neck are clearly part of a separate individual and were joined to the specimen to form a composite. This is particularly unfortunate because the removal of the anterior portion of the slab precludes identification of species by vertebral formula. Close examination confirms that the remaining portion of the skeleton, including the integument, is authentic.

Fig. 7.

Integument of Hyphalosaurus: A) Hyphalosaurus sp. (GMC V351) in dorsal view. Arrow indicates the area where cervical vertebrae and a skull from a separate individual have been artificially attached; box indicates area shown in (B). B) Close-up of integument impressions showing large, ovoid, keeled scutes surrounded by small irregular polygonal scales. C. Hyphalosaurus lingyuanensis (PKUP V1052) in ventral view; boxes indicate areas featured in (D) and (E). D) Skin impressions adjacent to dorsal ribs, including large keeled scutes (arrows) and small irregular polygonal scales. E) Skin impressions adjacent to caudal vertebrae, showing quadrilateral scales.

A parasagittal row of large, oval scutes topped with low midline keels runs along the flank of both specimens and is best preserved in GMC V351 (Fig. 7B). The row continues onto the base of the tail without significant decrease in size in PKUP V1052. These large keeled scutes are evenly spaced, with smaller, irregularly arranged, smooth polygonal scales occupying the area in between them. Smooth polygonal scales overlap the last preserved large keeled scutes, indicating the patch of integument bearing smaller keeled scutes has been folded over this part of the flank integument. A single small keeled oval scute with an area about one-fourth that of the others is preserved anterior of the femur in GMC V351. This keeled scute appears to represent another parasagittal row located either ventral or dorsal to the flank row. The small scales of the tail are quadrilateral and are more regularly arranged than those of the flank. In PKUP V1052, the tail is twisted at a level between the14th and 15th caudal vertebra, so that a lateral view of the tail and integument is exposed posterior to this point. Integument impressions extend for a significant distance from both the dorsal and ventral margins of the caudal vertebrae, suggesting the height of the tail may have been increased by soft-tissue projections. The polygonal scales surrounding the hindlimb skeleton are irregular and smaller than the polygonal scales separating the keeled scutes.

Integumentary impressions have previously been reported in Champsosaurus(Brown, 1905; Erickson, 1985) and Monjurosuchus splendens (Gao et al. 2000). The integument of Champsosaurus gigas is made up of fine scales. Scales of the upper back are small and rhomboid with sharp crests, scales on the lateral surface of the tail are low, smooth and pustulate, and scales on the lower flank are large and circular (Erickson, 1985). There is no evidence of large keeled scutes in Champsosaurus, but patches of integument are currently known from only a few regions of the body. The integument of Monjurosuchus consists largely of small overlapping scales, which are largest on the dorsal surface of the body, at least at the hind limb (Gao et al. 2000). Monjurosuchus also possesses large ovoid, keeled scutes, arranged in parasagittal rows and surrounded with smaller papilloid scales. The striking similarity between these scales in Monjurosuchus and Hyphalosaurus provides additional evidence that the apomorphic hyphalosaurids belong within Choristodera. Integument impressions indicate that Monjurosuchus possessed webbed feet (Gao et al. 2000). The integument is incompletely preserved around the manus and pes in both specimens of Hyphalosaurus, but a patch of scales visible between the proximal phalanges of the right pes in PKUP V1052 suggests some degree of webbing in Hyphalosaurus.

New observation and taxonomic revision on Hyphalosaurus baitaigouensis

Species Hyphalosaurus baitaigouensisJi et al. 2004b

Holotype: CAGS-IG-03-7-02, incomplete skull and partial postcranial skeleton (excluding the 11 leathery-shelled eggs from the original designation; see remarks below).

Type locality and horizon: Baitaigou village near Toutai, Yixian County, Jinzhou City of Liaoning Province; Early Cretaceous Yixian Formation (see remarks below).

Referred specimens: PKUP V1056-1058, BMNHC V014- 053, LPMC R-00052, LPMC R-00065-00066, and many unnumbered specimens in BMNHC, IVPP, LPMC and PKUP collections.

Diagnosis (revised from Ji et al. 2004b): Hyphalosaurid choristodere differing from the congeneric species H. lingyuanensis in having the frontal/parietal suture located at the midlevel between the orbit and the supratemporal fenestra; fused postorbitofrontal medially having extensive sutural contact with both the frontal and the parietal; possessing greater elongation of the neck, consisting of 24 cervical vertebrae; 19 dorsal vertebrae; more than 60 caudal vertebrae; deep and horizontal groove developed on the lateral surface of the dorsal vertebral centra; interclavicle rhomboid in shape, lacking a well-developed cross bar; iliac blade vertically oriented.

Remarks: The original designation of the holotype is problematic, as the holotype skeleton and the eggs outside the skeleton must be considered to be different individuals. Following the ICZN rules (ICZN, 1999: 73.1), the 11 leathery-shelled eggs associated with the holotype skeleton must be excluded from the holotype designation; therefore, the original type designation should be corrected to include only the incomplete skull and its associated partial postcranial skeleton as a single specimen.

The original description of the holotype is also problematic. The holotype skull is badly damaged and as illustrated in Ji et al. (2004b: Fig. 1) it contains at least five questionable structure identifications. Moreover, many key features (e.g. external nares) were not shown in the poorly preserved skull. A thorough description is needed to clarify the cranial morphology of the species (see below). Based on our observation of a large number of well-preserved specimens, we have extensively revised the diagnosis of the species as shown above. The single diagnostic feature mentioned in the original publication was the total of 26 cervical vertebrae (see Ji et al. 2004b). However, as described below, all the fully articulated specimens that we have examined have 24 cervical vertebrae. The miscount by the original authors was probably caused by including the two anteriormost dorsal vertebrae as cervical vertebrae. As shown in the original publication (Ji et al. 2004b: plate 1), the cervical series was distorted and the incompletely preserved pectoral girdle and right arm have been slightly displaced. Because of distortion in the holotype, the original authors miscounted the dorsal vertebrae as 17 in number, while multiple fully articulated specimens that we have observed show 19 dorsal vertebrae as described and figured below.

In the original publication, the type horizon was described as the upper part of the Chiufotang Formation (Ji et al. 2004b). Detailed stratigraphic study, however, has shown that the fossil horizon is probably part of the Yixian Formation (Wang et al. 2004). The latter stratigraphic designation is followed in this paper.

Osteology of H. baitaigouensis

The description below is based primarily on the study of several fully prepared specimens (BMNHC V014, BMNHC V050, PKUP V1056, LPMC R-00065, LPMC R-00066). These specimens are presumed to represent a range from medium-sized subadults to large fully-grown adults, based on the range of sizes (see Table 1) and degree of limb ossification observed in our study. Among these, LPMC R-00065 and R-00066 (Fig. 8) are the largest specimens known for this species. The former has a skull length of 66 mm (entire body length of > 1050 mm), and the latter specimen of 65 mm (entire body length of 1100 mm). We examined approximately 50 additional specimens of H. baitaigouensis in the BMNHC, GMC, LPMC and PKUP collections, though most of these were of varying completeness, quality of preservation and stages of preparation. Study of these specimens reveals previously unknown information regarding several cranial elements (premaxilla, frontal, and quadratojugal), and allows us to revise the description of several other elements (postorbital, postfrontal, lacrimal, and squamosal) to clarify the osteology of the species.

Fig. 8.

New specimens of Hyphalosaurus baitaigouensis: A) LPMC R-00065; B) LPMC R-00066; both nearly complete skeletons from the upper Yixian Formation exposed near Yizhou.

Cranial skeleton

The skull is small proportional to body size, with paired and elongate nares, large orbits and posteriorly flared supratemporal fenestrae (Figs 8–11). The infratemporal fenestrae are fully closed in adults but remain open in small juveniles (pers. comm. from Hou Lianhai on LPMC specimens). The shape of the skull also shows minor ontogenetic variations: the skull of small to medium sized individuals is generally lancelet shaped, with a pointed snout and strong posterior flaring of the supratemporal fenestrae, whereas large individuals tend to have a blunter snout as shown in LPMC R-00065. The skull appears to be strongly flattened, a feature seen in other choristoderes. However, all fossils examined in this study have undergone post-mortem compression in variable degrees and the true degree of flattening may have been somewhat less in life.

Fig. 11.

Skull roof and palate of Hyphalosaurus baitaigouensis: A) skull roof in BMNHC V014; B) skull roof in BMNHC V050; C) palate in an unnumbered specimen in the LPMC collections.

Fig. 9.

New specimen of Hyphalosaurus baitaigouensis: BMNHC V050; a nearly complete skeleton from the upper Yixian Formation exposed near Yizhou.

Skull roof

The premaxillae are unknown from the holotype (Ji et al. 2004b: Fig. 1), but are well preserved in several specimens used in this study. As best shown in the fully articulated skull of LPMC R-00065, the premaxilla has a blunt tooth-bearing base that carries at least four teeth. Four teeth are also observed on the right premaxilla of BMNHC V050. The dorsal process is slender and long, terminating at a point slightly posterior to the mid-level of the external narial openings. The narial openings are clearly paired, elongate and dorsally placed, as best shown in LPMC R-00065. The anterior surface of the premaxilla is penetrated by several small and rounded foramina, as seen in other choristoderes, and the posterolateral end of the premaxilla has a small notch for articulation with the maxilla (best shown in BMNHC V050 and LPMC R-00065).

The maxilla, as shown on both sides in BMNHC V014, has a small anterior process that inserts into the small notch on the premaxilla. The dorsal rim of this anterior process is slightly notched for the lateral border of the external narial openings. The middle part of the maxilla lacks a dorsal process, but has a relatively straight sutural contact with the nasal. At the level of the anterior border of the orbit, the maxilla is dorsolaterally notched, accepting the anterior tongue of the jugal, which forms the entire lateral rim of the orbit (best shown on the right side in BMNHC V014; Fig. 11A). The posterior extent of the maxilla is unknown from this specimen, but can be inferred as reaching the posterior border of the orbit, where the marginal tooth row ends in the lower jaw.

The nasal was identified on the right side of the holotype skull but labeled with a question mark (Ji et al. 2004b: Fig. 1). Several well-preserved specimens used in this study show that the nasals are paired and elongated. The nasals have a short midline suture anteriorly, but diverge posterolaterally between the maxilla and the prefrontal. Each nasal is anteriorly notched for the posterior border of the narial opening, and each has a small process abutting the dorsal process of the premaxilla. The posterior extension of the nasal has a small process that wedges between the prefrontal and the lacrimal (Fig. 10). The limited midline contact between the nasals contrasts with the more extensive contact or fusion seen in other choristoderes. This feature may be taxonomically significant at the species level or a higher level, pending information on the nasals in other hyphalosaurids.

Fig. 10.

Skull roof of Hyphalosaurus baitaigouensis: photograph and line drawing of LPMC R-00065.

The prefrontals are paired and share a long medial sutural contact. The prefrontals are widest anterior to the orbit between the lacrimals, tapering both anteriorly towards the contact with the nasals and posteriorly towards the contact with the frontals. The anterior half of the prefrontal shares an oblique suture with the nasal laterally. The anterior process extends to the level close to the posterior border of the nares. Posteriorly, the prefrontal/frontal articulation is roughly at the midlevel of the orbit and is formed by the small posterior process of the prefrontal overlapping the small notch of the frontal (Figs 10, 11A,B).

The status of the lacrimal was uncertain in the original publication, where a long strip-like structure was questionably identified as the lacrimal. On the left side of BMNHC V050, there appears to be a small triangular bone forming the anterior rim of the orbit. This bone articulates medially with the prefrontal and laterally with the maxilla. In BMNHC V014, the sutural patterns of the nasal with its surrounding elements are unclear, but large specimens in the LPMC collections (LPMC R-00065, R-00066) preserve a modest-sized, triangular lacrimal, and display similar sutural patterns as in BMNHC V050. In LPMC R-00065, the lacrimal foramen is visible.

The frontals are unknown from the holotype, but a large number of well-preserved specimens used in this study show that they are paired, slender and elongate. The frontals narrow anteriorly to form an interorbital beam that articulates with the prefrontals and posteriorly become slightly widened and they share an oblique suture with the postorbitofrontals. The frontals form the larger half of the interorbital beam but barely contribute to the posterior rim of the orbit. The interdigitating frontal/parietal suture is roughly transverse and is located at the midlevel between the orbit and the supratemporal fenestra, a diagnostic feature of the species (best shown in BMNHC V014, V050, LPMC R-00065).

The postorbital and postfrontal were misidentified as separate elements and were illustrated as disarticulated and far apart in the holotype specimen (Ji et al. 2004b: Fig. 1). However, all well-preserved specimens used in this study clearly show that the two bones are fused into a single postorbitofrontal (Figs 10, 11). In one relatively small specimen (BMNHC V014) there are differences in the coloration of the two bones. On both sides, the supposed area of the postfrontal (the more medial element) is slightly darker than the postorbital part, although no suture can be identified. A similarly sized individual (BMNHC V050) and a large specimen (LPMC R-00065), however, show no evidence of a suture or other sign of differentiation between the two parts of the postorbitofrontal. The postorbitofrontal has a slightly notched anterior margin forming the entire posterior rim of the orbit. Medially, the postorbitofrontal has a straight, posterolaterally oriented sutural contact with both the frontal and the parietal (different from H. lingyuanensis; see above). Laterally, the postorbitofrontal shares a straight sutural contact with the deepened posterior portion of the jugal. The posterior border of the fused postorbitofrontal is notched to form the entire anterior rim of the supratemporal fenestra and a well-developed posterolateral process attaches to the medial side of the squamosal, forming the anterior part of the dorsal rim of the supratemporal fenestra as well.

The jugal has a slender and blade-like anterior process that forms the ventral rim of the orbit. The tongue-like anterior end of the process articulates with the lacrimal and maxilla (best shown on the right side in BMNHC V014; Fig. 11A). The anterior process of the jugal is well preserved on the left side in BMNHC V050, but is slightly disarticulated from the maxilla as a preservational artifact. On the right side in BMNHC V050, the process is still in articulation with the maxilla, though the tip of the process is broken. As shown on both specimens, the jugal lacks a dorsal process at the posterior border of the orbit; instead, the dorsal border of the jugal has a roughly straight sutural contact with the postorbitofrontal (Fig. 11A,B). Posterior to the orbit, the jugal is deepened and is notched for articulation with the quadratojugal.

The quadratojugal was not identified in the holotype. This bone is incompletely exposed on the right side of BMNHC V050, and is often obscured by dorsal elements in other specimens. The posterior extent and the role of this element in the closing of the infratemporal fenestra in adult individuals are unclear, but its anterior articulation with the jugal is clearly by a process fitting in the notch of the jugal as shown on the right side of BMNHC V014 and the left side of BMNHC V050 (Fig. 11A,B). In the two largest specimens (LPMC R-00065, R-00066), the quadratojugal is either concealed by other bones or damaged by breakage, thus providing no reliable information on the morphology of this skull element.

The parietals are paired elements that form the entire medial border of the supratemporal fenestra. Anteriorly, the parietals form a roughly trapezoidal table that shares a transverse suture with the frontals anteriorly and a slanting suture with the postorbitofrontal laterally. The pineal foramen is clearly absent in all observed specimens, as in other choristoderes. The posttemporal processes are slender and long, diverging posterolaterally to the posterior extremity of the skull, where the parietal meets the medial extension of the squamosal. Along the medial border of the supratemporal fenestra, the parietal develops a downward flange to meet with several other elements and complete the lateral wall of the braincase (see below).

The squamosal was incorrectly labeled in the holotype (see Ji et al. 2004b: Fig. 1), as the structure labeled ‘squamosal’ is obviously part of the postorbitofrontal. In fact, the squamosal is an elongate element as best shown in BMNHC V050. In dorsal view, the squamosal has the anterior process in contact with the posterior extension of the postorbitofrontal, and thus it forms a large part of the lateral border of the supratemporal fenestra. This anterior process may ventrally contact both the jugal and quadratojugal, but the actual sutural pattern between these elements is unclear. The posterior part of the squamosal develops a roughly triangular medial wing that meets the posttemporal process of the parietal. Within the supratemporal fossa, the slightly curved anterior border of the squamosal wing sutures with the quadrate.

Palate and braincase: The vast majority of H. baitaigouensis specimens are exposed in dorsal view, and thus the palate remains incompletely known for this species. Most of our information on the palate of the species is derived from an unnumbered specimen in the LPMC collection (Fig. 11C), the only fully prepared specimen exposed in ventral view.

The small, subcircular internal choanae show a moderate degree of posterior displacement, opening at approximately one-third of the length from the anterior to the posterior end of the marginal tooth row. A nasopalatal trough runs posteriorly from the choana in Neochoristodera and is also present as a shallow groove in Monjurosuchidae. However, the presence or absence of this structure cannot be confirmed in H. baitaigouensis. The suborbital fenestra is visible on both sides of the skull and is subtriangular in shape.

Due to preservation and overlying portions of the mandible, details of the anterior tip of the snout are unclear and it is not possible to identify the sutures between the vomer, palatine and pterygoid with complete confidence. Regardless, it is clear that the vomers share a long midline contact as in H. lingyuanensis and bear at least one row of palatal teeth. The exact relationship of the vomer to the maxilla is less certain. It appears these bones contact each other as in other choristoderes, but the extent of the contact is uncertain. No clear suture between the pterygoid and palatine is identifiable in this specimen, but we presume that the region medial to the suborbital fenestra includes the contact between these two elements. If this is correct, the palatal tooth row running between the suborbital fenestra and the internal choana is part of the palatine.

There is a significant gap between the pterygoids (Fig. 11C), unlike the condition in Monjurosuchidae and Neochoristodera, in which the pterygoids share a long midline suture. Two rows of small palatal teeth stud each pterygoid, beginning at the suture with the parasphenoid and continuing to the level of the suborbital fenestra. Each row appears to be a single tooth position wide. Whether the transverse flange of the pterygoid is also toothed cannot be determined with certainty. As in other choristoderes, the palate is akinetic, with the pterygoid and parasphenoid sharing a tight sutural contact. The interpterygoid vacuity is enclosed by the parasphenoid rostrum posteriorly and the pterygoids anteriorly. In size, the interpterygoid vacuity is proportionally larger than that of Champsosaurus or Simoedosaurus, but smaller than that of Philydrosaurus or Tchoiria. The interpterygoid vacuity does not extend as far as the posterior edge of the suborbital fenestra.

The ectopterygoid is exposed in dorsal view through the orbit in several specimens (BMNHC V014, V050). The bone is triradiate with short anterior and posterior processes and a well-developed pterygoid process. The pterygoid process is slightly constricted at its lateral end, as in Cteniogenys(Evans, 1990).

The ventral surface of the braincase is well exposed in the unnumbered specimen from the LPMC collections (Fig. 11C). The basisphenoid is unexpanded posteriorly and the parasphenoid rostrum is short and broad. This gives the ventral surface of the braincase a flask-like shape, with the basisphenoid forming the base and the parasphenoid forming the thick neck. Projection of the basal tubera is very weak. The craniomandibular joint is located at roughly the same level with the occipital condyle as in the type species (see above). Fine details of the occipital condyle itself are not observable.

The lateral wall of the braincase is best observed on the right side of BMNHC V050. The parietal table projects a ventral flange. The squamosal sends a triangular medial process to contact the posttemporal process of the parietal. The posterior end of this triangular plate forms the posterior extremity of the skull and the anterior border of this triangular process sutures with the quadrate. Anterior to the quadrate, an inverted triangular bone forms part of the lateral wall of the braincase. This bone may represent the neomorph present in Neochoristodera, but better specimens are desirable for confirmation of this important feature. Anterior to the inverted triangular bone is a robust ventral extension from the parietal table. The left side of the skull in BMNHC V050 shows that a major portion of the lateral braincase wall has been crushed into small pieces within the supratemporal fossa, though the triangular plate-like medial extension of the squamosal is well preserved.

In BMNHC V014, the braincase is well exposed on both sides of the skull. Examination of the squamosal confirms observations made in BMNHC V050. The quadrate shows a well-defined anterolateral extension (quadratojugal process), which approaches the anterior interior border of the supratemporal fossa. The anterior margin of the quadrate shares sutures with the element identified as the possible neomorph and the ventral flange of the parietal (Fig. 11A). At this suture a small foramen opens. If the tentative identification of the neomorph proves correct, this opening is the quadratojugal foramen.

Axial skeleton

All the specimens used in this study consistently show that the cervical vertebrae are 24 in number, contrary to the counting of 26 in the original description by Ji et al. (2004b). Large specimens (LPMC R-00065, R-00066) also display a sharp distinction between the cervical and dorsal series: in the cervical vertebrae the neural spine is a low and simple crest, but starting from the first dorsal vertebra the neural spine becomes expanded with a rugose dorsal surface (Fig. 12). The centrum of each cervical is slightly longer than tall. BMNHC V050 shows breakage of the 21st cervical vertebra, but the number of cervical vertebrae has not been artificially altered in any manner. The 17th through the 24th cervicals of this specimen show slight shifting of the neural arches towards the right, but the arches are still partly in articulation with the centra. This slight disarticulation indicates that the neural arches and centra are unfused as typical in choristoderes.

Fig. 12.

Vertebral column of Hyphalosaurus baitaigouensis (LPMC R-00065). Note the sharp distinction of the neural spines between the cervical and dorsal series.

The dorsal vertebrae are well exposed in BMNHC V050 and many other specimens. There are 19 dorsal vertebrae, along with the cervicals providing a total of 43 presacral vertebrae for H. baitaigouensis. The dorsal vertebrae are shortened anterior-posteriorly and are slightly taller than long. In many of the dorsal vertebrae in BMNHC V050, the neural arch is disarticulated from the centrum, leaving the articular surface of the centrum exposed in dorsal view. The centrum is platycoelous, typical for choristoderes. The centrum of the 13th dorsal vertebra is exposed in lateral view on the left side, and the lateral surface of the centrum has a deep and horizontal groove, also clearly visible in PKUP V1056 (Fig. 13), collected from the same formation and horizon as the holotype of the species. This feature is absent in H. lingyuanensis and is recognized in this study as diagnostic for H. baitaigouensis.

Fig. 13.

Vertebral column of Hyphalosaurus baitaigouensis (PKUP V1056). Arrows pointing to deep horizontal grooves developed in the lateral surface of the centrum in the dorsal series.

In marked contrast to the cervical series, the neural spines of the dorsal vertebrae have a laterally expanded and rugose dorsal surface (best shown in LPMC R-00065, R-00066). Erickson (1987) hypothesized that the rugose neural spine apices of Simoedosaurus dakotensis provided attachment surfaces for a sagittal row of large scutes. We posit a similar function for the dorsal neural spines of H. baitaigouensis. The distinct expansion of the neural spine ends at the third sacral vertebrae, and the following caudal vertebral neural spines have a markedly different tall and triangular spike-like shape. The dorsal ribs are greatly expanded and terminate abruptly at their distal ends.

Three sacral vertebrae are identified in BMNHC V050 and other specimens used in this study. In all three sacral vertebrae the neural arch is disarticulated from the centrum and the neural spine is strongly expanded, as in the dorsal vertebrae. As exposed on the left side of the specimen, the sacral ribs are also unfused to the centrum, and have been slightly disarticulated from their associated sacral vertebra. The first sacral rib is the longest, the second the most robust, and the third is shortest.

The tail is greatly elongated, accounting for more than half of overall body length. There are more than 60 caudal vertebrae preserved in BMNHC V050. This caudal count is consistent with other well-preserved and prepared specimens we observed, including LPMC R-00066. The anterior and middle caudal vertebrae are slender, elongate and laterally compressed, bearing a tall and spike-like neural spine. The neural arch of the first caudal is disarticulated from the centrum, but in all posterior caudal vertebrae the neural arch is fused to the centrum. The spike-like neural spines remain well developed throughout almost the entire caudal series, but the last few vertebrae are reduced to tiny cylinders lacking recognizable spines. The tips of the transverse processes of the first few vertebrae are posteriorly deflected, but for the rest of the caudal series the transverse processes are oriented perpendicular to the centrum.

Appendicular skeleton

The interclavicle is exposed in dorsal view on the left side of the vertebral column in BMNHC V050 (Fig. 14). Although partly obscured by the scapula blade, clavicle and several ribs, the base of the interclavicle is clearly rhomboid in shape with a well-developed triangular anterior process. However, the interclavicle lacks a well-developed crossbar. This configuration of the interclavicle is different from the T-shaped condition in the type species H. lingyuanensis (Gao et al. 1999) and is another diagnostic feature of the species recognized in this study. The interclavicle has a long and slender stem, which is partly obscured by several ribs and the scapula. The left clavicle is a curved bone, and is disarticulated from but closely associated with the interclavicle. The right clavicle is also partly exposed in close association with the right scapula. The left scapula is well exposed in association with the interclavicle and the coracoid. The scapular blade is narrow and tapering. The coracoid is a roughly rounded plate. The coracoid foramen is located close to the anterior border of the coracoid plate as shown in several other specimens of the same species.

Fig. 14.

Photograph and line drawings of BMNHC V050, showing the pectoral girdle and the forelimbs of Hyphalosaurus baitaigouensis. Note the interclavicle lacks a crossbar.

The left humerus is completely preserved in BMNHC V050. The humerus is slightly sigmoid with widened proximal and distal ends, but the shaft is greatly constricted. The left humerus is exposed in dorsal view. A small groove is observable close to the anterior border of the distal part of the bone, though an ectepicondylar foramen cannot be identified. In BMNHC V014 the well-preserved left humerus clearly preserves a well-defined ectepicondylar foramen. The right humerus of BMNHC V050 is completely preserved, but the distal part of the forelimb has been added artificially from a different specimen. The distal part of the left forelimb is authentic, and the ulna is slightly shorter than the radius. The radius exhibits slight curvature. The proximal end of the ulna is thicker than the distal end, but no olecranon process is present. Based on comparisons of specimens from a variety of size classes, there is a clear ontogenetic trend in reduction of the epipodial length relative to propodial length as growth progresses. In BMNHC V050, a smaller individual, the humerus is approximately 150% the length of the ulna, whereas in the large individual LMPC R-00066 the humerus is about 170% the length of the ulna.

At the distal end of the ulna in BMNHC V050, three ossified mesopodial elements are identified. The ossification of only three carpal elements indicates subadult status, as larger individuals show several additional elements. A slightly larger specimen (PKUP V1056) shows better ossification in this part of the limb and includes seven carpal elements. The ulna contacts three elements, which we identify as the pisiform, ulnare and intermedium based on position. The pisiform is suboval and positioned postaxially at the ulna/ulnare contact. The ulnare is a large square bone, capping the distal end of the ulna. As preserved, the ulnare directly contacts metacarpal V distally without intervention of a distal carpal. The intermedium is an elongate element that contacts the ulna proximally and lies against the preaxial surface of the ulnare. The centrale and distal carpals 2, 3 and 4 are also ossified. Distal carpal 1 and 5 were apparently unossified or absent, at least at the ontogenetic stage represented by PKUP V1056. Distal carpal 5 is present in Monjurosuchus(Gao et al. 2000), but has not been reported in neochoristoderes. A pisiform has been identified in Monjurosuchus, but is not present in an articulated forelimb of the neochoristodere Ikechosaurus pijiagouensis(Liu, 2004). The third and the fourth metacarpals of H. baitaigouensis are equal in length and are the longest in comparison with other elements. The first metacarpal is the shortest and most robust. The phalangeal formula of the manus is 2–3–4–4–3 based on multiple specimens.

The ilium is best exposed in medial view on the right side of BMNHC V050 (Fig. 15). The iliac blade is roughly rectangular with a blunt dorsal end, and meets the acetabular base at an angle greater than 45°. The iliac neck is poorly defined. The medial surface of the ilium is smooth and slightly concave. The left ilium is well-exposed in lateral view in the same specimen. This view indicates the lower part of the ilium is expanded and concave, forming a large part of the acetabulum. The blade is slightly ridged for attachment of pelvic muscles. The blade is nearly vertically oriented in examined specimens, differing from the horizontal orientation in the type species H. lingyuanensis. The pubis has a concave anterior border, and shares a sutural contact with the ischium posteriorly. The ischium is a roughly triangular plate, with a well-developed tubercle projecting from the posterior border. This tubercle is similar to that seen in most neochoristoderes, but is sharply different from the spike-like process in monjurosuchids (Gao & Fox, 2005; Gao & Li, 2007).

Fig. 15.

Photograph and line drawing of BMNHC V050, showing pelvis and hind limbs of Hyphalosaurus baitaigouensis.

The hind limbs are well preserved on both sides in BMNHC V050. The femur is essentially straight, lacking significant sigmoid curvature. A large specimen (LMPC R-00065), however, shows some sigmoid curvature of the femur. This feature is also ontogenetically variable in M. splendens (Gao et al. 2000). The tibia is slightly stouter and slightly longer than the fibula. The fibula is a simple straight bone with an expanded distal end. The femur is roughly 150% of the length of the tibia in BMNHC V050, but large specimens (LPMC R-00065, R-00066) display different proportions, with the femur being approximately 165–170% of the length of the tibia. The hind foot is completely preserved on both sides in BMNHC V050, but a slightly larger specimen (PKUP V1056) is more completely ossified.

In the articulated tarsus of PKUP V1056, at least five discrete elements can be identified. The astragalus is the largest tarsal element and takes the shape of an irregular hexagon in dorsal view. Proximally, this bone bears two articular surfaces for the tibia and fibula. A non-articulatory surface is directed medially. Two distal surfaces serve as articulations with a moderate-sized element we identify as distal tarsal 4 and with the first metatarsal. The remaining lateral surface is for the articulation with the calcaneum. The calcaneum is the second largest element, and articulates with the fibula but not the tibia. It possesses a strong lateral projection and appears to be largely or completely separated from the 5th metatarsal. The distal surfaces of the astragalus and calcaneum together form a smooth concave articulation for distal tarsal 4. These three tarsal elements are tightly articulated in PKUP V1056. Distal tarsal 4 clearly articulates with the fourth and fifth metatarsals. Adjacent to distal tarsal 4, a square element, presumably distal tarsal 3, articulates with the 3rd metacarpal. Distal tarsal 3 is clearly exposed on the left side of this specimen, but obscured on the right. At least one small tarsal element is present medial to distal tarsal 3. The possibility that larger individuals had additional ossified tarsals remains, but in the largest specimens examined in this study the tarsus is not as well preserved as in PKUP V1056.

The pes is longer and more asymmetrical than the manus. Metatarsals III and IV are subequal in length as in the type species, and these significantly exceed metatarsals I, II and V in length. Digit IV is the longest, followed by the digit III, II, V and I. The claws are short and weakly curved, with poorly developed plantar tubercles. The digital formula of the pes is 2–3–4–4–3, as in other choristoderes.

Paleoecology of the family Hyphalosauridae

The Hyphalosauridae exhibit extensive skeletal modifications that suggest they were the most exclusively aquatic choristoderes. One such modification is the great elongation of the neck. The number of cervical vertebrae in hyphalosaurids varies from at least 16 to 24, and the number of dorsal vertebrae varies from 16 to 19. Amongst other clades of choristoderes, the presacral vertebral series includes eight cervical vertebrae and 16 dorsal vertebrae in the Monjurosuchidae (Gao et al. 2000; Gao & Fox, 2005; Gao & Li, 2007), eight cervical vertebrae and 16 dorsal vertebrae in the neochoristodere Simoedosaurus lemoinei (Sigogneau-Russell, 1981), and nine cervical vertebrae and 17 dorsal vertebrae in the neochoristodere Champsosaurus (Champsosaurus ambulator: AMNH 981; C. gigas, Erickson, 1972). Thus, the elongation of the neck of Hyphalosaurus appears to have been achieved through addition of cervical vertebrae rather than incorporation of dorsal vertebrae into the cervical series. The addition of dorsal vertebrae has also occurred in H. baitaigouensis, with three additional vertebrae added to the reconstructed plesiomorphic count of 16.

The expanded pachyostotic ribs and gastralia of Hyphalosaurus would provide negative buoyancy and also restrict the flexion of the trunk. The appendicular skeleton is relatively unspecialized, but the tail is deepened throughout most of its length by the high neural spines. Based on these features, the most likely means of underwater propulsion was undulation of the tail, with little involvement of the limbs. The well-ossified carpus and tarsus suggest that hyphalosaurids retained some capacity for terrestrial locomotion, though the great elongation of the neck and tail relative to the rest of the body would have made movement on land cumbersome. Based on anatomical details (including integument pattern) from a large number of specimens, an accurate life reconstruction of the animal can be presented (Fig. 16).

Fig. 16.

Life reconstruction of Hyphalosaurus lingyuanensis. Artwork by Kristin Lamm.

The super-abundance of hyphalosaurids throughout the Yixian Formation indicates that they played a particularly important role in the aquatic food chain of the Early Cretaceous ecosystem in the area. Our understanding of the jaw musculature of Hyphalosaurus remains poor due to the two-dimensional preservation in available specimens. However, the simple peg-like dentition, small skull size, and greatly elongated neck suggest pursuit of small prey items. As in other choristoderes, both species of Hyphalosaurus have a flattened skull. A flattened skull has been associated with prey capture through a sideways strike in aquatic reptiles (Taylor, 1987). Peg-like teeth in both upper and lower jaws seem to be designed for grasping prey with a quick sweeping bite, as seen in the extant fish Lepisosteus. The small and pointed palatal teeth are not suited for crushing, but may have been employed in holding small prey. The great range of neck orientation in articulated skeletons includes animals with the head oriented at a 180° angle relative to the body, suggesting the neck was highly flexible. All indications suggest these individuals were buried in situ by ash fall in moderate to deep water, indicating these poses were attainable in life and not the result of rigor mortis or desiccation.

The reduction of the cervical neural spine height may have served to reduce drag during lateral sweeping of the head. Rieppel (2002) suggested suction feeding may have been employed by some pachypleurosaurs, but the Hyphalosaurus specimens at hand do not reveal details of the hyoid apparatus necessary to evaluate this possible feeding strategy. Regardless of whether a sweeping bite or suction was used to capture prey, Hyphalosaurus appears to have been an active hunter, in contrast with the sit and wait ecology proposed for long-snouted neochoristoderes (Erickson, 1972). Small fish and arthropods are both abundant in Yixian Formation fossil beds and seem to be the most appropriately sized prey items. Interestingly, the holotype skeleton of H. lingyuanensis is preserved in close association with six individuals of Lycoptera and many other specimens of Hyphalosaurus are also associated with this small fish. These associations provide only circumstantial evidence for a predator–prey relationship, but Hyphalosaurus likely consumed Lycoptera fish at least on occasion. Gut contents are preserved only in one specimen, LPMC R-00066. Small, disarticulated ribs are visible enclosed between the ribs and gastralia of this individual. This indicates Hyphalosaurus consumed vertebrate prey at least occasionally. It is uncertain whether the dearth of gut contents in other specimens reflects a diet composed mainly of soft-bodied prey or is due to gut contents being lost and/or remaining uncovered during preparation.