The endocranium and trophic ecology of Velociraptor mongoliensis

Abstract Neuroanatomical reconstructions of extinct animals have long been recognized as powerful proxies for palaeoecology, yet our understanding of the endocranial anatomy of dromaeosaur theropod dinosaurs is still incomplete. Here, we used X‐ray computed microtomography (µCT) to reconstruct and describe the endocranial anatomy, including the endosseous labyrinth of the inner ear, of the small‐bodied dromaeosaur, Velociraptor mongoliensis. The anatomy of the cranial endocast and ear were compared with non‐avian theropods, modern birds, and other extant archosaurs to establish trends in agility, balance, and hearing thresholds in order to reconstruct the trophic ecology of the taxon. Our results indicate that V. mongoliensis could detect a wide and high range of sound frequencies (2,368–3,965 Hz), was agile, and could likely track prey items with ease. When viewed in conjunction with fossils that suggest scavenging‐like behaviours in V. mongoliensis, a complex trophic ecology that mirrors modern predators becomes apparent. These data suggest that V. mongoliensis was an active predator that would likely scavenge depending on the age and health of the individual or during prolonged climatic events such as droughts.


| INTRODUC TI ON
Velociraptor mongoliensis Osborn, 1924 is a velociraptorine dromaeosaur found in Late Cretaceous formations of China and Mongolia (Osborn, 1924;Godefroit et al., 2008) that has been made famous in recent years thanks to its portrayal in numerous Hollywood movies. V. mongoliensis has also been the subject of a number of cranial and postcranial publications (Sues, 1977;Norell et al., 1997;2004;Barsbold and Osmólska, 1999;Turner et al., 2007;Manning et al., 2009), with the cranial osteology, including the braincase, being well-known thanks to the exceptionally preserved specimens found in Mongolia (Barsbold and Osmólska, 1999). Despite this heightened attention, the endocranial anatomy of V. mongoliensis has not yet been described.
Indeed, the endocranial anatomy of Dromaeosauridae as a whole is still relatively poorly known despite the initial osteological description of Dromaeosaurus having occurred almost a century ago (Matthew and Brown, 1922). Since that time, the endocast for Bambiraptor feinbergi (Burnham, 2004) has been partially described and portions of the endocast of Saurornitholestes langstoni, V. mongoliensis, Tsaagan mangas, and Deinonychus antirrhopus have been measured for quantitative analysis or otherwise imaged (Witmer and Ridgely, 2009;Zelenitsky et al., 2011;Balanoff et al., 2013). There is, however, a distinct lack of described endocasts with which to expand the palaeobiology of dromaeosaurs in the formal literature to date. Although a few publications have noted and discussed the implications of the large endocranial space in dromaeosaurs (Hopson, 1977;Currie, 1995;Norell et al., 2004), relating the endocranial anatomy of velociraptorine dromaeosaurs to their trophic ecology has yet to be done in any capacity.
With this in mind, we explored the neuroanatomy of V. mongoliensis (IGM 100/976) in order to better estimate the trophic ecology and sensory aptitude of this species-thus providing much needed sensory and behavioural data for dromaeosaurs. Here, we describe the anatomy of the hindbrain and inner ear of V. mongoliensis using cranial endocasts and compare its neuroanatomy to extant reptilian (including birds) taxa in order to place its sensory abilities into a broad palaeoecological context.

| ME THODS
IGM 100/976 was collected as a part of the 1991 Joint Expedition of the Mongolian Academy of Sciences and American Museum of Natural History. This specimen was recovered from the Djadokhta Formation at Tugrugeen Shireh, Mongolia (Norell et al., 1997) and consists of a partial skeleton, including an incomplete braincase that is missing the bones anterior to the basisphenoid and supraoccipitals (Figure 1a,b). The braincase is comprised of a few incomplete elements-the exoccipitals, supraoccipital, and basioccipital. These four elements are fused to form an incomplete adult endocranial space where the sutures are obliterated along the surface (Norell et al., 2004). Because of its incomplete nature, the endocast preserves the entire hindbrain but only a featureless portion of the midbrain. The mean and high hearing frequencies for IGM 100/976 were calculated following the method outlined in Walsh et al. (2009). To accomplish these reconstructions, we took measurements from the anterior-most extent of the basisphenoid to the posterior-most margin of the occipital condyle along with the length of the cochlear duct ( Table 1). The two measurements were then used to calculate a cochlear duct-basisphenoid ratio and then logarithmically transformed. This normalized value was placed into pre-calculated formulae found in Walsh et al. (2009).

| Cranial endocast
The identifiable regions of the brain preserved in the specimen are  (Table 1).
The medulla is wider than tall and forms an almost oval shape at the foramen magnum. As seen in most other maniraptorans, the medulla is antero-posteriorly short and narrower than the rest of the hindbrain (Kundrát, 2007;Balanoff et al., 2009;Lautenschlager et al., 2012) (Table 1). Anteriorly, the medulla exhibits a gentle dorsolateral constriction between it and the cerebellum. Anteriorly, there is a 132.94° angle between the hindbrain and midbrain. This pontine flexure (Table 1) implies that the brain exhibited a gentle curvature and was not all located along the same horizontal plane. This curvature is unsurprising due to its presence in many non-maniraptoran theropods (Sampson and Witmer, 2007;Witmer and Ridgely, 2009), basal therizinosaurs (Lautenschlager et al., 2012), and oviraptorosaurs (Kundrát, 2007;Balanoff et al., 2014).
As a whole, few anatomical structures are preserved on the endocast of the cerebellum. The hindbrain lacks a prominent dorsal dural peak overlying the cerebellum that is found in some other maniraptorans such as Conchoraptor (Kundrát, 2007) and large-bodied derived tyrannosaurs (Osborn, 1912;Witmer and Ridgely, 2009;Bever et al., 2011;Brusatte et al., 2016). The absence of a large dural peak is consistent with another velociraptorine dromaeosaur,  T. mangas (personal observation by the authors) and basal tyrannosaurs (Kundrát et al., 2018); however, it is possible that this portion of the endocast was not preserved.

| Cranial nerves and vasculature
Both Velociraptor, although these could not be reconstructed digitally. The occurrence of small veins in Velociraptor would not be surprising considering birds and their close relatives have a thin dural envelope and a majority of their braincase filled with neural tissue (Norell et al., 2004;Evans, 2005). Little of the venous architecture is preservedonly the posterior middle cerebral veins are observable along the posterodorsal surface of the cerebellum (Figure 1e).

| Endosseous labyrinth
Both endosseous labyrinths are preserved in IGM 100/976 (Figure 2), although the posterior portion of the left labyrinth, i.e. where the posterior semicircular canal meets the lateral canal, is not preserved (Figure 2f). In many respects, the vestibular anatomy of V. mongoliensis is similar to that of other non-avian theropods (Balanoff et al., 2009;Witmer and Ridgely, 2009;Lautenschlager et al., 2012).  The cochlear duct, which would have housed the basilar papilla, the receptor organ for hearing, is relatively long in V. mongoliensis compared with most non-avian theropods. It is also relatively wide and follows a similar anteriorly orientated course as the crus communis. The separation between the fenestra vestibuli and fenestra cochleae (oval and round windows, respectively) cannot be differentiated in this scan (Figure 2a,b,f). suggest that quick movements and a stable gaze were essential to its everyday life (Figure 1). This interpretation fits well with the current idea that V. mongoliensis was a nimble predator that relied heavily on its agility while pursuing and attacking prey. Moreover, as enlarged floccular lobes have also been proposed to be indicative of strong vestibulo-ocular (VOR) and vestibulocollic (VCR) reflexes (Hopson, 1977;Witmer and Ridgely, 2009), it can be reasoned that V. mongoliensis was able to track moving objects easily. With that being said, because the optic lobes were not preserved it is impossible to say at this point to what degree IGM 100/976 relied on sight rather than other senses. Based on the enlarged floccular lobes, elongated semicircular canals, and large orbit size of the species, it can be assumed that the visual acuity and field of view of V. mongoliensis was high (Stevens, 2006;Schmitz and Motani, 2011;Torres and Clarke, 2018).

| Sensory abilities of Velociraptor
This heightened optical sensitivity is not surprising considering the hypothesized predatory lifestyle of V. mongoliensis. When combined with its large flocculi and potentially sensitive VOR and VCR, it is likely that V. mongoliensis was easily able to track and pursue its prey smoothly based on its sensory neuroanatomy and stereoscopic vision (as determined from the position of the orbits).
In life, the endosseous cochlear duct of V. mongoliensis would have housed the basilar papilla-the auditory organ of tetrapods (Gleich et al., 2005;Walsh et al., 2009). As the length of the cochlear duct has been interpreted as a rough measurement of the basilar papilla, the length of the duct can be used as an estimator of hearing frequencies in non-avian dinosaurs (Witmer and Ridgely, 2009;Lautenschlager et al., 2012). Moreover, the relationship between the length of the cochlear duct and the basisphenoid has also been shown to correlate with hearing frequencies in modern archosaurs (Walsh et al., 2009), thus providing a way to calculate mean and high frequencies of non-avian dinosaurs. A recent study using extant turkeys demonstrated that a shape analysis of a single endosseous labyrinth can be used to represent an entire population (Cerio and Witmer, 2019); we therefore suggest that the hearing frequencies calculated in this study can be used as proxies for high and average hearing frequencies for V. mongoliensis. By measuring and logarithmically transforming the ratio between the cochlear duct length and the total length of the basisphenoid, a mean hearing range (2,368 Hz) and high-frequency hearing limit (3,965 Hz) was calculated for IGM 100/976-a range that is comparable to birds such as the common raven (Corvus corax) and the African penguin (Spheniscus demersus) (Walsh et al., 2009). Unsurprisingly, the scaled anteroposterior width and length of the cochlear duct were much more similar to birdsspecifically neognaths such as budgerigars (Melopsittacus undulatus), storks (Ciconia ciconia), and mute swans (Cygnus olor) -than to more basal archosaurs and other reptiles (Figure 3).
Our results indicate that V. mongoliensis could hear, hunt, and perhaps vocalize most efficiently in the range of 2,400 Hz. When compared with other maniraptorans for which data are available, the mean hearing frequency of V. mongoliensis is high. For instance, the mean and high hearing frequencies estimated for the basally diverging therizinosaur Falcarius utahensis are 1,630 Hz and 4,000 Hz, respectively (Lautenschlager et al., 2012). Similarly, the frequency range of the more derived therizinosaurid Erlikosaurus andrewsi has a high range between 1,600 and 4,000 Hz (Lautenschlager et al., 2012). Although the mean frequency range of V. mongoliensis is notably higher than that of therizinosaurs, the high-frequency values are roughly the same.
The elongate nature of the cochlear duct supports the hypothesis that V. mongoliensis was capable of detecting a wide range of sounds, indicating that hearing was likely an important sensory system in this taxon (Manley, 1990;Walsh et al., 2009;Witmer and Ridgely, 2009;Brusatte et al., 2016;Carabajal et al., 2016).
In fact, IGM 100/976 plots closest to the social vocal learner Melopsittacus undulatus (budgerigar), making it feasible that V. mongoliensis utilized hearing in social interactions as well as active predation (Figure 3).

| The trophic ecology of Velociraptor
Dinosaur feeding styles, such as predation vs. scavenging, have been a topic of popular interest in the past but remain difficult to diagnose in the fossil record (Holtz, 2008). Our current understanding of the trophic ecology of V. mongoliensis is provided by several sources. The most famous of these, the 'fighting dinosaurs' (IGM 100/25) of Mongolia, preserves what has been interpreted by some palaeontologists as a predation attempt of a Velociraptor on a Protoceratops (Carpenter, 1998). However, further evidence has emerged in recent years suggesting that

V. mongoliensis was not an obligate predator. This includes
Velociraptor tooth marks on bones, which have been interpreted as late stage scavenging, and the preserved gut contents of a subadult individual (Hone et al., 2010;. Each of these specimens indicate that scavenging was a part of the trophic ecology of V. mongoliensis. The neuroanatomical results described in this study help flesh out the degree to which scavenging contributed to the diet of V. mongoliensis.
Our current understanding of the neuroanatomy of V. mongoliensis suggests that predation likely made up a large part of its diet.
As with therizinosaurs and oviraptorosaurs, IGM 100/976 possesses relatively enlarged flocculi (Figure 1c,e;Lautenschlager et al., 2012;Balanoff et al., 2018). However, the flocculi of IGM 100/976 surpass those of observed therizinosaurs and almost completely fill the interior space between the semicircular canals. Enlarged flocculi have been used to predict prey tracking capabilities and may imply that the species had an acute vestibulo-ocular reflex (Walsh et al., 2013).
This evidence in conjunction with the wide field of binocular vision, extended hearing range (as determined from this study), and skeletal morphology indicates that rather than being equally or more reliant on scavenging, V. mongoliensis was well-equipped to be an active predator.
The fossil record, however, indicates that scavenging was at least a small part of the diet in V. mongoliensis (Hone et al., 2010(Hone et al., , 2012. Opportunistic scavenging is supported by gut contents recovered from MPC-D100/54 that include a 75-mm-long bone of an unidentified pterosaur. Whether this represents an act of osteophagy or scavenging due to an injury or its small size, it is probable that the pterosaur was dead prior to being eaten, given its incomplete nature. In the case of IVPP V16137, a probable Protoceratops, multiple bone fragments including a dentary, exhibited bite marks characteristic of velociraptorine dromaeosaurs. Velociraptorine teeth (IVPP V16138) were also found in association with IVPP V16137, further indicating that a velociraptorine dromaeosaur was feeding on the carcass of IVPP V16137. Some of these tooth drag marks found along the anterior portion of the dentary suggest that this was an instance of late-stage scavenging by V. mongoliensis due to the lack of significant muscle mass located along a dentary during life. While this evidence for scavenging can be interpreted as being somewhat circumstantial, we accept that the specimens nevertheless show enough evidence to be considered acts of scavenging rather than active predation based on the conclusions of previous studies (Hone et al., 2010;. This type of flexible hunting strategy is not surprising given that modern predator diets are a spectrum rather than an 'either/ or' scenario in which seasonality, fitness, and other ecological constraints are the primary drivers (Mattisson et al., 2016) birds such as Aquila chrysaetos, often resort to changes in hunting behaviour, or even scavenging, when prolonged weather patterns, injury or ontogenetic stage forces them to find alternative food sources (Tjernberg, 1981;Marchetti and Price, 1989;Wilmers et al., 2003;Mattisson et al., 2016). It follows that the neuroanatomy of V. mongoliensis suggests a behaviour that is adapted for active predation (Carpenter, 1998); however, young or injured individuals and those experiencing dietary constraints brought on by local climate would have actively sought out carcasses for an easy meal.

| CON CLUS IONS
The neuroanatomy and sensory capabilities of V. mongoliensis are described here for the first time and are subsequently used to put this taxon into a larger palaeobiological context. Evidence from the hindbrain and labyrinth of IGM 100/976 suggests that V. mongoliensis had an average hearing range near 2,400 Hz (similar to modern social birds such as ravens and penguins), highly sensitive vestibulo-ocular and vestibulocollic reflexes, and a fine-tuned sense of balance-all of which would have been advantageous as an active predator. Although previous studies have used gut contents and tooth marks on Protoceratops mandibles to provide evidence that scavenging was a part of V. mongoliensis trophic ecology, most evidence, including the neuroanatomy, suggests an active predatory lifestyle. Therefore, we interpret the presence of scavenging as a facet of the trophic ecology for V. mongoliensis. Based on the behaviour of modern bird taxa, our better understanding of velociraptorine senses, the apparent case of a predation event in the 'fighting dinosaurs', and the age/health/environment of the scavenging individuals, it is likely that V. mongoliensis was an active predator that would readily rely on carrion in the event that a ready source of prey items was not available.

ACK N OWLED G EM ENTS
We thank Mark Norell (AMNH) for granting access to the specimen and Matthew Colbert (University of Texas at Austin) for conducting the scanning. Additionally, we would like to thank Steve Brusatte

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

AUTH O R CO NTR I B UTI O N S
J.L.K. and J.M.N conceived the study. J.L.K reconstructed the 3D brain endocast, performed the sensory calculations, and drafted/ revised the manuscript. J.M.N reconstructed and described the labyrinths and assisted in writing the manuscript. J.S.S. and J.A.G.
provided the CT data, gave anatomical considerations for the project, and edited the initial drafts of the manuscript. A.M.B. provided expert knowledge and edited the final drafts of the paper.

DATA AVA I L A B I L I T Y S TAT E M E N T
The 3D models produced in this study are openly available from