Osteology of the two-fingered oviraptorid Oksoko avarsan (Theropoda: Oviraptorosauria)

Abstract

Oviraptorosaurs are among the most diverse and best-known extinct theropod groups. Their bizarre anatomy and their social and reproductive behaviour are now well understood. Among the oviraptorid subclade, the two-fingered Oksoko avarsan is particularly well-represented. It is known from several exquisite skeletons, preserving not only the entire skeleton, but multiple stages through ontogeny, providing an exemplar for understanding the anatomy of oviraptorids and the changes that their skeletons experienced over their lifetimes. Here I comprehensively describe the osteology of Oksoko avarsan and comment on its ontogenetic variation. Excellent preservation of the specimens provides unparalleled detail into the anatomy of an oviraptorid, providing valuable context for interpreting other oviraptorosaurs. Several changes are observed to occur through ontogeny relating to robustness of the bones and proportions of the skeleton, but there is little evidence suggesting that discrete features like the cranial crest arose late in ontogeny. Instead, early development of the cranial crest in oviraptorids, and its internal connection with the nasal passages and other pneumatic spaces, argue in favour of a role in vocalization, perhaps alongside sexual display. Detailed anatomical and ontogenetic data, like those provided by known specimens of Oksoko avarsan, are necessary to help to underpin ongoing research into the palaeobiology and macroevolution of Oviraptorosauria.

Introduction

Oviraptorosaurs are among the most bizarre but well-known extinct theropod dinosaurs. The first oviraptorosaurs were named in 1924: Chirostenotes pergracilisGilmore 1924 by Gilmore (1924), followed by Oviraptor philoceratopsOsborn 1924, described by Osborn (1924), from Late Cretaceous beds in North America and Asia, respectively. Although Osborn (1924) commented on the similarities between the forelimbs of Oviraptor and Chirostenotes, only much later were they united in a clade, by Osmólska (1976). Shortly thereafter, this clade would come to be known as Oviraptorosauria (Barsbold 1976a).

Throughout the 1970–90s, Rinchen Barsbold and Halszka Osmólska would continue to describe new specimens and species of oviraptorosaurs, revolutionizing our understanding of this enigmatic group and their link to other theropod dinosaurs (Barsbold 1976b, 1981, 1983, 1986, 1988, Osmólska 1981, Maryanska and Osmólska 1997, Barsbold et al. 2000). Since that time, the pace of research and discovery of oviraptorosaurs has only accelerated (Fig. 1). In the 1990–2000s, expeditions by the American Museum of Natural History collected some of the finest oviraptorosaur skeletons known to date, representing the exclusively-Asian Oviraptoridae (Norell et al. 1994, 1995, 2001, 2018, Clark et al. 1999, 2001, 2002, Balanoff and Norell 2012). These included some exquisite skeletons preserved atop nests of eggs (Norell et al. 1995, 2018), reversing the perception of oviraptorosaurs as egg-thieves into that of nurturing mothers. The torrid pace of new oviraptorosaur discoveries continued early in the new millennium (Lü 2002, Xu et al. 2002, 2007, Lü et al. 2004, 2005, Lü and Zhang 2005), and the first phylogenetic approaches on the group expanded their membership and extended their lineage further back to the Early Cretaceous (Maryańska et al. 2002, Osmólska et al. 2004).

Subsequently, in the last 20 years, nearly 30 new oviraptorosaur species have been named (Fig. 1). This diversity records a radiation of species throughout the Campanian–Maastrichtian, wherein oviraptorosaurs become the most diverse group of Late Cretaceous non-avian theropods (Funston et al. 2020) and disperse across Laurasia, as far south as Mexico (Serrano-Brañas et al. 2022) and as far west as Uzbekistan (Currie et al. 1993). Oviraptorosaurs were thus one of the last dinosaur clades to undergo a major diversification event, paralleling similar radiations in ceratopsian (Sampson and Loewen 2010) and hadrosaurid (Gates et al. 2012) dinosaurs.

Also in that interval, considerable advances have been made in understanding the anatomy, biology, and behaviour of oviraptorosaurs. The once-enigmatic caenagnathid dinosaurs are now represented by multiple relatively complete skeletons (Lamanna et al. 2014, Funston and Currie 2016, 2020), linking the unusual but often-isolated dentaries to the slender manus and elongate hindlimbs. At the same time, more egg–skeleton associations have revealed the reproductive biology of oviraptorosaurs (Sato et al. 2005, Wang et al. 2016, Yang et al. 2019, Bi et al. 2021, Xing et al. 2022, Yang and Sander 2022), showing that oviraptorids, in particular, retained the paired oviducts of more basal archosaurs and brooded their speckled blue eggs (Wiemann et al. 2018) with body heat, rather than substrate (Tanaka et al. 2015, 2018). Finally, spectacular assemblages of multiple individuals preserved together demonstrate that oviraptorosaurs were social animals, with numerous species consorting as juveniles and at least some forming long-lived groups of mixed ages (Funston et al. 2016b, 2019, 2020).

Despite these major advances and our excellent understanding of oviraptorosaurs, their fossil record still holds some surprises. For example, only recently was it recognized that some oviraptorids underwent a trend of digit and forelimb reduction, culminating in the loss of a functional third manual digit altogether (Funston et al. 2020). This didactyl oviraptorid from the Late Cretaceous of Mongolia (Fig. 2), Oksoko avarsanFunston et al. 2020, is among the best-represented oviraptorosaurs, known from multiple complete skeletons with skulls (Figs 3, 4). These specimens provide an opportunity to detail the anatomy, intraspecific variation, and ontogenetic changes of the oviraptorid skeleton. This latter aspect of oviraptorid morphology is particularly salient, as despite multiple species being represented by specimens from varying ages, little work has qualified the changes occurring throughout oviraptorosaur ontogeny. It thus remains unresolved whether features like cranial crests, sacral vertebral count, and hindlimb fusion, among others, might be expected to change over the course of an individual lifetime.

Here, I describe the osteology of Oksoko avarsan in detail and compare its skeleton to those of other oviraptorosaurs. Furthermore, I describe the variation apparent in the ontogenetic series known for Oksoko, and comment on the ramifications for taxonomic utility of certain characters. Finally, I highlight some remaining gaps in our understanding of oviraptorid dinosaurs, and exciting new research directions that have begun to open.

Institutional abbreviations

CMN, Canadian Museum of Nature, Ottawa, Ontario, Canada; MPC-D, Institute of Paleontology, Mongolian Academy of Sciences, Ulaanbaatar, Mongolia; ROM, Royal Ontario Museum, Toronto, Ontario, Canada; TMP, Royal Tyrrell Museum of Palaeontology, Drumheller, Alberta, Canada; UALVP, University of Alberta Laboratory for Vertebrate Palaeontology, Edmonton, Alberta, Canada.

Materials and methods

The four known specimens of Oksoko avarsan (MPC-D 100/33, MPC-D 102/11, MPC-D 102/12, and MPC-D 102/110) comprise at least six individuals. MPC-D 100/33 and MPC-D 102/12 each represent a single partial skeleton, whereas MPC-D 102/11 includes a relatively complete skeleton (MPC-D 102/11.a) and three cranial bones of another individual (MPC-D 102/11.b). The main block of MPC-D 102/110 includes two relatively complete skeletons and a partial skeleton of a third individual. These individuals are sub-numbered MPC-D 102/110.a, MPC-D 102/110.b, and MPC-D 102/110.c, in descending order of completeness. MPC-D 102/11 and MPC-D 102/110 were probably collected from the same assemblage (Funston et al. 2020), which has a minimum number of four individuals based on right quadrates. As discussed by Funston et al. (2020), the provenance of this assemblage is uncertain because the specimens were confiscated from poachers, who excavated the skeletons illegally. Several lines of evidence suggest that MPC-D 102/110 was collected from either Bugiin Tsav or Guriliin Tsav in the north-western part of the Nemegt Basin. Geochemical fingerprinting on MPC-D 102/110 strongly suggests that the specimens are from the Nemegt Formation, specifically the Nemegt locality, although the Bugiin Tsav and Guriliin Tsav localities were not included in that sample (Fanti et al. 2017). Another line of evidence is an abandoned ankylosaur skeleton re-discovered in 2018 at Guriliin Tsav, poached using similar methods and materials as used to collect MPC-D 102/110 (Fig. 2C, D). Both field jackets are made of thin pre-prepared blaster bandages with blue fibres, and yellow plastic bags were used as a separating layer. As these features of the specimens are unusual, this suggests, minimally, that the same group of poachers collected both specimens. Lastly, legitimately collected specimens of Oksoko avarsan are known from Bugiin Tsav (MPC-D 100/33) and Guriliin Tsav (MPC-D 102/12), showing that this taxon was present in the area.

Specimens of Oksoko avarsan and other oviraptorosaurs housed in the collections of the CMN, MPC, ROM, TMP, and UALVP were examined firsthand and measured using digital calipers (±0.1 mm) or a fabric measuring tape (±1 mm). Information about other specimens was taken from the literature. The specimens were photographed using a Nikon D5000, Nikon D7200, or Nikon COOLPIX AW120 camera with a variety of lenses.

CT scans of MPC-D 102/110 were performed at The National Museum of Natural History and Science in Tokyo, Japan. Despite minimal matrix adhered to the specimen, the scans suffered from severe beam hardening artefacts and ghosting, such that the borders of many elements at the surface are difficult to discern. Contrast within the block is relatively poor, especially within the endocranial cavities, which cannot be segmented. Contrast surrounding each cranium is particularly low, and thus details of their morphology cannot be ascertained from the CT scans. However, many of the more robust postcranial bones, including the vertebrae and limbs, can be adequately distinguished from the matrix and each other, and thus are useful for visualizing which parts of the individuals are preserved and how they are arranged, although their morphologies cannot be reconstructed. These images verify the arrangement of the specimens, their positions with respect to each other, and the associations between skulls and postcrania (Fig. 5). This latter aspect of the arrangement of the skeletons was initially ambiguous because of the unusual positions of the skulls appressed to the sternae. However, CT images show that the cervical vertebrae of each specimen are curled to form a spiral, confirming the ownership of each skull, as interpreted by Funston et al. (2020).

Description

Between the six skeletons known, every skeletal element of Oksoko avarsan is represented (Figs 3, 4). Each of these is exquisitely preserved and has suffered minimal post-mortem scavenging or weathering, and moderate to no crushing. These specimens reveal an oviraptorid with a distinctive cranial crest, short forelimbs with only two functional digits, elongate hindlimbs, and a relatively short tail (Fig. 3). MPC-D 102/110.a is the most complete skeleton, including a complete skull and skeleton, missing only the distal caudal vertebrae. MPC-D 102/110.b preserves a slightly disarticulated skull, parts of the vertebral column, a forelimb, the pelvis, and the hindlimbs. MPC-D 102/110.c preserves a partial ilium, some vertebrae, and a complete tibiotarsus that was revealed by CT scans. Thus, the individuals in MPC-D 102/110 (Figs 3, 5) provide excellent representation of the skeleton of Oksoko avarsan. Nonetheless, articulation of these specimen means that some elements are not visible in all views, and so some bones are better represented by the completely prepared specimens: MPC-D 100/33, MPC-D 102/11.a, MPC-D 102/12, and the fully prepared manus of MPC-D 102/110.a.

Cranial skeleton

The skulls of MPC-D 102/110.a and MPC-D 102/110.b are complete (Figs 6, 7), but are crushed mediolaterally. The posterior portion of the skull of MPC-D 102/110.b, including the braincase and suspensorium, is disarticulated slightly from the anterior part of the skull and rotated so that it is oriented perpendicular to the remainder of the cranium (Fig. 6). The skull of MPC-D 102/11.a is relatively complete, but is missing most of the anterior parts of the face, palate, and mandible (Figs 8, 9). The left side of the skull is roughly articulated and is well preserved. The right and posterior sides have been crushed and lie on a single plane parallel to the left side of the skull. Three extra skull bones from a second individual (MPC-D 102/11.b) are preserved alongside the more complete skull of MPC-D 102/11.a: the postorbital, quadratojugal, and quadrate (Fig. 10). The other specimens (MPC-D 100/33 and MPC-D 102/12) lack cranial elements.

Premaxilla

The premaxilla (Figs 6, 7) is completely preserved in MPC-D 102/110.a and MPC-D 102/110.b, but only a small portion is present in MPC-D 102/11.a (Fig. 8). It is tall dorsoventrally and constricted anteroposteriorly. Dorsally, it is divided by the naris into two processes: the nasal process extending dorsally and the subnarial process directed posterodorsally. The nasal process is much narrower than the subnarial process in lateral view, unlike in Citipati osmolskaeClark et al. 2001, Khaan mckennaiClark et al. 2001, Nemegtomaia barsboldi (Lü et al. 2004), and Rinchenia mongoliensis (Barsbold 1986), where these processes are subequal in width (Barsbold 1986, Clark et al. 2001, 2002, Lü et al. 2004, Balanoff and Norell 2012, Funston et al. 2018). In contrast, the nasal process is wider than the subnarial process in Banji longXu and Han 2010, Huanansaurus ganzhouensisLü et al. 2015, and Tongtianlong limosusLü et al. 2016. The nasal process of Oksoko curves posterodorsally so that it forms a small part of the continuous semicircular crest with the nasals, frontals, and parietals, but not to the same extent as the anteroposteriorly broad premaxilla of Tongtianlong limosus. In Oksoko avarsan, the nasal process extends dorsally to the ventral third of the naris, whereas in Banji long and Rinchenia mongoliensis the premaxilla extends far above the naris (Xu and Han 2010, Funston et al. 2018). The subnarial process of the premaxilla is broad and tapers posteriorly. Posteriorly, it separates the lacrimal and nasal anteriorly and prevents the maxilla from contacting the nasal on the lateral surface of the skull. The lateral surface of the body of the premaxilla is pierced by multiple small foramina. Ventral to the small, oval naris, there is a lateral depression in the premaxilla similar to that of Citipati osmoslkae (Clark et al. 2001), but shallower than the prominent fossa in Banji long (Xu and Han 2010). The occlusal margin of the premaxilla has at least two denticulations, but this area is broken in both individuals and there may have been more. The occlusal edge of the premaxilla is relatively longer anteroposteriorly than in Rinchenia mongoliensis or Corythoraptor jacobsiLü et al. 2017, more comparable to Citipati osmolskae or Nemegtomaia barsboldi. The palatal surface of the premaxilla cannot be seen on any of the specimens.

Maxilla

The maxilla (Figs 6, 7) is missing in MPC-D 102/11.a and poorly preserved in both individuals of MPC-D 102/110, but best observed in MPC-D 102/110.b (Fig. 6B). The antorbital fossa is small and the antorbital fenestra is divided in two by a dorsally expanding strut of bone (Fig. 6B), as in most oviraptorids. The jugal process is relatively short and extends only partway under the orbit. The labial–buccal transition on the lateral side of the maxilla is marked by a ridge, ventral to which there is a pronounced lateral tubercle, as in Rinchenia mongoliensis (Funston et al. 2018). The presence of longitudinal palatal ridges on the maxilla, as exhibited in Citipati, cannot be determined because of overlying matrix. Likewise, the maxillovomeral tubercle (=palatal ‘tooth’), which is present in all oviraptorids, is obscured by matrix in MPC-D 102/110, so the contribution of the maxilla cannot be determined.

Nasals

The fused nasals (Figs 6–9, 11) are complete but crushed in both individuals of MPC-D 102/110. Only the posterolateral wings of the nasals are preserved in MPC-D 102/11.a (Figs 8, 9, 11). In this individual, the nasals are fused along the midline, but posteriorly a suture is still visible (Fig. 11). Like Rinchenia mongoliensis and Corythoraptor jacobsi, the great size of the nasals is mostly due to expansion of the lateral descending processes. Similar to Citipati osmolskae (Clark et al. 2001) and Khaan mckennai (Balanoff and Norell 2012), the posterodorsal part of the premaxilla excludes the maxilla from contributing to the naris and from contacting the nasal on the lateral surface of the skull. The elliptical naris is displaced far dorsally, unlike Huanansaurus ganzhouensis and Tongtianlong limosus, and overlies the antorbital fenestra, similar to Citipati osmolskae (Clark et al. 2001), but differing from the elongate naris of Banji long (Xu and Han 2010), which extends further ventrally. In Oksoko, the naris extends above the dorsal margin of the orbit; posteriorly its ventral margin is level with the top of the orbit. In contrast, the naris is smaller and situated further anteroventrally in Rinchenia mongoliensis and probably Corythoraptor jacobsi (contraLü et al. 2017). In Oksoko, the anterior midline process forms less than half the length of the nasal, and curves anteroventrally to meet the posterodorsal process of the premaxilla. It lacks the step-like process present in Banji long that demarcates the posterior extent of the premaxillo-nasal crest. In lateral view (Figs 6, 7, 8), the anterior midline process of Oksoko is broad dorsoventrally, the result of a septal ridge on its ventral surface. The lateral descending processes host several pneumatic cavities, which vary in position between each specimen. In the better-preserved MPC-D 102/11.a (Figs 8, 11), the midline rami of the fused nasals are thickened and rounded dorsally, delimiting a lateral recess in which the pneumatic cavities lie. These complex asymmetric pneumatic pockets are similar to those of MPC-D 100/42 and Citipati osmolskae and Banji long, and appear to fully penetrate the nasal and open into the nasal passage, but they are not as extensive as in Corythoraptor jacobsi or Rinchenia mongoliensis, where they form a honeycomb-like internal structure. The buttress formed by the fused nasals above the pneumatic pockets in Oksoko is confluent with the similarly thickened dorsal parts of the frontals and parietals. Together, they form a raised rim on the dorsal margin of the cranial crest. The posterior suture with the frontal is simple and straight, like in Citipati osmolskae (Clark et al. 2001), and does not have the interdigitating irregularity present in Khaan mckennai and Tongtianlong limosus (Balanoff and Norell 2012, Lü et al. 2016). The fused nasals are arched transversely so that their wing-like posterolateral processes are nearly vertically oriented, as in Corythoraptor jacobsi and Rinchenia mongoliensis. Although this may be exaggerated by transverse crushing of the skulls in MPC-D 102/110, a similar arched morphology is present in an undeformed skull of Citipati osmolskae (MPC-D 100/978; Clark et al. 2001), where the nasals contribute to a cranial crest.

Lacrimal

The lacrimal (Fig. 6B) is well preserved in all three skulls, but best observed in MPC-D 102/110.b. The lacrimal contacts the nasal dorsally, the maxilla and premaxilla anteriorly, and the frontal posteriorly above the orbit. Like other oviraptorids, the lacrimal has a poorly developed anterior nasal process, and lacks the T-shape present in Caudipteryx zouiQiang et al. 1998 and deinonychosaurs. The foramen for the lacrimal duct faces anteriorly and opens into a shallow subhorizontal channel on the lateral surface of the lacrimal. The pneumatic pockets seen on Banji long (Xu and Han 2010), Citipati osmolskae (Clark et al. 2001), Khaan mckennai (Balanoff and Norell 2012), and Tongtianlong limosus are present on only one individual, MPC-D 102/110.b. The frontal process is long, but does not extend posteriorly past the nasal. A triangular process of the frontal separates the lacrimal from the dorsal edge of the orbit, creating a Z-shaped suture in lateral view. The jugal process curves posteroventrally to meet the jugal, forming an inclined posterodorsal margin to the antorbital fenestra. The preorbital bar of the lacrimal is anteriorly pierced at its midlength by a small, vertical slit that communicates with the orbit. The preorbital bar is flared slightly laterally, its posterior margin is broadly concave, and it forms the entire anterior margin of the orbit.

Frontal

The frontal (Figs 6–9, 11) is well preserved in each skull. It is taller dorsoventrally than long anteroposteriorly, which is unusual for a theropod. It forms most of the dorsal part of the orbit, with the postorbital, and is invaded by the supratemporal fenestra posteriorly. It tapers dorsally, so that it is longer anteroposteriorly above the orbit than on the dorsal margin of the cranial crest. It contacts the lacrimal and nasal anteriorly, the postorbital laterally, the parietal posteriorly, and the laterosphenoid posterolaterally. The postorbital process is elongated dorsoventrally to accommodate the tall frontal process of the postorbital, and continues dorsally as a distinct ridge separating the supraorbital and supratemporal portions of the frontal. The supraorbital part of the frontal is concave laterally, and has a dorsoventrally oriented pneumatic depression on most of its lateral surface. The supratemporal part of the frontal is not pierced by pneumatic openings, but the breakage pattern on the left side of MPC-D 102/11.a suggests that it may have been hollow above the braincase, as in Citipati osmolskae (Clark et al. 2001). Unlike MPC-D 102/110, the frontals of MPC-D 102/11.a are not fused, but all three individuals have a simple, straight contact. The frontals are separated only slightly posteriorly by the parietals, with which they have a simple, obtusely angled contact. The supraorbital rim lacks a supraciliary lip, which is the case in Citipati osmolskae (Clark et al. 2001), but not in Khaan mckennai (Balanoff and Norell 2012). The dorsal surface of the orbit is badly crushed in all individuals, but appears to have a groove posteriorly.

Parietal

The parietal (Figs 6–9, 11) is preserved in each skull. It is tall dorsoventrally, similar to Tongtianlong limosus, but this has probably been slightly exaggerated by transverse crushing in each individual. The parietals of MPC-D 102/11.a are fused completely, although a furrow is still visible on the dorsal midline (Fig. 11). The lateral surface of the parietal is deeply concave, to accommodate mandibular adductor musculature. The sagittal crest is tall, extending about 20 mm above the apex of the nuchal crest in MPC-D 102/11.a, and is transversely expanded, tapering posteriorly. The nuchal crest is pronounced, like in Huanansaurus ganzhouensis, and merges at the midline with the sagittal crest. The posterior end of the sagittal crest is near the apex of the cranial crest, posterior to which the skull roof descends steeply (Figs 6, 7, 12). A similarly sloped skull roof is present in most oviraptorids, except Banji long (Xu and Han 2010), Conchoraptor gracilisBarsbold 1986 (Funston et al. 2018), Khaan mckennai (Balanoff and Norell 2012), and Yulong mini Lü et al. 2013. The parietal contributes only to the posterior half of the medial surface of the supratemporal fenestra, contrasting with Khaan mckennai (Balanoff and Norell 2012), where it forms the majority of this surface. The transversely straight but dorsally arched occipital margin of the parietal is the widest point of the bone (Fig. 11). In MPC-D 102/110.a, b, the arched margin of the parietal contacts the supraoccipital and exoccipital posteriorly and the squamosal laterally.

Jugal

The jugal (Figs 7, 8) is best observed in MPC-D 102/110.a and MPC-D 102/11.a. It is triradiate and relatively robust compared to other oviraptorids, but remarkably similar to that of Banji long (Xu and Han 2010) and somewhat similar to Huanansaurus ganzhouensis and Tongtianlong limosus (Lü et al. 2015, 2016). Unlike most oviraptorids where the ventral margin of the jugal is straight or dorsally arched, in Oksoko it is slightly sinuous, also similar to Huanansaurus ganzhouensis and Tongtianlong limosus. The maxillary process is dorsoventrally broad towards its posterior base and tapers anteriorly where it meets the lacrimal and maxilla. The postorbital process is wide in lateral view, and has an anterior facet for the postorbital that extends ventrally only a third of the length of the postorbital process. In Citipati osmolskae (Clark et al. 2001) and Khaan mckennai (Balanoff and Norell 2012) this facet extends much further ventrally, almost to the junction of the postorbital process and the other two processes of the jugal. The quadratojugal process of the jugal is short and laterally overlies the quadratojugal. It is much shorter and more robust than that of Rinchenia mongoliensis (Funston et al. 2018), where it is bifurcated posteriorly to interfinger with the quadratojugal.

Postorbital

The postorbital (Figs 6–8, 10) is preserved in all of the articulated skulls and an isolated postorbital is present with MPC-D 102/11.b (Fig. 10). The postorbital is tall and its frontal process is vertical, which distinguishes this taxon from all other oviraptorids, except Rinchenia mongoliensis (Funston et al. 2018). In other oviraptorids, the frontal process is oriented anterodorsally, and is typically shorter than the jugal process. In MPC-D 102/11.a (Figs 8, 10) and MPC-D 102/110.b, the jugal process is shorter than the frontal process, and only forms half of the posterior orbital margin. It is slightly longer in MPC-D 102/110.a (Fig. 7), where it forms almost two-thirds of the orbital margin. Overall, the jugal process is less robust and more gently curved than in Corythoraptor jacobsi, Huanansaurus ganzhouensis, or Tongtianlong limosus (Lü et al. 2015, 2016, 2017), but is similar in these respects to Citipati osmolskae (Clark et al. 2001). The anterior (orbital) margin of the postorbital is strongly sinuous, with a concave orbital portion and a convex frontal portion. In most oviraptorids, the anterior margin of the postorbital is smoothly concave. The squamosal process is unbroken only in MPC-D 102/110.b, where it is dorsoventrally broad and anteroposteriorly short.

Squamosal

The squamosal (Figs 6–9) forms the posterodorsal corner and upper margin of the subrectangular infratemporal fenestra. The postorbital process has a lateral rugosity and a dorsolateral groove for the postorbital. The lateral aspect of the squamosal is divided into two main parts by a curved ridge that delimits the corner of the infratemporal fenestra (Fig. 12). Anteroventral to this ridge, the squamosal is laterally depressed and probably would have anchored adductor musculature. Medially the squamosal has an anteroposteriorly wide contact with the parietal, best observed in MPC-D 102/110.b (Fig. 6B). The squamosal bifurcates ventrally, as in all oviraptorids. The posterior process contacts the paroccipital process of the exoccipital and encapsulates the external auditory meatus. Ventromedially, the squamosal contacts and fuses to the quadrate, and ventrolaterally it contacts the quadratojugal, where it borders the external auditory meatus.

Quadratojugal

The triradiate quadratojugal (Figs 6–10) is not fused to the quadrate, even in the large, isolated individual (MPC-D 102/11.b; Fig. 10) associated with MPC-D 102/11.a. The anterior process lies medial to the jugal and forms about two-thirds of the ventral margin of the infratemporal fenestra. Dorsally, the ascending process forms at least half of the posterior border of the infratemporal fenestra, although its full extent is obscured by fusion to the squamosal. The ascending process contacts the quadrate along most of its length, but is separated ventrally by a quadrate foramen. The posterior process of the quadratojugal forms a cap on the lateral surface of the quadrate, and has a tab-like posteroventral extension. The posterior process is relatively long, comparable to Banji longXu and Han 2010, Citipati osmolskae (Clark et al. 2001), and Nemegtomaia barsboldi (Fanti et al. 2012).

Quadrate

The quadrate (Figs 6–10) is poorly exposed in MPC-D 102/110, and in MPC-D 102/11.a it is badly damaged on the left and preserved only in posterior view on the right (Fig. 9). The isolated right quadrate from MPC-D 102/11.b is incomplete, but useful for determining morphology (Fig. 10). Medially, the quadrate contacts the parasphenoid, pro-otic, and pterygoid; laterally, the quadratojugal; and dorsally, the squamosal. It does not appear to contact the exoccipital/opisthotic extensively, although on both sides of MPC-D 102/11.a these bones are disarticulated and this area is not visible in the skulls of MPC-D 102/110. The optic wing of the quadrate is oblique to the midline, extending antero-dorso-medially to postero-ventro-laterally, and covers most of the lateral surface of the braincase. Anteriorly, the optic wing of the quadrate contacts the epipterygoid. At this point, it also contacts the pterygoid ventrally. The condyles of the quadrate are saddle-shaped as in other oviraptorosaurs. There is a large, vertically oriented, oval quadratojugal foramen, formed entirely by excavation of the quadrate. Just medial to this, on the posterior side of the quadrate, there is a deep depression. There is a vertical ridge on the posterior surface of the quadrate, just lateral to the optic wing. The lateral surface of the quadrate contacts—but is not fused with—the quadratojugal, and this contact is raised into a distinct process with a cotyle bifurcated by a ridge.

Palatal skeleton

The palatal skeleton is incomplete in MPC-D 102/11.a (Figs 8, 9), with only the pterygoids, ectopterygoids, and a small part of the right palatine preserved. In MPC-D 102/110, the palatal skeleton is mostly obscured by the overlying mandible (Fig. 6). The ectopterygoid has a dorsally curved maxillary process, which would have contacted the maxilla dorsally and the palatine anteriorly. Anteriorly, the pterygoid has a crescentic contact with the ectopterygoid in lateral view, as in almost all oviraptorids, although this suture appears more pointed in Banji long (Xu and Han 2010) than in Oksoko avarsan. The ramus of the pterygoid is short anteroposteriorly, tall dorsoventrally, and concave on its ventral surface. The pterygoid ramus lacks the deep dorsal excavation present in Banji long (Xu and Han 2010). The pterygoids are separated by an interpterygoid vacuity like in Jiangxisaurus ganzhouensis (Wei et al. 2013), and like other oviraptorosaurs (Clark et al. 2002, Balanoff et al. 2009), the pterygoid lacks a transverse flange, although this area is poorly exposed in MPC-D 102/11.a. The pterygoid has a broad posterodorsally facing contact with the quadrate, from which it tapers anteriorly. Posteriorly, it underlies the optic wing of the quadrate, and is mostly obscured by crushing. At its posterior end, it is dorsoventrally tall and transversely thin. At this point, it contacts the basipterygoid process of the basisphenoid medially.

Occiput

The occiput is not visible on either skull of MPC-D 102/110 (Fig. 6). In MPC-D 102/11.a (Figs 9, 13) it is well preserved and none of the bones of the occiput are fused, although the basioccipital and basisphenoid have begun to co-ossify. The trapezoidal supraoccipital is disarticulated from the rest of the occiput. The facets for the exoccipitals on the supraoccipital are separated by a groove, indicating that the supraoccipital contributed to the foramen magnum. The supraoccipital tapers anterodorsally, where it would have fit between the squamosal processes of the fused parietals. The dorsal surface of the supraoccipital bears two longitudinal, shallow, elliptical depressions. The ventral surface of the supraoccipital has a lateral lamina on each side, which would have formed the walls of the foramen magnum. The exoccipitals were not yet fused to the basioccipital, similar to Banji long, the holotype of which probably also represents a juvenile individual (Xu and Han 2010). Both exoccipitals have been rotated anterolaterally from life position, which exposes their medial sides (Fig. 13). A medioventral process indicates that the exoccipital formed only the dorsolateral corner of the occipital condyle. The exoccipital is thickened dorsomedially where it contacts the supraoccipital, and tapers laterally towards the dorsal border of the paroccipital process. The paroccipital process curves ventrally, and has a raised, undulating lateral edge. Medially, the base of the paroccipital process has a depression, which is bordered laterally by a rounded ridge that extends along the long axis of the paroccipital process. On the medial (internal) surface, which forms the posterior wall of the foramen magnum, there are multiple foramina (Fig. 13D). The largest, the metotic foramen, remains filled with matrix, but presumably housed the foramen for cranial nerves IX, X, and XI, as in Incisivosaurus and deinonychosaurians (Currie and Zhao 1993, Currie 1995, Balanoff et al. 2009). The opening of the metotic foramen on the external surface of the braincase is not visible in MPC-D 102/11.a because of the dislocation of the exoccipitals, and this area has not been fully prepared in MPC-D 102/110.a and MPC-D 102/110.b. Posteroventral to the metotic foramen are three internal foramina for cranial nerve XII (Fig. 13D), one extra compared to Incisivosaurus, where there are two. Cranial nerve XII also exits through two foramina on the exterior surface of the exoccipital ventrolateral to the occipital condyle (Fig. 13C). On the internal surface of the foramen magnum, dorsal to the foramina for cranial nerve XII, is a small foramen set in a concavity, for the ductus endolymphaticus (Fig. 13D). The basioccipital is articulated with the basisphenoid, and although they have begun to co-ossify in this individual, they were not yet fully fused. The occipital condyle is kidney-shaped, and has a ventrally constricted neck. The basal tubera are situated ventral to the occipital condyle in posterior view (Fig. 13), rather than level with it as in Citpati osmolskae (Clark et al. 2001: fig. 6). The basal tubera are not widely spaced and are relatively small, which may be a consequence of the early developmental stage of MPC-D 102/11.a. The basisphenoid is not well exposed in MPC-D 102/11.a, but several features can be discerned. It has begun to co-ossify with the basioccipital, but there is still a suture. The basipterygoid processes face ventrolaterally, and are separated by a dorsoventrally oriented groove. This groove extends dorsally to the basisphenoid recess. Lateral wings of the basisphenoid extend dorsally to encapsulate the basioccipital, and contact the exoccipital and probably pro-otic.

Braincase wall

Most of the lateral wall of the braincase is obscured by the overlying bones in MPC-D 102/110 (Fig. 6), but on the left side of MPC-D 102/11.a parts of the laterosphenoid and parasphenoid are exposed (Fig. 8). The laterosphenoid extends dorsally into the supratemporal fenestra, terminating just dorsal to the supratemporal bar in lateral view. Only the anterior part of the pro-otic is exposed, where it is pierced by the large foramen ovale. For such a delicate element, the parasphenoid rostrum is exceptionally preserved in MPC-D 102/11.a (Fig. 8). It occupies the space dorsal to the interpterygoid vacuities, and has been taphonomically shifted dorsally so that it lies at the centre of the orbit. It is transversely narrow but dorsoventrally tall and straplike. The anterior end is modified into a ‘boot’, superficially similar in shape to the pubic boot of most oviraptorosaurs. At its posterior end, it is pierced by two small foramina.

Scleral ossicles

Dozens of small, crushed plates of bone are present in the orbits of both MPC-D 102/110.a–b (Figs 6, 7). The more complete plates are roughly square, with rounded corners. The largest is 6.5 mm in height and 9 mm in length, about 20% the anteroposterior length of the orbit.

Mandible

The lateral surface of the mandible is well preserved in MPC-D 102/110.a (Figs 6, 7), but is mostly missing from MPC-D 102/11.a. The dentary is tall and downturned anteriorly like in Jiangxisaurus ganzhouensis (Wei et al. 2013), but with a pronounced ventral chin. The occlusal margin is concave anterodorsally and the labial surface is marked by minute foramina. The posterodorsal ramus is broad and strap-like, tapering posteriorly where it contacts the surangular above the heart-shaped external mandibular fenestra, which does not extend as far anteriorly into the dentary as in Banji long (Xu and Han 2010). The posteroventral ramus of the dentary is long, extending as far as the surangular fossa on the lateral side of the surangular, and it tapers where it underlies the angular. The coronoid process is tall and protrudes far above the rest of the mandible, but does not appear to be medially deflected like it is in Jiangxisaurus ganzhouensis (Wei et al. 2013). The dorsal margin of the mandible has only a single apex, in the surangular, rather than two apices, as is the case in Huanansaurus ganzhouensis and the Dzamyn Khondt oviraptorid. The surangular prong is broken in MPC-D 102/110.a, but appears to be present in MPC-D 102/110.b. The angular is straplike and extends to the anterior end of the external mandibular fenestra. The surangular has a deep recess, which may have housed a surangular foramen, but this region is broken. The articular has a tall, convex articular ridge and a small, posteriorly directed retroarticular process. As in Rinchenia mongoliensis (Funston et al. 2018), and unlike all other oviraptorosaurs, the retroarticular process of MPC-D 102/110.a is composed predominantly of the surangular, rather than including a contribution from the angular. The presence of a coronoid, reported only occasionally but probably present throughout Oviraptorosauria (Clark et al. 2002, Balanoff et al. 2009, Funston and Currie 2020), cannot be determined because none of the mandibles can be observed in medial view.

Ceratobranchial

A long, rod-shaped ceratobranchial is preserved just lateral to the mandible of MPC-D 102/110.a (Fig. 7). The anterior end of the element is expanded dorsoventrally. The shaft is straight and cylindrical, unlike the curved ceratobranchial of Citipati osmolskae (Clark et al. 2001). The ceratobranchial is nearly half the length of the mandible, and about one-third the length of the skull.

Axial skeleton

Alongside the anteriormost cervical vertebrae, only the posterior sacral vertebrae and the ventral surfaces of the caudal vertebrae are exposed in MPC-D 102/110 (Fig. 3). The axial skeleton of MPC-D 102/11.a is represented by an incomplete atlas, a partial axis, two anterior cervical vertebrae (Fig. 9), four posterior dorsal vertebrae, a sacrum composed of three co-ossified vertebrae, and a complete caudal series. A nearly complete vertebral column is known from MPC-D 102/12, although it is missing the sacrum and the anterior cervical vertebrae. Only the sacrum and caudal vertebrae are currently mounted with MPC-D 100/33, but photographs taken in September 2001 by P. Currie show a nearly complete axial series, including the atlas–axis, anterior cervical vertebrae, dorsal vertebrae, a sacrum, and caudal vertebrae. Together, the vertebrae from all specimens represent the entire axial column, and most positions are represented by multiple individuals.

Cervical vertebrae

There are 12 cervical vertebrae, including the atlas and axis (Fig. 14). The anterior cervical vertebrae of MPC-D 102/11.a (Fig. 9) are incompletely ossified, and their articulation with the base of the skull obscures the morphology of the centra. The neural arches are not fused to the centra and have low neural spines. The right side of the neural arch of the atlas is exposed and is separate from the left, but the two halves would probably have fused later in life. The atlas intercentrum and the odontoid process of the axis are missing, so their morphology cannot be discerned. The axis has an anteroposteriorly long neural spine, which is transversely thickened distally and extends posteriorly past the centrum. The first postaxial cervical vertebra has a relatively tall and fingerlike neural spine, although the rest of the neural arch is broad and dorsoventrally flattened, typical of oviraptorosaurs (Balanoff and Norell 2012). MPC-D 102/12 and MPC-D 100/33 provide more information on the entire cervical vertebral series (Fig. 14). The atlas–axis (Fig. 14C) of MPC-D 100/33 is tightly adhered, but sutures are still visible between the atlas intercentrum and the axis. The neuropophyses of the atlas are fused to each other along the midline and have begun to co-ossify with the intercentrum. The neural spine of the axis is missing in this specimen, but its morphology is bulbous in MPC-D 102/110 and MPC-D 102/11.a (Fig. 9). In contrast, the neural spine of the third cervical is peglike (Fig. 9). The centra of the anterior cervical vertebrae have steeply inclined articular faces and widely spaced parapophyses, resulting in a triradiate appearance in ventral view (Fig. 14F). Each has a deep, lateral pleurocoel and a concave, posterior articular surface. The neural arches are as wide as they are long and have large, circular, dorsomedially facing, anteriorly extending prezygapophyses. These are connected to the postzygapophyses by a broad lamina, from which the transverse processes barely protrude. Large, moundlike epipophyses (Fig. 14E) sit on the dorsal surfaces of the postzygapophyses, seemingly comparable to those in Jiangxisaurus ganzhouensis and Tongtianlong limosus (Wei et al. 2013, Lü et al. 2016), and these appear to become larger in more posterior vertebrae, like in Huanansaurus ganzhouensis (Lü et al. 2015). The neural spines are low and square. Based on their absence, the cervical ribs had not yet fused in MPC-D 100/33, but in MPC-D 102/12, some appear to have fused to the parapophyses. MPC-D 102/12 is missing the anterior cervical vertebrae but preserves the mid and posterior cervical vertebrae. The centra become relatively taller posteriorly along the cervical vertebral series (Fig. 14B), and this is accompanied by broadening of the neural arches so that they are wider than they are long, like in Khaan mckennai and Tongtianlong limosus. The centra retain large pleurocoels, but the articular faces become less inclined posteriorly along the series. Regardless, the posterior articular face remains concave throughout the series. The transverse processes become better developed and fuse to the cervical ribs, which decrease in relative length successively. The epipophyses are large until about the eighth postaxial cervical vertebra, after which they decrease in size.

Dorsal vertebrae

A complete dorsal vertebral series is preserved with each of MPC-D 102/12 and MPC-D 100/33, although they are better preserved in MPC-D 102/12 (Fig. 15). The posterior dorsal vertebrae, the sacral vertebrae, and four of the anterior caudal vertebrae are articulated with the right ilium of MPC-D 102/11.a (Fig. 16). The dorsal vertebral series comprises 10 vertebrae, which increase in size successively. The anterior three vertebrae have hypapophyses, which are largest on the anterior one and smallest on the posterior one. The centra are barrel-shaped with a ventral keel and develop a ventral curve in lateral view towards the posterior end of the series. This is accompanied by an increase in the size of the lateral pleurocoel—which is present on all dorsal vertebrae—and a transverse broadening of the centrum. In MPC-D 102/12, the neural arches are fused to the centra, and in most cases the suture is closed (Fig. 15). This is not the case in MPC-D 100/33, where the neural arches are not fused and in many cases have become disarticulated. The parapophyses of MPC-D 102/12 are large and concave. They become more dorsally positioned posteriorly along the dorsal vertebral series, transitioning from a location exclusively on the centrum (D1–D5), to bridging the neurocentral suture (D6–D8), to exclusively on the neural arch (D9, D10). The neural arches are deeply excavated by infraprezygapophyseal, infradiapophyseal, and infrapostzygapophyseal fossae, which in some cases have merged, leaving a strut of bone remaining (Fig. 15H). The infraprezygapophyseal fossae become shallower in more posterior vertebrae, whereas the infradiapophyseal and infrapostzygapophyseal fossae remain deep throughout. The neural spines become taller until D8, after which they are slightly shorter.

Sacral vertebrae

There are six sacral vertebrae (Fig. 16), which are all fused in MPC-D 100/33. In contrast, only three vertebrae have been co-ossified to form the sacrum in MPC-D 102/110.a, MPC-D 102/110.b, and MPC-D 102/11.a (Fig. 16B), which reflects the early ontogenetic stages of these individuals. Like all oviraptorosaurs, the centra of the sacrum have large, lateral pleurocoels and are flattened ventrally. However, they are not flattened to the same degree as in caenagnathids and the pleurocoels sit above the ventral surface of the sacrum rather than opening to the ventral surface. The second sacral vertebra shows an incipient ventral keel, whereas sacral vertebrae 3–6 have a midline groove. The anterior sacral neural arches of MPC-D 100/33 are missing, but the posterior ones have fused together into a fan-like sheet of bone. In MPC-D 102/11.a, the neural spines remain separate dorsally (Fig. 16A), but their ventral bases have begun to fuse. Unlike in caenagnathids, the transverse processes and their accompanying sacral ribs do not vary in position along the sacral series. Rather, in each vertebra they are consistently located at the level of the neurocentral suture. As in caenagnathids, however, the transverse process and sacral rib of sacral vertebra 5 appear to be the largest, although not forming the same hatchet-shaped process. The lack of sacral fusion in MPC-D 102/11.a provides insight into the somitic origin of the sacral series. The three fused sacral vertebrae probably represent the primordial sacral vertebrae, based on the extent of their fusion early in life. Indeed, three sacral vertebrae are fused even before hatching in oviraptorids (Norell et al. 2001). Accordingly, the anteriormost sacral vertebra must have been recruited from the dorsal series and two caudosacral vertebrae must have been incorporated from the tail (Fig. 16A). This is supported by the morphology of these vertebrae, which most closely resemble the dorsal and caudal vertebrae, respectively.

Caudal vertebrae

The complete caudal vertebral series of Oksoko avarsan would have had 29 caudal vertebrae, the last three of which fuse into a pygostyle later in life for a total count of 27 caudal vertebrae (Fig. 17). MPC-D 102/11.a preserves 27 of the 29 caudal vertebrae (Fig. 17B), and is missing only the last two pygal vertebrae. MPC-D 102/12 has the complete pygal series, but is missing the proximal caudal vertebrae (Fig. 17C), which were presumably lost at the same time as the sacral vertebrae, resulting in a total of 27 vertebrae, the last three of which are fused into the pygostyle. MPC-D 100/33 has 23 caudal vertebrae from the middle part of the tail, missing both the proximal and distal vertebrae. MPC-D 102/110.a–c preserve seven, four, and seven caudal vertebrae, respectively, from the base of the tail, but are missing the distal parts of the tail. The centra of the proximal caudal vertebrae have pleurocoels, like in Jiangxisaurus (Wei et al. 2013), although in Oksoko they are reduced in size relative to the dorsal and sacral vertebrae. From the 19th caudal vertebra to the tip of the tail, pleurocoels are absent. In MPC-D 102/12, some of the proximal pleurocoels have become infilled with bone, but their borders can still be discerned (Fig. 17D). The centra of the caudal vertebrae are barrel-shaped, rather than anteroposteriorly elongated as in many theropods. Posteriorly along the vertebral series, each centrum is slightly more elongate relative to its height (Fig. 17H–J), but not to the degree seen in theropods like deinonychosaurs, ornithomimids, and tyrannosaurs. In MPC-D 102/11.a, the anterior neural arches are not fused to the caudal vertebrae, but the 17 posteriormost vertebrae have neural arches that are fused with a closed suture. In contrast, all of the caudal vertebrae of MPC-D 102/12 have fused neurocentral sutures. The neurocentral fusion of the posterior caudal vertebrae and the lack of fusion in all the other vertebrae of MPC-D 102/11.a suggests that closure of the neurocentral sutures proceeds posterior to anterior, as in crocodylians (Brochu 1996, Irmis 2007). This lends support to previous suggestions that closure of the neurocentral sutures in the cervical vertebrae provides evidence of maturity in oviraptorosaurs (Funston and Currie 2016). There is a relatively large infradiapophyseal fossa below the high transverse process on the anterior caudal vertebrae. In MPC-D 102/12, this is accompanied by a supradiapophyseal fossa on the anterior two caudal vertebrae. The transverse processes descend progressively towards the lateral surface of the centra posteriorly and become shorter mediolaterally. Their orientation also changes from being directed posteroventrally to more directly laterally. In MPC-D 102/11.a, the transverse processes persist until the eighth last vertebra, whereas in MPC-D 102/12 they persist until the last vertebra preceding the pygostyle. However, these distal transverse processes are anteroposteriorly elongate and hatchet-shaped in dorsal view (Fig. 17I), barely protruding from the centrum. Similar transverse processes are present in MPC-D 100/33, but they do not extend as far down the tail, probably representing an intermediate ontogenetic stage of development.

Ribs and gastralia

The dorsal ribs are poorly exposed in MPC-D 102/110 (Fig. 3), but CT scans show that they are present in MPC-D 102/110.a (Fig. 5). In this individual, some of the proximal parts of the posterior dorsal ribs are exposed on the left side of the individual, and the distal portions of the shafts of the dorsal ribs are exposed where they are articulated with the gastral basket. Ventral (sternal) ribs are also preserved on the right side of MPC-D 102/110.a (Fig. 3), where four relatively straight, strap-like ventral ribs are associated with the right sternal plate, as in Jiangxisaurus, although the posterior rib is not fused in Oksoko. Each has been displaced to some degree from their natural articulations, but it is clear that the anteriormost and posteriormost ventral ribs are distinctly smaller than the other two ventral ribs, of which the third is largest. The proximal end of each ventral rib is positioned between the sternocoracoidal process and lateral trabecula of the sternum. Two posterior rib heads are preserved with MPC-D 102/11.a, but they provide little information. Four partial ribs are preserved with MPC-D 102/12 (Fig. 18). The head of the ribs are relatively simple, lacking the pneumatization present in caenagnathids like Apatoraptor pennatusFunston and Currie 2016. The largest of the four ribs has a broad capitulum, suggesting that it is from the middle part of the dorsal series. On the posterolateral surface of the shaft, there is a facet for the attachment of an uncinate process (Fig. 18D). However, the uncinate process was not recovered with the skeleton. Uncinate processes are not visible in the CT scans of MPC-D 102/110.a, suggesting that they had not yet ossified at this ontogenetic stage. Whether they ossified later in life is unclear. The gastralia are well preserved and articulated in MPC-D 102/110, but not the other specimens. The complete gastral basket is preserved in MPC-D 102/110.a. There are 12 rows of medial gastralia, and lateral gastralia are associated with most of these rows. The right medial gastralia are offset anteriorly from the left gastralia, and anteriorly some of the medial gastralia fuse, as in other theropods.

Chevrons

The anterior chevrons are preserved in MPC-D 102/110.a–c (Fig. 19A, B), and relatively complete series of chevrons are preserved with MPC-D 100/33, MPC-D 102/11.a, and MPC-D 102/12. The anterior chevrons of MPC-D 102/110.a, MPC-D 102/110.c, MPC-D 102/11.a, MPC-D 100/33, and MPC-D 102/12 are elongate and taper distally to a bulbous process (Fig. 19C), similar to Heyuannia yanshini (Barsbold 1981) (Funston et al. 2018). However, the first chevron of MPC-D 102/110.b is unusual and extremely small compared to the other specimens (Fig. 19B), despite similarity in the size of the associated caudal vertebrae. Indeed, the first chevrons of MPC-D 102/110.a, c are more than twice the dorsoventral height of MPC-D 102/110.b. This condition is reminiscent of the dimorphism described in the chevrons of Khaan mckennai, although it is more extreme than in Khaan mckennai (Persons et al. 2015). Without a larger sample size, the nature of this dimorphism cannot be determined, although the similarity in size and morphology of MPC-D 102/110.a and MPC-D 102/110.b suggests that it is unlikely to be the result of ontogenetic differences. All of the distal chevrons are preserved with MPC-D 102/11.a (Fig. 17B), and some of the series is preserved with MPC-D 100/33. These are all elongate dorsoventrally, rather than becoming platelike, as is the case in caenagnathids and MPC-D 100/42, the Dzamyn Khondt oviraptorid.

Pectoral girdle

The complete pectoral girdle of MPC-D 100/33 is preserved and was disarticulated during preparation, allowing for detailed description. Both halves of the pectoral girdle are present but the left scapula is missing its distal end and the right coracoid is slightly damaged. The pectoral girdles of MPC-D 102/110.a, b are probably complete, but are mostly obscured by the overlying bodies. The pectoral girdles of MPC-D 102/11 and MPC-D 102/12 are unknown.

Scapula

The scapula is long and gracile (Fig. 20A–E). The distal end is slightly expanded and has a rounded end. In cross-section, the lateral surface of the scapula is flat, whereas the medial side is rounded, which produces a lens-shaped outline. The ventral edge of the scapula is sharp, but the dorsal edge is rounded. The scapular blade thickens transversely and curves medially towards the glenoid. About 30 mm distal to the glenoid, there is a small protrusion on the ventral edge of the blade that may have anchored musculature. Just anterior to this, on the dorsal edge of the blade, there is a shallow groove. The acromion process is small and rounded in dorsal view (Fig. 20E). Its dorsal surface is flat, but this flattened area does not extend far posteriorly, and although it probably contacted the furcula, there is no distinct area marking its articulation. The lateral edge is dorsally upturned and has a rounded, thickened edge. The anterior edge is thick and barely protrudes from the region where it connects medially to the body of the scapula. The acromion does not extend anteriorly past the contact of the scapula and coracoid. The glenoid of the scapula is approximately rectangular in articular view. Its lateral edge extends anterodorsally, whereas its medial edge is parallel to the scapular blade. As a result, the anterior part of the articular surface is exposed laterally (Fig. 20D). The articular surface is slightly concave and tapers transversely towards the posterior side. The unfused contact between the scapula and coracoid is crescentic (Fig. 20B). The anterior surface of the scapula is convex, whereas the posterior side of the coracoid is concave. Accordingly, the scapula has a relatively large, dorsal flange anterior to both the glenoid and acromion, which differs from other oviraptorids like Heyuannia yanshini (Funston et al. 2018), where the acromion is the most anterior part of the scapula, or caenagnathids where it is set posterior to the glenoid (Funston et al. 2021). On the medial side of the head of the scapula, there is a proximodistal groove that extends to the same level as the glenoid. This groove is continuous with a groove leading to the coracoid foramen, so it probably accommodated vasculature and nerves.

Coracoid

The coracoid is long dorsoventrally (Fig. 20B, D). The contact for the scapula is concave and tapers in transverse thickness dorsally. The glenoid is approximately square and faces completely posteriorly, with a slight lateral exposure. The coracoid foramen is oval and oriented with its long axis anteroventrally to posterodorsally. On the medial surface, it is connected to a deep groove that extends to the scapulacoracoid contact. The biceps’ tubercle is relatively large and circular. Its apex is rounded, rather than rugose, and there are no other ridges on the lateral surface of the coracoid. On the medial surface there are two fossae separated by a trabecula that correspond in position to the biceps tubercle (Fig. 20B). The body of the coracoid has two main processes: the posteroventral process and an anteriorly projecting flange—the acrocoracoid process. The latter process is rounded in profile and its apex is thickened. There is a concavity in the edge of the coracoid separating this process from the posteroventral process. This notch is shallower than a similar feature present in the coracoid of Heyuannia yanshini (MPC-D 100/30). The posteroventral process curves strongly posteriorly. It tapers in transverse thickness towards all edges and the apex, except that the apex itself is thickened and bulbous.

Furcula

The furcula is excellently preserved (Fig. 20F–J), and is missing only the very distal ends. The hypocleidium is long and pointed, but is relatively gracile, especially compared to the robust hypocleidium of Citipati osmolskae (Clark et al. 1999, Norell et al. 2018). In Tongtianlong limosus, the hypocleidium is small (Lü et al. 2016). The entire furcula is gracile and its curvature follows a rounded V-shape (i.e. the epicleidal processes are not parallel), compared to the U-shaped furcula of Tongtianlong limosus. Each epicleidal process expands transversely to its midpoint, and then tapers again distally. At the midpoint, there is a ventral facet where the furcula contacts and rests upon the acromion process of the scapula (Fig. 20G, H). In lateral view, this facet invades the lateral edge of the bone, which accommodates the upturned lateral edge of the acromion. In articulation, the hypocleidium of the furcula extends nearly to the acrocoracoid process of the coracoid, but a relatively large, lens-shaped triosseal fenestra remains.

Sternum

Both sternal plates are partly exposed in MPC-D 102/110.a, but not in the other individuals of MPC-D 102/110. The sternals are well preserved in MPC-D 100/33 (Fig. 20K, L), but were difficult to observe because they were mounted behind glass at the time of observation. The sternal plates are not fused along the midline and their posterior ends are separated. The sternocoracoidal process and lateral trabecula are both well developed, and are separated by an incised notch where the ventral ribs articulated. However, there are no distinct articular facets for the ventral ribs in this notch. Whereas the sternocoracoidal process is pointed in MPC-D 102/110.a, it is rounded and bulbous in MPC-D 100/33, possibly the result of older age and increased ossification. Sternal plates are described and illustrated for Jiangxisaurus ganzhouensis as oval plates, which would be unusual for an oviraptorid. However, this morphology does not match the photographs provided in the article, where the sternal plates appear more similar to other oviraptorids, with well-developed sternocoracoidal processes and lateral trabeculae. Unfortunately, no further comparison with Oksoko is possible based on the limited information available. In Oksoko, at least one foramen consistently pierces the sternal plate, but its position varies. In MPC-D 100/33, it is closer to the sternocoracoidal process (Fig. 20K, L), whereas in MPC-D 102/110.a, it is further medially, near the midline, and consists of two foramina. The right sternal of MPC-D 100/33 has a large fenestra near the centre of the plate (Fig. 20K). It is possible that this is pathological, but it could also be the result of variable ossification of the plates. Unfortunately, the detailed examination necessary to support these hypotheses was not possible.

Forelimb

The right forelimb of MPC-D 100/33 is completely preserved (Fig. 21), although it appears to be either missing phalanx III-1 or this element had not ossified. The left forelimb is represented by the humerus, ulna, and radius, but the carpals, metacarpals and ungual II-3 are missing. The right humerus, ulna, radius, and manus of MPC-D 102/110.a are exposed (Fig. 21), as is the left manus. Only the left ulna, radius, and manus of MPC-D 102/110.b are visible, although it is likely that the right forelimb is preserved under the body of MPC-D 102/110.a. The quarry of MPC-D 102/12 was revisited in 2018 and a manual ungual I-2 was recovered, but otherwise the forelimb of that individual is unknown.

Humerus

Both humeri of MPC-D 100/33 are well preserved and identical in size and shape. The humeral head is modestly developed (Fig. 21A, C), but does protrude slightly from the shaft. It is anteroposteriorly thin and appears more like a crest than a condyle. The proximal end is roughly parallelogram-shaped in proximal view (Fig 21E). On the posterior side of the humerus, the articular surface overhangs the surface of the rest of the bone. The deltopectoral crest extends distally from the lateral side of the head, which is anteriorly deflected. The crest thickens towards its apex, which is not downturned like in Heyuannia yanshini (MPC-D 100/30). The edge of the crest is rounded and slightly rugose on either side. The apex of the crest is just under half the length of the humerus (47%) from the proximal end, similar to Jiangxisaurus ganzhouensis and Tongtianlong limosus (Wei et al. 2013), but much greater than in citipatiines like Corythoraptor jacobsi, Huanansaurus ganzhouensis, and the Dzamyn Khondt oviraptorid. The anterior surface of the crest is concave, whereas the posterior surface has a plateau with a slightly depressed surface. The ridge outlining its anterior side has faint striations for muscle attachment, but there is no rugose mound like the one in Heyuannia yanshini (MPC-D 100/30). The depression is slightly triangular, tapering distally. The shaft of the humerus is almost perfectly cylindrical, but the anterior face is slightly flattened. There are no ridges or features on the shaft, which has less torsion and is more gracile than that of Heyuannia yanshini and Tongtianlong limosus (Lü et al. 2016, Funston et al. 2018). The distal end is about as wide as the head and is roughly rectangular in distal view. The entepicondylar tuber is larger than the ectepicondylar tuber (Fig. 21A, F), but both are small compared to the large, anteriorly curving ones of Heyuannia yanshini (Funston et al. 2018). In MPC-D 100/33, the entepicondylar tuber is dorsally hooked but does not protrude more than the ectepicondylar tuber, which itself extends proximally as a rounded ridge. In Heyuannia yanshini, this ridge is large and extends far anteriorly to become wing-like. In MPC-D 100/33, the medial side of the condyle is swollen and larger than the lateral side; the opposite is true in Heyuannia yanshini.

Ulna

The ulnae and radii of MPC-D 100/33 and MPC-D 100/110.a, b (Fig. 21G, H) are preserved, but those of MPC-D 100/33 were mounted and unavailable for detailed examination. Like other heyuannines, the antebrachium is shorter than the humerus, whereas in citipatiines, it is longer. The ulna is robust, expanding towards both the proximal and distal ends. The proximal end has a tall, bulbous coronoid process but a poorly developed olecranon, so that the socket for the humerus is poorly pronounced. The shaft tapers in dorsoventral thickness to the distal end, where it instead becomes transversely broad. The distal end of the ulna is crescentic in outline, with a distinct medial process, similar to the one in Heyuannia yanshini (MPC-D 100/30).

Radius

The radius is also robust (Fig. 21G, H), but is only half the thickness of the ulna throughout the shaft, comparatively more gracile than in Jiangxisaurus ganzhouensis. Its proximal end is square and about the same dimensions as the shaft. A slight ridge extends distally from the ventromedial edge, probably to accommodate the interosseum membrane. The distal end is expanded but does not appear to have a styloid process; however, this region is broken in MPC-D 100/33 and not visible in MPC-D 102/110.

Carpals

The carpals (Fig. 22) of the left hand of MPC-D 102/110.a are excellently preserved, and provide considerable information on the homology and development of the oviraptorid carpals. The radiale is the most proximal carpal (Fig. 22G, H), but it differs in shape from the angular, trapezoidal radiales of most theropods. Instead, it is more rounded and essentially featureless, although it is slightly wedged dorsally (Fig. 22H–M). The semilunate carpal (Fig. 22A–G) is the largest of the wrist and it covers the proximal ends of metacarpals I and II. It is roughly dumbbell-shaped, with a flat distal surface and a rounded proximal surface. Its proximal surface forms a distinct trochlea, with which the radiale and the crescentic distal end of the ulna articulate. The dorsal side of the trochlea is slightly smaller than the ventral side, but both are semicircular in lateral view. The flat distal side of the semilunate carpal is divided into two distinct faces separated by a shallow ridge (Fig. 22F). The medial face would have articulated with metacarpal I, although it did not overlie its entire proximal surface (Fig. 22B, D, E). The lateral facet for metacarpal II is concave, and in this depression sit two miniscule carpals, which are closely appressed if not fused (Fig. 22F, G). The larger of these is roughly triangular, and the smaller one is spherical. These carpals would have separated the proximal ends of the metacarpal I and metacarpal II in life (Fig. 22B). It is unclear whether these minute carpals are sesamoid bones, or if they represent the vestiges of the intermedium and ulnare, which typically lie lateral to the semilunate carpal and cover the proximal ends of metacarpals II and III. In the latter case, the larger element would more likely be the intermedium, whereas the smaller element would be the ulnare (Fig. 22F, G). However, this would necessitate a reorganization of the carpal region, as these carpals are typically proximal to the semilunate carpal, rather than distal. Thus, the interpretation that they represent sesamoid ossicles is preferred here. Further work and disarticulation of the carpal regions of other well-preserved oviraptorosaurs may clarify the homologies of these extra bones. Previous work has suggested that one or more of these carpals are missing in oviraptorids (Osmólska et al. 2004, Balanoff and Norell 2012) but are present in caenagnathids (Zanno and Sampson 2005). The wrist of Oksoko is apparently similar to that of Khaan in having only two carpal bones (Balanoff and Norell 2012), although this would indicate multiple losses of the ulnare and intermedium in oviraptorids, as other oviraptorids more closely related to Oksoko, like Heyuannia spp. (Lü et al. 2005; MPC-D 100/30), have three.

Manus

The manus of MPC-D 100/33 and MPC-D 102/110 are well preserved. The left hand of the latter specimen was disarticulated and provides detailed information on the elements (Fig. 23). As in other heyuannines (e.g. Heyuannia, Jiangxisaurus, and Nemegtomaia), the first digit is the most robust, whereas digits II and III are much more gracile. Metacarpal I is roughly rectangular (Fig. 23A–F). The proximal end is kidney-shaped in proximal view (Fig. 23E), with a convex medial side and a concave lateral side. It is inclined so that the medial side reaches further proximally. The lateral side of the metacarpal has a concavity (Fig. 23E), which is deeper proximally, to accommodate metacarpal II. The edge of this concavity prevents metacarpal II from reaching the ventral surface of the metacarpus in life. In distal view, the distal end of the metacarpal is roughly rectangular, but with a deep notch in its medial side. The lateral condyle is larger than the medial one, and both are transversely constricted about halfway up their height (Fig. 23F). The condyles are only weakly ginglymoid and almost straight in mediolateral view. Manipulation of phalanx I-1 with the condyles of metacarpal I results in a restricted range of motion when the condyles are kept in full contact. Phalanx I-1 is the largest of the hand (Fig. 23T) and exceeds metacarpal I in length. Its proximal articular surface is relatively flat, rather than deeply excavated, which contrasts with most theropod manual phalanges. In dorsal view (Fig. 23B), the shaft curves slightly medially. The collateral ligament pits are relatively shallow but the medial one is deeper, and the medial condyle is larger than the lateral one. The ungual I-2 is strongly recurved and has a well-developed flexor tubercle (Fig. 23A, D). The proximal articular surface lacks a proximodorsal lip, unlike Huanansaurus ganzhouensis and other citipatiines, and there is no groove between it and the rounded flexor tubercle. The vascular grooves are shallow and the lateral one is positioned further dorsally.

Metacarpal II is the longest of the hand (Fig. 23T), but is about half the transverse width of metacarpal I. Its proximal end is strongly compressed mediolaterally, and sits entirely within the concavity on metacarpal I. The shaft is straight and cylindrical, lacking any ridges or grooves. The medial condyle is slightly larger than the lateral one (Fig. 23L), but this disparity is not as great as in metacarpal I. When articulated with the first metacarpal, metacarpal II is deflected laterally (Fig. 23T). Phalanx II-1 is small, about half the length of phalanx I-1, but subequal in length to II-2. It is transversely compressed and minimally ginglymoid. The collateral ligament pits are shallow. Phalanx II-2 is more gracile than phalanx II-1, but overall similar in shape and size. It lacks the lateral groove on the distal end described for Jiangxisaurus (Wei et al. 2013). Ungual II-3 is relatively straight and has a poorly developed flexor tubercle. Like ungual I-2, it lacks a proximodorsal lip, but it has a more poorly developed proximal articular surface. The flexor tubercle is small and just dorsal to it there is a foramen on the lateral side.

Metacarpal III is diminutive and unusual in morphology (Fig. 23M). Its proximal end is tongue-like and deflected laterally. In articulation, it does not reach the carpus (Fig. 23T). The shaft is transversely compressed and less than half the transverse diameter of metacarpal II. The distal end is unusual for an oviraptorosaur, and indeed any theropod. The condyle is bulbous and spherical, rather than being divided into a true trochlea, and is overhung by a dorsal process (Fig. 23M). In articulation, this restricts the mobility of phalanx III-1 to mild flexion. Phalanx III-1 is also unusual. It is exceptionally small, less than 1 cm in length, and has poorly developed articular surfaces (Fig. 23M–S). Whereas the proximal articular surface is conventional, the distal end is blunted and transversely convex (Fig. 23S). As a result, there is no distal articular surface, which starkly contrasts with the condition in all other oviraptorids. This suggests that digit III of the manus was comprised only of the metacarpal and a single phalanx. This is supported by the absence of any more distal phalanges in all three hands visible in MPC-D 102/110, despite preparation from fresh matrix and the preservation of delicate elements like sclerotic plates. The combined length of metacarpal III and phalanx III-1 in articulation does not exceed the length of metacarpal II (Fig. 23T), so it is unlikely that the third digit would have protruded from the manus in life.

Pelvic girdle

The pubes, ischia, and some parts of the ilium are visible on MPC-D 102/110, but are best seen in MPC-D 102.11.a, where they are exquisitely preserved (Fig. 24). All six bones of the pelvis are complete in MPC-D 102/11.a (Figs 4, 24–26), but the right ischium is broken into two pieces. The pubes and ischia of MPC-D 100/33 are preserved, but the ilia are missing. The right ilium and left ischium of MPC-D 102/12 are known.

Ilium

The ilium (Fig. 24) is dolichoiliac and the pre-acetabular and postacetabular blades are nearly equal in length, although the postacetabular blade is slightly longer. The two ilia diverge posteriorly, and do not contact dorsally, as the neural spines of the sacrum extend past the dorsal margins of the ilia (Fig. 24A). The pre-acetabular blade has a rounded anterior margin, and is expanded anteroventrally anterior to the cuppedicus fossa. This anteroventral process is rounded, like in Corythoraptor jacobsi, and extends ventrally level with the dorsal margin of the acetabulum. The cuppedicus fossa is shallow, but posteriorly its medial border is demarcated by a sharp dorsal ridge. The pubic peduncle extends slightly further ventrally than the ischial peduncle, but it is equal in length anteroposteriorly to the transversely narrow ischial peduncle. The pubic peduncle has a flattened, ventrally facing surface where it meets the pubis, to which it was not fused or co-ossified in MPC-D 102/11.a or MPC-D 102/12. There is no supra-acetabular crest, but there is a bulge above the ischiadic peduncle, which probably represents a poorly developed antitrochanter. The ischiadic peduncle is triangular in lateral view and projects laterally past the lateral surface of the iliac blade. The brevis fossa is modestly developed and short, extending anteriorly about halfway as far as the base of the ischial peduncle. The brevis shelf of Oksoko avarsan is unique amongst oviraptorosaurs in that it is not continuous with the ischiadic peduncle (Fig. 24E, G). Instead, the brevis shelf is short and the postacetabular blade has an extra posterodorsally inclined ridge, separated from the brevis shelf by a groove (Fig. 24E, G). This unique morphology is clearly demonstrated in MPC-D 102/11.a and MPC-D 102/12, but is absent in all other oviraptorosaurs. The dorsal margin of the ilium is nearly flat from the pre-acetabular blade to the anterior margin of the brevis fossa, where it tapers dorsoventrally. The postacetabular blade is squared off posteriorly, as in Nemegtomaia barsboldi and Rinchenia mongoliensis, whereas it is more tapered in Corythoraptor jacobsi, Khaan mckennai, Nankangia jiangxiensis, and the Dzamyn Khondt oviraptorid.

Pubis

The pubis (Fig. 25) is strongly curved anteriorly, a feature shared with all other oviraptorids. When articulated with the rest of the pelvis (Fig. 3), the pubis extends anteriorly far past the anterior margin of the ilium. The iliac and ischiadic contacts of the pubis are widely separated by the rounded margin of the acetabulum. The iliac contact is long anteroposteriorly, with an anterior process, and wide transversely. The ischiadic contact is oriented vertically, and is tall dorsoventrally but very narrow transversely. It protrudes posteriorly from the shaft of the pubis, and is offset ventrally from the shaft by a square notch. Medial to the ischiadic contact, there is a shallow concavity that lacks the posterior circumscription of caenagnathids (Sullivan et al. 2011). The shafts of the pubes are separated by a transversely narrow pubic apron (Fig. 25A, B). Ventral to the pubic apron there is a long oval fenestra separating the shafts of the pubes before they converge again at the symphysis. Even in MPC-D 102/11.a, the pubic symphysis is fused, but there are grooves both dorsally and ventrally where the pubes meet and a wide anterior cleft separating the pubes. The pubic boot is longer anteriorly than posteriorly (Fig. 25C, F).

Ischium

The ischium is long, gracile, and concave posterodorsally (Fig. 26). The ischium is much longer in Oksoko avarsan than in most other oviraptorids, particularly contrasting with the short ischia of Corythoraptor jacobsi and Nankangia jiangxiensis (Lü et al. 2013b). There is a proximal groove that separates the pubic and iliac contacts, which represents the minimal involvement of the ischium in the acetabulum. The pubic contact is pitted and rugose, indicative of a cartilaginous element separating it from the pubis. The anterior margin of the obturator process is gently convex, curving towards the apex. This contrasts with the concave anterior edges of the ischium in caenagnathids. The obturator process is more than halfway down the shaft of the ischium, and forms a square point. The obturator process is thin and delicate, and is broken in both MPC-D 102/11.a and MPC-D 102/12, although it is completely represented in the former. The ischia of both individuals of MPC-D 102/110 are excellently preserved, but they differ slightly in morphology. Whereas the shapes of the ischia of MPC-D 102/110.a are identical to those of MPC-D 102/11.a, MPC-D 102/12, and MPC-D 100/33, those of MPC-D 102/110.b differ. This individual has a deep notch in the ventral edge of the ischium (Fig. 26B), separating the obturator from the distal end. Although this is incipiently developed in the other specimens, in MPC-D 102/110.b it is about three times as deep. It is important to note that this individual also has dimorphic chevrons, and therefore these elements may differ for the same reasons.

Hindlimb

Hindlimb elements are known for all individuals except MPC-D 102/11.b (although these bones may represent the cranium of MPC-D 102/110.c). All of the hindlimb elements are exquisitely preserved, with minimal crushing, a high degree of articulation, and pristine surface textures. The distal right hindlimb of MPC-D 102/11.a was left articulated and serves as an excellent reference for the in vivo positions of the bones, whereas the left hindlimb was disarticulated and shows the morphology of the bones more clearly. The hindlimb elements of MPC-D 102/12 were likewise disarticulated and so show the conditions in a skeletally mature individual.

Femur

The femora are complete in MPC-D 102/110, but are not fully exposed. Both femora are preserved in MPC-D 102/11.a, but the right femur (Fig. 27A–E) is more complete than the left, which is represented by only the proximal half. The right femur lacks the medial side of the distal end, but the lateral condyle is present, so length can be estimated. The femur of MPC-D 102/12 is complete but badly damaged and comminuted (Fig. 27F–K). The femora of MPC-D 100/33 are both well preserved but could not be observed because they were mounted. The femoral head is directed medially and has only a slightly constricted neck. The anterior face of the head is continuous with the neck, but the posterior edge projects posteriorly past the surface of the neck. In medial view (Fig. 27C, H), the posterodorsal corner of the femoral head is depressed and the anterodorsal corner is more bulbous. There is no rugosity for the capitate ligament, although this area is damaged in the larger MPC-D 102/12. The greater trochanter is broadly curved, but does not extend far above the neck of the head. It does not form a crest, but rather a rounded mound. The lesser trochanter is narrow and fingerlike, appressed to the anterior surface of the greater trochanter throughout its length (Fig. 27A, F). However, there is a small cleft between these structures proximally, which continues into a short groove distally. The shaft of the femur is cylindrical and curved anteriorly. It lacks a fourth trochanter and instead there is a posteromedially located patch of rugose bone for m. caudofemoralis. Distal and lateral to this, there is a dorsolateral to ventromedially inclined muscle scar just above the popliteal fossa. The lateral surface of the femur has no obvious muscle scars, but there is a slight mound just ventral to the greater trochanter, which continues distally as a posterolateral ridge. The anterior surface of the shaft has a long muscle scar that extends distally from the lateral groove of the lesser trochanter to just distal to the level of the insertion of m. caudofemoralis. This scar twists from the lateral side of the shaft to the medial side. On the distolateral part of the anterior surface, there is a pronounced rugosity with a mounded border. The popliteal fossa is very deep compared to most oviraptorids (Fig. 27I), but is not overhung by the crista tibiofibularis, which is the case in Rinchenia mongoliensis (Funston et al. 2018). The crista tibiofibularis is divided by a deep notch, separating the more bulbous fibular condyle from the larger tibial condyle. The ectocondylar tuber is mounded and rugose, and appears to become larger through ontogeny.

Tibia

Like the femora, the tibiae of MPC-D 102/110 are intact but not completely visible. The left tibia of MPC-D 102/11.a (Fig. 28J, K) is missing its proximal end, but the right tibia is completely preserved in articulation with the fibula, tarsals, and complete foot. The right tibia and fused astragalocalcaneum are preserved with MPC-D 102/12 (Fig. 28), and both tibiae were recovered for MPC-D 100/33, but were not available for examination. The cnemial crest is proximodistally short but is relatively well pronounced. It is only slightly everted laterally, and its apex is at its ventral end, rather than the dorsal edge as in ornithomimids. The fibular condyle has two main lobes, separated by a narrow groove into which the fibula inserts. The posterior lobe is larger than the anterior one. These two lobes are separated from the posterior surface of the femoral condyle by a notch. In MPC-D 102/12, the fibular condyle and femoral condyle coalesce external to this notch, leaving a circular tunnel (Fig. 28B). The main portion of the femoral condyle is kidney-shaped in proximal view, and extends further proximally than the fibular condyle. The fibular crest is poorly defined but is thick and rugose, rather than platelike (Fig. 28A, D). Posterior to the fibular crest there is a shallow groove, but it is not continuous with the large nutrient foramen that opens dorsally. The shaft of the tibia has a flat anterior surface, but there is a slight ridge at the distal end of the shaft, near the ascending process of the astragalus. The posterior surface of the tibia is rounded but the apex of the curvature is more medially located. The result is that the anteromedial corner of the tibia is sharp, whereas the lateral corner is more rounded. There is no facet or groove for the fibula, instead it rests upon the rounded lateral corner of the tibia. The anterior side of the distal end is obscured by the overlying astragalocalcaneum in each specimen, but it is clear that the medial malleolus protrudes anteromedially to create a bowl into which the astragalus fits (Fig. 28C). The lateral malleolus is posteriorly deflected and has a modest postfibular flange that does not extend far proximally.

Fibula

Fibulae (Fig. 28E–G) from each specimen are preserved. Unlike conventional reconstructions, each of the articulated fibulae is oriented with the broadest portion of the head oriented transversely, rather than anteroposteriorly. The head is concavoconvex and crescentic in dorsal view, with a larger lateral portion than medial portion. The medial part of the head is fingerlike in proximal view, whereas the lateral side is bulbous, which results in a central groove extending to a fossa on the posterior face. The lateral edge of the fibula distal to the head is sharply attenuated to a ridge (Fig. 28G), and this continues distally to become the lateral edge of the shaft. Distal to the head, the shaft thickens and has a thick, rugose anteromedial ridge. This ridge lies adjacent to the fibular crest on the tibia, and likely accommodated the interosseum membrane. The remainder of the fibular shaft is slender and concavoconvex, with the concavity oriented towards the tibia. The distal end has a bulbous head and curves slightly posteriorly. It appears to be separated from the calcaneum in each specimen, although this may vary depending on the position of the leg.

Astragalocalcaneum

The astragalus (Fig. 28H, I) is obscured in MPC-D 102/110 and the right foot of MPC-D 102/11.a by the overlying feet, but is exposed in MPC-D 100/33 and MPC-D 102/12. The medial condyle is much larger than the lateral one, and its medial surface is inclined to fit on to the medial malleolus of the tibia. The condyle is anteroposteriorly thin, which contrasts with the robust condyles of caenagnathids and some other oviraptorids. There is a concavity at the base of the ascending process that has a pronounced anterior lip (Fig. 28I). Distal to this, there is a fossa in the anterolateral part of the intercondylar space. On the distal surface of the astragalus, there is another depression in the intercondylar sulcus. The posterior edge of the astragalus is relatively straight, rather than curved. The lateral condyle has a sinuous anterolateral edge, which overhands the calcaneum dorsally but is excavated by it ventrally. The ascending process covers the entire surface of the tibia at its base, and extends at least 30% of the length of the tibia. The lateral edge of the ascending process is vertical, whereas the medial edge inclines proximolaterally to give the ascending process its taper. The calcaneum is unfused in the smaller specimens (MPC-D 102/11.a, MPC-D 102/110, and MPC-D 100/33), but it is fused in MPC-D 102/12 (Fig. 28H), which suggests that it fuses through ontogeny (see Discussion). The calcaneum is kidney-shaped, with the convex side facing anteriorly. Its lateral surface is concave, surrounded by a transversely thickened circumferential lip. The calcaneum is thicker at its anterior end than its posterior side.

Distal tarsals

Distal tarsals III and IV (Fig. 29) are preserved with each specimen. Distal tarsal III is roughly trapezoidal and, as in all oviraptorosaurs, thickens towards its posterior side. It covers the posterior half of metatarsal III in proximal view (Fig. 29A, B), but even in the mature MPC-D 102/12, it does not expand anteriorly. However, although it covers only metatarsal III in MPC-D 102/11.a and MPC-D 102/110, it has expanded medially in MPC-D 102/12 to cover the posterolateral corner of metatarsal II. In MPC-D 102/11.a, the posterior edge of distal tarsal III is rounded, whereas it becomes more square and much thicker through ontogeny in MPC-D 102/12 (Fig. 29A, B). In this individual, it has also begun to fuse to metatarsal III [Fig. 29C), which resembles the condition in some derived caenagnathids [Elmisaurus rarusOsmólska 1981 and Citipes elegans (Parks 1933)]. The medial side of distal tarsal III is rounded and bulbous. Although this is also the case for the lateral side in MPC-D 102/11.a, in the older MPC-D 102/12, the lateral edge is straight where it abuts—but does not fuse to—distal tarsal IV.

Distal tarsal IV is circular except for a rounded process extending from the lateral side (Fig. 29A, B). This process is probably homologous with the proximodorsal process of caenagnathids. Although oviraptorids generally lack a well-developed process here, the distal tarsals are typically poorly described, and so this feature may be more prevalent. The distal tarsal is disc-shaped and both sides are equal in thickness, but it tapers in thickness towards each edge. The lateral process is bulbous and thicker than the neck leading to it. In MPC-D 102/12, this process has become greatly enlarged and projects dorsolaterally (Fig. 29B, D), more closely resembling the proximodorsal process of caenagnathids. Furthermore, instead of remaining circular and disc-like, the fourth distal tarsal of MPC-D 102/12 is thickened posteriorly and has a straight medial edge where it meets distal tarsal III (Fig. 29B).

Pes

Both feet (Fig. 30) are preserved in their entirety in all specimens, except MPC-D 102/12, which preserves only the right metatarsus, metatarsal III from the left side, and a single phalanx III-1 from the right. All five metatarsals are represented in most specimens, but the feet of MPC-D 102/11.a appear to lack metatarsal V. It is possible that it was disarticulated during preparation and is represented by indeterminate splint-like bones accessioned with the specimen. Alternatively, metatarsal V may not have ossified yet in this individual.

Metatarsal I has a flat shaft and a small condyle (Fig. 30B, D, G). The shaft has a triangular proximal process and a tab-like posterior process. The lateral side is flat, whereas the medial side is rounded. The condyle is roughly triangular in distal view, with a narrower anterior side. The medial ligament pit is shallow and small, and the lateral one is large and deep. The posterior side of the condyles each have a small ridge, separated by a small depression. Phalanx I-1 is about the same length as the metatarsal and ungual. The proximal articular surface is inclined to face dorsomedially and is deeply concave. The shaft of the phalanx twists laterally and slightly dorsally. The condyle is narrow and the medial ligament pit is shallow but equal in size to the lateral one. Ungual I-2 is small and relatively straight, except for a slightly hooked tip. The proximal surface is crescentic and there is only a slight transverse constriction distal to it. The flexor tubercle is poorly developed. The medial and lateral grooves are poorly developed but the lateral one is deeper and slightly further dorsal.

Metatarsal II (Fig. 30) is the shortest of the weight-bearing metatarsals, but has a large proximal end. The proximal end is trapezoidal in proximal view, with the wider side facing metatarsal III and inclined about 50° mediolaterally. The shaft tapers from the proximal end but is consistent in thickness throughout most of its length. It is thicker than the other metatarsals anterioposteriorly but equal to metatarsal IV in transverse breadth. There is an incipient posteromedial ridge, but it is not well developed in the smaller specimens. In MPC-D 102/12, this ridge becomes larger and has a rugose surface. The proximal end of II-1 is inclined dorsomedially in proximal view. There are two ridges on either side of its ventral edge. The shaft of the phalanx is slightly curved laterally. The condyle is not ginglymoid and there is a large depression on the dorsal surface. The medial ligament pit is shallow, but the lateral one is deep. Phalanx II-2 is relatively symmetrical, but the proximal end is slightly skewed laterally. The phalanx is about half the length of phalanx II-1. Ungual II-3 is the largest ungual of the foot, but is only slightly larger than ungual III-4. The former is dorsoventrally deeper but slightly shorter in length. The flexor tubercle is weak and the claw is modestly recurved.

Metatarsal III is the longest of the foot (Fig. 30B, G) and the widest at its distal end. Its proximal end is wider than the other two metatarsals posteriorly, but it is anteriorly pinched—albeit not to the same degree as in caenagnathids. The proximal end is, therefore, triangular in proximal view, with a flat posterior edge. There is a flattened shelf on the posterior surface of the head (Fig. 30B, H) reminiscent of the posterior protuberance of Elmisaurus rarus (Osmólska 1981, Currie et al. 2016, Funston et al. 2021), but much smaller. This raised area is continuous with similar platforms on metatarsals II and IV. The shaft of metatarsal III is square in cross-section, with sharp posterior corners and flat sides. The distal condyle is asymmetrical, with a larger medial condyle than lateral condyle. The postcondylar ridges are well developed but end abruptly, rather than continuing proximally. Digit III is the longest and widest (Fig. 30B, G). Phalanx III-1 has a semicircular proximal face with two poorly developed ventral ridges. The shaft and condyle are symmetrical. The condyle is more ginglymoid than the phalanges of digit II and the collateral ligament pits are equal in depth. The more distal phalanges are virtually identical to phalanx III-1 in morphology, but they are each about 30% shorter than the previous one. Ungual III-4 is the longest of the foot but is more gracile than II-3. It is nearly perfectly symmetrical, including equally deep vascular grooves positioned equally far dorsally. The flexor tubercle is small, but larger than that of ungual II-3, and the claw is slightly more recurved than the latter.

Metatarsal IV (Fig. 30) has a large, semicircular proximal end in proximal view, with the flat edge against metatarsal III. The raised posterior area is triangular with a 45° inclination to the dorsal edge. The shaft of the metatarsal is compressed anteroposteriorly so that it is wider than deep. There is no sign of a posterolateral ridge in the smaller specimens, but in MPC-D 102/12 there is a rugose patch on the posteromedial side. The condyle is not deflected laterally and the lateral condylar ridge is small. The lateral ligament pit is relatively shallow. Digit IV is about equal in length to digit II, including the unguals. Phalanx IV-1 is wider distally than proximally. It has a deeply concave, triangular proximal end. The shaft is directed slightly medially. The other phalanges are short, with barely any shaft separating the proximal and distal condyles. These phalanges are symmetrical except for a slight lateral skew to the proximal articular surfaces. Ungual IV-5 is small and straight with a weak flexor tubercle and greater transverse constriction than the other unguals. The medial vascular groove is deeper and more dorsally positioned.

Metatarsal V is missing in MPC-D 102/11.a (Fig. 30), either as a result of preparation or poor ossification, and in MPC-D 102/12, probably because it was lost before collection. In the other specimens, it is a tapering splint tightly appressed to the posterolateral surface of metatarsal IV. Its proximal end is expanded and rounded. The shaft curves slightly anteriorly, but not to the same degree as in caenagnathids like Chirostenotes pergracilis (Gilmore 1924, Funston 2020). The distal end has a small bulb at its apex, but is otherwise simple. Its absence in MPC-D 102/12 suggests that it never fused throughout ontogeny, unlike in derived caenagnathids (Elmisaurus rarus and Citipes elegans), where it fuses to the proximodorsal process of distal tarsal IV (Currie et al. 2016, Funston et al. 2016a, 2021).

Discussion

New anatomical insights

The cranial anatomy of oviraptorosaurs, especially oviraptorids, has been well characterized. The skull of Oksoko avarsan fits well within the general schema of oviraptorids, and does not greatly change our understanding of its construction or variation, although it highlights several notable features. Most prominent of these is the highly pneumatized, domed crest formed of the nasals, frontals, and parietals. Oksoko avarsan is, so far, the only oviraptorid with a significant participation of the parietals in the cranial crest; although Lü et al. (2017) speculated that the parietal might have participated in the crest of Corythoraptor jacobsiLü et al. 2017, it is clear that the parietal contribution would be much less than in Oksoko avarsan. Furthermore, despite the much larger crest in Rinchenia mongoliensis, the contribution of the parietal to the crest is much smaller than in Oksoko avarsan. Other notable features of the skull of Oksoko avarsan include a postorbital with a vertically oriented frontal process, and a relatively robust jugal that contrasts with the typically rod-like jugals of most oviraptorids. Together with a well-developed postorbital ridge on the frontal, these bones combine to create a relatively robust column at the posterior margin of the orbit.

The forelimbs of Oksoko avarsan, on the other hand, provide several new insights into oviraptorosaur pectoral morphology. In particular, the excellent preservation of the carpal region in Oksoko avarsan (Fig. 22) provides new data on the carpus in oviraptorosaurs. Conventionally, four carpals have been identified as the ancestral arrangement in oviraptorosaurs, representing the semilunate (a fusion of distal carpals I and II), the radiale, the intermedium, and the ulnare (Osmólska et al. 2004, Zanno and Sampson 2005). At least some oviraptorosaurs are understood to have lost one or more carpals, most likely the ulnare and/or intermedium, resulting in two or three carpals in the wrist (Qiang et al. 1998, Osmólska et al. 2004, Qiu et al. 2019). The discovery of small ossifications distal to the semilunate carpal in Oksoko avarsan (Fig. 22F) raises questions about their homologies. Whereas it is possible that these minute ossicles represent vestiges of the distal carpals, this would challenge the consensus homology of the semilunate carpal as developing from distal carpals I and II, and would require co-option of other carpals into the semilunate, while retaining the general proximal–distal arrangement of the wrist. Instead, if they are not sesamoid bones, it is more likely that these are the homologues of the ulnare and intermedium (or even distal carpal III), but in a highly reduced state and unusual position. Discovery of a similar arrangement of the carpals in other closely-related oviraptorosaurs would provide more support for this hypothesis, which is presently difficult to test. Nevertheless, these may be sesamoid bones as originally described, although sesamoids are unknown in other theropod dinosaurs and it may be more parsimonious to interpret missing carpals as present, rather than inferring that two ossicles arose de novo. These parts of the carpal regions of the other individuals are not exposed, and so it cannot be determined whether these bones are consistently present in the carpus of Oksoko avarsan. Demonstration that these minute bones are consistently present among individuals would provide more support for their interpretation as the ulnare and intermedium, because sesamoid bones tend to be more variable (Vickaryous and Olson 2007).

The radiale (Fig. 22H–M) is retained but simplified in Oksoko avarsan, appearing closer to spherical than the triangular or trapezoidal radiales of other oviraptorosaurs. These simplifications of the carpal region may have been adaptations for enhanced flexibility, as they create a ball-and-saddle joint between the semilunate carpal and radiale that would allow considerable flexion of the wrist. Alternatively, they may have been the result of general reduction of the forelimb, indicating reduced reliance on its functionality. The latter hypothesis may also be supported by progressive reduction of the lateral digits in heyuannines (Funston et al. 2020) resulting in greater disparity of the digits and eventual loss of the third digit (as is the case in Oksoko avarsan; Fig. 23). However, most heyuannines retain and even expand the deltopectoral crest, ectepicondylar tuber, and medial process of the ulna (Fig. 21), indicating strong musculature of the forelimb, which conflicts with this hypothesis. It is possible that these changes reflect a change to a function of the forelimb less reliant on manipulation with the digits and more focused on powerful retraction of the arm. One possibility is scratch-digging, either for foraging or nest-building, which might explain the shorter, more robust digit I with a trenchant ungual. Detailed myological reconstructions and biomechanical analysis of the range of motion of the forelimb might provide more evidence on the matter.

The unusual, functionally didactyl hands of Oksoko avarsan are also worthy of note. As thoroughly discussed by Funston et al. (2020), they appear to reflect a progressive trend of digit reduction in heyuannine oviraptorids after their dispersal to the Gobi area. The metacarpal region provides an excellent example of the typical relationship between oviraptorid metacarpals, some features of which are frequently used in phylogenetic matrices. The lateral surface of metacarpal I is excavated by a shallow concavity, into which the proximal end of metacarpal II sits, covering it somewhat in flexor view. The proximal ends of these metacarpals are angled relative to their long axes (Fig. 23B, H), reflecting the angled articulation of the semilunate carpal and carpal region. The highly ginglymoid distal articular facet of metacarpal I (Fig. 23F) would have restricted phalanx I-1 to the flexor–extensor plane, whereas digit II was probably slightly more flexible based on the shallower articular groove in metacarpal II. Metacarpal III is much more gracile than most other oviraptorids, reflecting the reduction of the third digit, and its proximal end has a dorsal process (Fig. 23M) that effectively locks phalanx III-1 into place. Overall, the manus follows the typical heyuannine structure, but with emphasized disparity of digits I and III, and generally limited mobility between the metacarpals, and between the metacarpals and first phalanges.

Ontogeny

The osteohistological results of Funston et al. (2020) allow ontogenetically variable characters of Oksoko avarsan to be identified and tied to particular developmental stages. Most of the individuals known (MPC-D 102/110.a–c and MPC-D 102/11.a) are juveniles and, although histological sections for MPC-D 100/33 are not available, its intermediate size and degrees of vertebral fusion suggest an intermediate ontogenetic stage between the juveniles and the skeletally mature MPC-D 102/12. The zone of parallel-fibred bone surrounding the femoral cortex of MPC-D 102/12 suggests that this individual had slowed its growth considerably and can be considered skeletally mature. This supports fusion of neurocentral sutures on dorsal and cervical vertebrae and of the pygal vertebrae as ontogenetically progressing features. Expansion of the distal tarsals and their fusion to the proximal metatarsals may also be features of advanced age, but the absence of these characters in other large, presumably adult, oviraptorids with comparable degrees of fusion elsewhere in the skeleton may indicate that they are diagnostic of this taxon.

The juvenile specimens of Oksoko avarsan (MPC-D 102/11.a and MPC-D 102/110) show that cranial crests are developed early in ontogeny (Figs 6, 8). This contrasts with interpretations by previous authors that cranial crests may have developed later in life (Wang et al. 2016). In this regard, it is important to note that there is presently no evidence that the cranial crests of oviraptorids developed only partway through ontogeny or expressed positive allometry through growth. The former idea was initially proposed by Norell et al. (2001), who observed that small oviraptorid skulls tended not to have crests, whereas larger ones did. However, additional collecting and taxonomic work shows that these specimens represent different taxa, rather than ontogenetic stages of a single taxon. Indeed, small skulls with crests are now known (e.g. Banji long), as are large skulls without prominent crests (e.g. MPC-D 100/79-D, and Khaan mckennai). Four oviraptorid embryos are known that preserve sufficient parts of the skull (i.e. nasals) to evaluate the presence of a crest, and of these, two clearly lack crests (IVPP V20182 and YLSNHM01266; Wang et al. 2016, Xing et al. 2022), but two display inflated nasals comparable to adult crested oviraptorids (MPC-D 100/1018 and MPC-D 100/1019; Weishampel et al. 2008; Funston personal observation). Future work linking juvenile or embryonic oviraptorids to adults may reveal more information about the development of the crest, but at present there is no reason to expect that very young individuals lack features associated with the crest later in life. Likewise, positive allometry of the crests in oviraptorids has never been tested, and there is presently insufficient data to test allometry through the ontogeny of a single species—only Oksoko (N = 3) and Nemegtomaia (N = 2) have multiple individuals with crests preserved. Although the adult morphology of the crest is unknown in Oksoko avarsan, it is clear that it was prominently developed in young juvenile individuals, where it already shows distinctive features from other oviraptorid taxa. The substantial crest already incorporates the nasals, frontals, and parietals, which suggests that it either does not progressively incorporate a larger portion or more elements of the skull throughout ontogeny, or that this happens early in life (i.e. before 1 year of age and <50% of adult mass; Funston et al. 2020). This contrasts with crest development in cassowaries (Casuarius casuarius), which is more complex and gradual, and incorporates a greater number of bones than observed in any oviraptorid (Green 2020, Green and Gignac 2021, 2023). Nevertheless, like in cassowaries, helmeted guinea fowl (Numida meleagris), and maleo (Macrocephalon maleo), the crest clearly develops in Okosko avarsan prior to the attainment of sexual maturity, although it appears to do so more rapidly than these species.

Some changes in the forelimb of Oksoko avarsan can be noted between MPC-D 102/110 and MPC-D 100/33, although what ontogenetic interval this represents is uncertain. The humerus of MPC-D 100/33 is slightly longer relative to the femur than that of MPC-D 102/110.b and the deltopectoral crest is longer relative to the humerus. The radius and ulna are consistent in relative length between the specimens, however. The proportions of the manus remain the same, except that manual ungual I-2 is appears to be strongly positively allometric, increasing from 13% (MPC-D 102/110.a) through 17% (MPC-D 100/33) to 22% (MPC-D 102/12) of the length of the femur.

The excellent hindlimb material of Oksoko avarsan allows for quantification of ontogenetic trends in robustness and limb proportions (Table 1). MPC-D 102/11.a and MPC-D 102/12 are ontogenetic endpoints for this taxon, and their hindlimb proportions differ, both between the hindlimb segments and within regions the bones themselves (Table 1). The femur increases in relative minimum anteroposterior diameter throughout ontogeny, from 94% of the minimum mediolateral diameter in MPC-D 102/11.a to 108% of the minimum mediolateral diameter in MPC-D 102/12—an increase of 14%. The same is true for the tibia and metatarsus, although these relative increases are much smaller (5.5% and 3%, respectively; Table 1). In the femur and tibia, these changes are caused solely by an increase in minimum anteroposterior diameter through ontogeny, because minimum mediolateral diameter remains constant relative to the lengths of the femur and tibia (Table 1). However, this is not true of the metatarsus, which increases in relative minimum width throughout ontogeny (by 2.5% of its length). This suggests that the anteroposterior expansion of the metatarsus is stronger than initially suspected, roughly equal to the change in the tibia. The proportions of the hindlimb bones to each other change throughout ontogeny as well (Table 1). The tibia is reduced in relative length from 124% of the length of the femur to 117% the length of the femur, a reduction of 7%. The metatarsals also decrease in relation to the femur (59% to 56%), but not relative to the tibia. It appears, therefore, that the change in proportions of the metatarsus and tibia are more strongly linked to each other, and that the distal hindlimb segments change as a single unit relative to the femur throughout ontogeny. This may reflect constraints on limb morphology related to tail-driven locomotion (Benson and Choiniere 2013, Rhodes et al. 2020).

Other qualitative changes occur in the hindlimb throughout ontogeny, particularly relating to fusion of various elements. The astragalocalcaneum becomes more fused throughout ontogeny, although the onset of this fusion cannot be determined because the astragalocalcaneum of MPC-D 100/33 is still unfused. In MPC-D 102/12, however, this fusion is relatively extensive and these bones are separated by only a few small gaps. This might suggest that fusion of the astragalocalcaneum occurs rapidly towards the approach of adult body size, but the degree to which this can be generalized to other oviraptorosaurs is unclear. The astragalocalcaneum and the tibia remain unfused throughout life, contrasting with the condition in avimimids (Funston et al. 2016b, 2019) and some caenagnathids (CM 96523; TMP 1985.065.0001). The distal tarsals expand posteriorly and the lateral process of distal tarsal IV thickens proximodistally into a proximodorsal process. This process had begun to protrude in MPC-D 100/33, but it was not as well developed as in MPC-D 102/12. In spite of their expansion and eventually extensive contact, the distal tarsals do not seem to fuse to each other, unlike in Elmisaurus rarus (Currie et al. 2016) and Citipes elegans (Funston et al. 2016a). They do, however, begin to fuse to the proximal ends of the metatarsals, although even in MPC-D 102/12, a suture is visible between these bones. This condition has not been reported in any other oviraptorids, so it is unclear whether it is related to advanced age in MPC-D 102/12, or if it is a unique feature of this taxon. Considering that both parts of the ankle in MPC-D 102/12 show a much greater degree of fusion than is typical of other oviraptorosaurs, it is likely that this results from a single developmental process. The articular parts of the metatarsus increase in size throughout ontogeny, especially the distal condyle of metatarsal II. Muscle insertions on all of the hindlimb bones, especially the metatarsals, become more defined and rugose, but these apparently do not expand in size throughout ontogeny.

Overall, these results suggest that there is relatively little ontogenetic change in discrete characters in the appendicular skeletons of oviraptorids, besides states of fusion, but that their proportions change throughout life. In contrast to the findings of Lü et al. (2013a) that oviraptorid hindlimbs are isometric to body size, these ontogenetic series show that the hindlimbs do become relatively shorter and stockier throughout life. These contrasting results probably arise from different approaches: Lü et al. (2013a) assessed interspecific allometry, and patterns of allometry through ontogeny do not always mirror those across species. Whether the ontogenetic trends revealed here apply to other oviraptorosaurs as well will require a larger dataset of specimens of a single species [the ontogenetic series used by Lü et al. (2013a) have been dismantled by further taxonomic work]. Regardless, this negative ontogenetic trend is not as pronounced as carnivorous theropods (Currie 2003), and, as suggested by Lü et al. (2013a), supports other evidence of herbivory in oviraptorids (Funston et al. 2018, Meade and Ma 2022).

Functions of the cranial crest

The large cranial crests present in some oviraptorosaurs have generally been assumed to function as sexual or social displays (Lü et al. 2017, Ma et al. 2020a). Several aspects of crest morphology, development, and evolution, however, suggest that they may have served other functions instead or as well.

The presence of a well-developed crest early in ontogeny in Oksoko (Figs 6, 8) argues against its exclusive use as a sexual display feature, because such features are expected to be prominently developed only later in life as the individual matures (Hone et al. 2012, Knell et al. 2013, Green and Gignac 2021, 2023). Considering the presumably large energy investment necessitated by the crest, which modifies three bones, two of which form the braincase roof, it seems unlikely that this expenditure would be favoured early in ontogeny if the crest functioned solely for display in adult individuals. Nevertheless, the early development of a crest may also have enabled it to signal social status and fitness once the individual reached sexual maturity.

Also worthy of note, despite the discovery of numerous new crested oviraptorids, crest shape appears to be somewhat conservative, in the sense that, among the 10+ crested species, most crests follow one of two general shapes (Fig. 31), which emphasize different parts of the nasal bone. This contrasts with the wide variety of ornament shapes in ceratopsids and hadrosaurids, which are widely recognized as sexual display structures (Hone et al. 2012, Knapp et al. 2021). Triangular nasal crests, emphasizing the anterior midline process of the nasal and premaxilla, are known in at least three species (Citipati osmolskae, Nemegtomaia barsboldi, and the Dzamyn Khondt oviraptorid, and possibly Banji long; Fig. 31A). Tall, domed crests, emphasizing the lateral descending processes of the nasal, are likewise known in at least four taxa (Fig. 31B; Corythoraptor jacobsi, Oksoko avarsan, Rinchenia mongoliensis, and Tongtianlong limosusLü et al. 2016, and possibly Huanansaurus ganzhouensisLü et al. 2015 and Oviraptor philoceratops). Surprisingly, both of these morphotypes are represented by taxa in both subfamilies of Oviraptoridae (Citipatiinae and Heyuanninae), suggesting that they evolved independently, and domed crests are even known in Caenagnathidae (Anzu wyliei). Perplexingly, species exhibiting the same morphotypes also coexist in the same formations, which is unexpected if the crest served a role in species recognition, as proposed for cassowaries (Green 2020). The high degree of crest shape convergence is unusual compared to other dinosaur ornaments, which tend to exhibit high disparity (Hone et al. 2012, Knell et al. 2013). This suggests that the shape of the crest in oviraptorids is governed by different factors than other dinosaurs. In light of this, the extensive pneumaticity of the crests and skull may be important. The association of cranial crests with large pneumatic recesses is well documented (Clark et al. 2002, Balanoff and Norell 2012), although the nasals in oviraptorosaurs appear to be pneumatized regardless of whether a crest is present (Funston et al. 2018). In Oksoko, the crest is extensively pneumatized, with pneumatic recesses covering the nasal and extending on to the frontal as well. Balanoff and Norell (2012) described the connection of these nasal pneumatic recesses to the tympanic sinus through the frontal and parietal, an interpretation supported by the extensive (albeit crushed) vacuities along the dorsal portion of the crest in MPC-D 102/11.a. The connection of a sinus system extending from the nasals to the tympanic region is unique to oviraptorosaurs. Kundrát and Janácek (2007) suggested that the pneumatized tympanic region of Conchoraptor gracilis enhanced its ability to detect low-frequency sounds, which they interpreted as possible anti-predator or prey-detection adaptations.

Thus, in addition to, or perhaps instead of, a socio-sexual signalling role, these lines of evidence may point to the cranial crest of oviraptorids functioning in sound production or resonation, leading to extensive pneumatization and adaptations for detecting a broader range of frequencies. This function may better fit with the presence of the crests early in ontogeny, the frequent convergence of crest shapes, and their connection to the nasal and tympanic systems. Detailed segmentation of endocranial spaces and their connections to the nasal passages would be helpful in assessing the likelihood of this alternative hypothesis, and the effect of crest and pneumatic diverticula shape on sound production (Weishampel 1981), if they are indeed involved.

External indicators of maturity

Osteohistological samples of Oksoko avarsan were described in detail by Funston et al. (2020), and will not be revisited here. However, based on the maturity estimates presented in that study, the external signs of immaturity in MPC-D 102/11.a and MPC-D 102/110, which are estimated to be young juveniles on the basis of osteohistology, may provide useful milestones for assessing maturity in other oviraptorids (Table 2).

The low degree of cranial fusion in MPC-D 102/11.a probably reflects its young ontogenetic stage. The braincase bones of MPC-D 102/11.a are loosely attached and have disarticulated post-mortem. The frontals are separated along their contacts, rather than being tightly sutured, and the parietals and nasals each have shallow clefts separating them adjacent to their frontal contacts (Table 2). Nevertheless, the contralateral elements are mostly fused, suggesting that both the nasals and parietals are united early in development. In MPC-D 102/110, the frontals are more tightly adhered, and there are no clefts in the nasals and parietals. This suggests, unsurprisingly, that the degree of fusion of the braincase increases with size and/or maturity (Table 2). Furthermore, it suggests that in oviraptorids, these elements fuse early in life and are probably not suitable indicators of skeletal maturity.

The vertebrae of these specimens may provide better indicators of maturity. Despite being roughly 50% of adult body mass, each of the young individuals (MPC-D 102/11.a and MPC-D 102/110) lacks complete fusion of the sacral vertebrae, and only the three primordial sacral vertebrae are fused (Fig. 16). The sacrum of the slightly larger MPC-D 100/33 has a fused dorsosacral vertebra, but the two caudosacral vertebrae remain unfused (Fig. 16). This suggests that consolidation of the sacrum begins with the dorsosacral vertebra at about 50% of maximum body mass and proceeds to the caudosacral vertebrae later in ontogeny (Table 2). The caudal vertebrae show a similar pattern of progressive fusion (Table 2). The neurocentral sutures of all but the most distal 15 caudal vertebrae of MPC-D 102/11.a are open, whereas some of the more proximal caudal vertebrae of MPC-D 100/33 have begun to fuse the neurocentral suture (although their exact positions cannot be determined). This suggests that fusion of the proximal caudal vertebrae begins relatively early in life, certainly before the achievement of adult body size. In contrast, the dorsal and cervical vertebrae have completely unfused neurocentral sutures in all specimens, except MPC-D 102/12, which shows osteohistological signs of maturity. This suggests that neurocentral suture closure of the presacral vertebrae occurs later in ontogeny, beginning after at least 50% of maximum body mass has been attained (Table 2). This suggests that fusion of the presacral neurocentral sutures is a better indicator of skeletal maturity than sacral or caudal vertebrae, but also highlights uncertainty in the sequence and timing of neurocentral fusion in oviraptorosaurs. In this sense, while absence of neurocentral suture fusion is an ambiguous indicator of immaturity (an individual with at least some unfused sutures may be close to adult size), fusion of neurocentral sutures in the cervical and dorsal vertebrae may be a better indicator that the individual is in the later stages of ontogeny. The discovery and osteohistological sampling of subadult individuals of Oksoko avarsan might help to constrain when neurocentral sutures close.

Future directions

The continued discovery of excellent specimens like MPC-D 102/110 has offered a myriad of opportunities for understanding the anatomy and biology of oviraptorosaurs. Thus, more so than in many other dinosaur clades, datasets are now available to test a range of hypotheses, including functional morphology, evolutionary trends and patterns, intraspecific variation, ontogeny and skeletal growth, among many others. Many of these lines of inquiry have lagged behind the rapid pace of discovery of new species, but the first forays into these topics have now begun to be undertaken. For example, a number of recent studies have tested macroevolutionary and palaeoecological hypotheses about oviraptorosaurs (Ma et al. 2017, 2020a, b, Tanaka et al. 2018, Wiemann et al. 2018, Funston et al. 2020, Meade and Ma 2022, Yang and Sander 2022). These provide major advances in our understanding of the biology and evolution of oviraptorosaurs. Studies on functional morphology and its impact on oviraptorosaur ecology and evolution (Ma et al. 2017, 2020b, Meade and Ma 2022) are particularly insightful, and continued work in this vein may help to resolve the selective pressures that led to the unusual morphologies of oviraptorosaurs.

However, these kinds of studies require a solid taxonomic and phylogenetic foundation, which continues to be elusive. The taxonomic framework of oviraptorosaurs is notoriously unstable, particularly on the caenagnathid branch of the lineage. Nonetheless, even among the comparatively well-known oviraptorids, there is contention about which specimens pertain to which species, and which characters are useful for distinguishing species. For example, material initially lumped into ‘Ingenia’ yanshini (most recently referred to as Heyuannia yanshini) has now been pulled into other genera (MPC-D 100/33, Oksoko), resolving some conflicts about character states in the former taxon. Similar issues for other oviraptorids (e.g. differentiating Nemegtomaia, ‘Ingenia’, and Conchoraptor) stand to be resolved simply by re-examination of historic holotypes and referred specimens in the light of new anatomical knowledge of the family, and this would make for straightforward but fruitful work.

Addressing issues with the most popular phylogenetic matrix for oviraptorosaurs will be perhaps less straightforward. Most research groups have built iteratively off the matrix developed by Maryánska et al. (2002) and Osmólska et al. (2004), but studies progressing in parallel mean that different versions of the matrix (Lü et al. 2017, Yu et al. 2018, Qiu et al. 2019, Funston et al. 2020) have not yet been reconciled. More problematic still, differing interpretations of vague character descriptions mean that scoring practices are different, and also that new characters have been introduced that are redundant under other interpretations. Furthermore, many characters from the initial matrix have become less useful as new taxa have become known; these characters strongly unite the major clades of oviraptorosaurs (e.g. Caenagnathidae and Oviraptoridae), but have too little variation or character states to resolve finer relationships within these groups. This is exacerbated by high proportions of missing data within the matrix, even for characters that could be scored based on known material. As a particularly salient example, the analysis of Funston et al. (2020) recovered Oksoko in a clade with Banji and Jiangxisaurus, but there is no single character that can be scored for all three of these taxa that unites them to the exclusion of other taxa. Instead, Banji is drawn to Oksoko because they share two characters of the palate (vomer distant from parasphenoid and vomer level with other palatal elements), whereas Jiangxisaurus is drawn into the clade based a single character, subequal lengths of metacarpals II and III. Revising these taxa for characters added to the matrix since their initial descriptions would probably change their tenuous positions within the phylogeny. In a separate vein, taxa removed by some workers, like Beibeilong (embryos are inappropriate for phylogenetic analysis) and Ningyuansaurus (probably a deinonychosaur) are reintroduced by others (e.g. Wei et al. 2022), making comparison between results challenging. Finally, logistical issues like the practice of adding new characters and taxa to the end of the matrix (Longrich et al. 2013, Lamanna et al. 2014, Yu et al. 2018), rather than in more logical orders, have made it unwieldy to comprehensively revise the phylogenetic matrix and score new taxa into it. Overall, these and other issues have resulted in a problematic phylogeny where topology is volatile, newly recognized outgroups like scansoriopterygids (Pittman et al. 2020) are difficult to incorporate, and characters are poorly constructed for resolving relationships within the major clades. A revised phylogenetic matrix to address at least some of these issues is currently in progress (Funston in prep.), but continued work and cooperation will be necessary to incorporate new data as they arise.

A further issue is that representation of oviraptorosaurs in other popular phylogenetic matrices is poor, leading to ingroup topologies that are inconsistent with morphology. For example, iterations of the popular Theropod Working Group (TWiG) matrix frequently recover paraphyletic Caenagnathidae and Oviraptoridae, or early-branching forms intermixed with late-branching Late Cretaceous taxa (Hartman et al. 2019, Pittman et al. 2020, Cau and Madzia 2021). These issues stem not only from taxon selection within Oviraptorosauria, but also the inadequacy of characters derived for other theropods to distinguish ingroup relationships within the anatomically unusual Caenagnathoidea. In order to bring the results of broad-scale analyses up to our current understanding of oviraptorosaur relationships based on more focused datasets, further effort is needed to improve the sampling and scoring within these matrices.

Conclusions

The exquisite preservation of multiple individuals of Oksoko avarsan provides a detailed record of its anatomy and insights into ontogenetic variation in a model oviraptorid. Comparison to a suite of recently described oviraptorids highlights regions of high variability that may be useful for construction of future phylogenetic characters, and also valuable for identification and characterization of less complete material. Ontogenetic change in Oksoko avarsan is surprisingly minimal, particularly with respect to the cranial crest, whose development early in life argues against sexual display as the sole function. Instead, connection with the nasal passages and pneumatic sinuses within the skull points to a possible role in vocalization. Whereas fusion of compound skeletal elements increases through ontogeny, this does not proceed in a consistent posterior–anterior sequence. Limb bones change minimally in proportion, but become more robust with advancing age. Documentation of these data will hopefully help in future work revising oviraptorosaur taxonomy and phylogeny, which is needed to underpin the macroevolutionary and palaeobiological studies that have begun to be tackled by multiple research groups.

Acknowledgements

This work was supported by NSERC, the Vanier-Banting Commission, the Royal Society (Grant NIF\R1\191527), and the Dinosaur Research Institute. My thanks to P. Currie and F. Fanti for photographs and discussion; and to Chinzorig Tsogtbaatar and Tsogtbaatar Khishigjav for access to specimens.

DATA AVAILABILITY

All data used in this study is presented in the manuscript.

Conflict of interest

I have no conflicts of interest.

References