Endocranial morphology of three early-diverging ceratopsians and implications for the behavior and the evolution of the endocast in ceratopsians

Abstract

Ceratopsian dinosaurs underwent great changes, including a shift of locomotion mode, enlarged horns and frills, and increased body size. These changes occur alongside the evolution of endocranial morphology and physiology such as the size and shape of the flocculus, hearing range, olfactory ratio, and the reptile encephalization quotient (REQ). However, the evolution of endocranial structures in early ceratopsians is still unclear because of a lack of information on the earliest ceratopsians. Here, we reconstructed the endocasts of three early-diverging ceratopsians including the Late Jurassic Yinlong, and the Early Cretaceous Liaoceratops and Psittacosaurus. These ceratopsians display obvious flocculi, large and separate olfactory bulbs, long and high anterior semicircular canals, and relatively long cochlear ducts. In the evolution of the earliest ceratopsians to early neoceratopsians, changes include the increasing size of the flocculus (which is reduced or absent in late-diverging ceratopsids), the attenuation of the semicircular canals, and the heightening of the anterior semicircular canal (which is shortened in late-diverging ceratopsids). The endocranial structures suggest early-diverging ceratopsians had a higher olfactory acuity and were adapted to hearing higher frequencies than late-diverging ceratopsians. Furthermore, the REQ suggests that Yinlong and Psittacosaurus were more highly encephalized than late-diverging ceratopsians and most extant reptiles. The angle of the lateral semicircular canal suggests that heads in ceratopsians display a transition from a forward posture to a more downward posture. Our new findings are significant for understanding the physiological changes during ceratopsian evolution and also have implications for the evolution of physiology in extant tetrapods.

Non-technical Summary

The horned dinosaurs underwent great changes throughout their evolution, including a shift in the locomotor mode (from bipedal to quadrupedal posture), enlargement of horns and frills, and increase in body size. Endocast reconstruction also suggests a decrease of olfactory acuity, hearing frequency, and the reptile encephalization quotient in derived horned dinosaurs. However, it is still unclear how these endocranial structures changed in early ceratopsian taxa due to a lack of information from the earliest horned dinosaurs. Here, we use virtual analytical methods to reconstruct the endocast of extinct dinosaurs and examine the evolution of the endocasts of horned dinosaurs that display some unique structures associated with auditory sense and smell. Based on the dataset and analytical detail on endocranial structures, we found that early-diverging horned dinosaurs (e.g., Yinlong and Psittacosaurus) had a high olfactory acuity and were adapted to hearing high frequencies, whereas the late-diverging horned dinosaurs (e.g., Triceratops) possessed lower olfactory acuity and hearing frequency. The early horned dinosaurs bear relatively large brain volumes, even higher than most extant reptiles. Head posture in ceratopsians displays a transition from forward facing to a downward tilt, indicating their different feeding preferences. These results are valuable for understanding the evolution of horned dinosaurs as well as extant tetrapods.

Introduction

The endocranial morphology in Ceratopsia has been studied for more than a century based on the observations of cranial cavities exposed in specimens or physical casts of the endocranial surface from specimens (Brown 1914; Brown and Schlaikjer 1940; Hopson 1979; Forster 1996). Computed tomography (CT) scanning has allowed new insights into the endocranial structure of extinct taxa. Recently, details of the endocranial anatomy of some ceratopsians have been described based on CT scanning, including Psittacosaurus (Zhou et al. 2007; Bullar et al. 2019; Napoli et al. 2019; Sakagami et al. 2023), Liaoceratops (Yang et al. 2023), Auroraceratops (Zhang et al. 2020), Pachyrhinosaurus (Witmer and Ridgely 2008; Tykoski and Fiorillo 2012), and Triceratops (Sakagami and Kawabe 2020). These studies concentrate on endocast anatomy, sensorineural attributes, and endocranial capabilities using quantitative comparisons. Differing from other ceratopsians, the endocast of Psittacosaurus reveals a large olfactory bulb and tall vertical semicircular canals, which suggests that they were agile animals (Zhou et al. 2007). The endocast of Liaoceratops more closely that of resembles Psittacosaurus than it does those of more late-diverging ceratopsians, possessing large olfactory bulbs, a discernible flocculus, and relatively long semicircular canals (Yang et al. 2023). In Auroraceratops, slender semicircular canals, a large olfactory bulb, and a swollen cerebrum appear transitional to ceratopsids (Zhang et al. 2020). Smaller olfactory bulbs, relatively shorter cochlear ducts, and smaller semicircular canals characterize late-diverging ceratopsians (Witmer and Ridgely 2008; Sakagami and Kawabe 2020). In general, the endocranial structures of ceratopsians can be seen to have undergone a series of transitions. However, the origin and early evolution of endocranial structures in ceratopsians have remained unclear due to a lack of information from the earliest-diverging bipedal ceratopsians.

Yinlong is one of the earliest ceratopsians from the Upper Jurassic Shishugou Formation of the Juggar Basin, Xinjiang, China (Xu et al. 2006). It is also the most complete and well-preserved Jurassic ceratopsian and is known from more than 30 individuals. Previous studies focused on morphology, phylogeny, and tooth replacement (Han et al. 2016, 2018; Hu et al. 2022). Here, for the first time, we present a detailed description of the endocast of an adult specimen of Yinlong created by micro-CT imaging, and provide a discussion about endocranial evolution among ceratopsians. In addition, we also provide the endocasts of the holotype of Liaoceratops and a new specimen of Psittacosaurus lujiatunensis for comparison. This study provides important new evidence in our understanding of the initial evolution and associated behavioral change of early-diverging ceratopsians.

Institutional Abbreviations

CUGW, China University of Geosciences (Wuhan), Wuhan, China; FPDM, Fukui Prefectural Dinosaur Museum, Fukui, Japan; IVPP, Institute for Vertebrate Paleontology and Paleoanthropology, Chinese Academy of Sciences, Beijing, China; PMoL, Palaeontological Museum of Liaoning, Liaoning, China; ZMNH, Zhejiang Museum of Natural History, Zhejiang, China.

Results

Endocranial Morphology of Yinlong downsi

The skull of IVPP V18637 is slightly compressed dorsolaterally and the braincase is well preserved. For the sake of simplicity, we refer to the digital casts of bone-bounded spaces that housed soft tissue as if they were the structures themselves. The digital reconstructions include the braincase (Supplementary Fig. A1), the cranial endocast, the endosseous labyrinth, the endocranial vasculature, and the cranial nerves (Fig. 1).

Figure 1. The 3D reconstructions of the skull, cranial endocast, and endosseous labyrinth of Yinlong downsi (IVPP V18637). Skull in lateral view (A). Cranial endocast in left lateral (B), dorsal (C), and ventral (D) views. Left endosseous labyrinth (E) and right endosseous labyrinth (F) in lateral view. Right endosseous labyrinth in dorsal view (G). Brain endocast is represented by yellow; cranial nerves by green; cerebral carotid artery and pituitary by red; metotic fissure by cyan; endosseous labyrinth by purple. Abbreviations: asc, anterior semicircular canal; cd, cochlear duct (cochlearis); car, cerebral carotid artery canal; cer, cerebrum; cp, complex of metotic fissure and CN IX-XI; crc, crus commune; de, dural peak; fl, flocculus; fv, fenestra ovalis (foramen vestibuli); lab, endosseous labyrinth; lsc, lateral semicircular canal; mo, medulla oblongata; mt(IX), the canal for CN IX (glossopharyngeal nerve) and the metotic fissure; ob, olfactory bulb; ol, optic lobe; ot, olfactory tract; pi, pituitary fossa; pt, prismatic protuberance; psc, posterior semicircular canal; vls, ventral longitudinal sinus. CN II, optic nerve canal; CN VI, abducens nerve canal; CN VII, facial nerve canal; CN X-XI, vagus and accessory nerves; CN XII, hypoglossal nerve canal.

Cranial Endocast

In IVPP V18637, the olfactory bulbs, the olfactory tract, the cerebrum, the optic lobe, the cerebellum, the pituitary, and the medulla oblongata are well reconstructed (Fig. 1A–D). Because it is not enclosed by bone, the ventral part of the olfactory region cannot be reconstructed (Fig. 1D). In dorsal view, the rostralmost part of the endocast is the oval olfactory bulbs (ob) which appear asymmetric due to compression (Fig. 1C). The olfactory bulbs are impressed into the ventral surface of the frontal bone. Caudally, the olfactory tract (Fig. 1C, ot) narrows from the olfactory bulbs and gradually widens, contacting the cerebrum after reaching its narrowest point. Caudal to the olfactory tract is the cerebrum (Fig. 1C, cer), visible as a rounded lateral swelling in the forebrain (Fig. 1B). The dorsal surface of the cerebrum is flat, as in Psittacosaurus (Zhou et al. 2007: fig. 2; Napoli et al. 2019: fig. 13),

Medial to the semicircular canals exists a pyramidal process projecting into the ring formed by the anterior semicircular canal (asc) and the crus commune, but not extending past it (Fig. 1B,C). The process is identified as the flocculus (fl) wrapped by the prootic (Fig. 1B). A relatively large flocculus is also known in Psittacosaurus and Liaoceratops (Sakagami et al. 2023: fig. 3; Yang et al. 2023: fig. 2). The flocculus plays a role in gaze stabilization and coordinating eye movements during movements of the head, the neck, and the body (Walsh et al. 2013). It is relatively small in most dinosaurs compared with that of flying reptiles and avians (Witmer et al. 2003: fig. 2; Walsh et al. 2013).

The midbrain is caudal to the cerebrum (Fig. 1B). The position of the optic lobes has been debated, and some researchers suggest that the optic lobes in ceratopsids are not visible (Hopson 1979: figs. 22, 23; Witmer and Ridgely 2008: fig. 2). However, in Yinlong, low swellings rostral to the dural peak and caudal to the cerebrum are identified as the optic lobes, similar to that in Psittacosaurus (Zhou et al. 2007: fig. 2; Fig. 1B). Caudal to the optic lobes is a caudally directed protuberance that is nearly parallel to the medulla oblongata and represents the dural peak (de), as seen in other dinosaurs (Yang et al. 2023: fig. 2; Fig. 1B). It was identified as the “cartilage area of the supraoccipital” by Zhou et al. (2007). In lateral view, and below the dural peak, the posterior edge of the cerebellum descends steeply to the medulla oblongata at an angle of 70°. This angle is larger than that in Psittacosaurus (Zhou et al. 2007: fig. 2; Fig. 1B). In ceratopsids, a smoother transition lacking this steep angle is seen between the cerebellum and medulla oblongata (Brown 1914: plate XXXIV; Witmer and Ridgely 2008: fig. 2; Sakagami and Kawabe 2020: fig. 2).

The ventral surface of the medulla oblongata (mo) is nearly parallel to the dorsal surface of the endocast rostral to the dural peak, which is the primitive archosaur pattern (Giffin 1989). The medulla oblongata narrows rostrodorsally and is constricted at the level of the otic capsule as in other dinosaurs (Hopson 1979). The endocast in Yinlong shows a longitudinal ridge ventral to the medulla oblongata that extends forward from the foramen magnum and widens immediately caudal to the constriction forming the prismatic protuberance (pt), as reported in Psittacosaurus (Zhou et al. 2007: fig. 3; Fig. 1D). The ventral longitudinal ridge has been identified as the sinus for the basilar artery in Psittacosaurus, Pachyrhinosaurus, and Triceratops (Forster 1996: fig. 8; Zhou et al. 2007: fig. 2; Witmer and Ridgely 2008: fig. 2), and the ventral longitudinal sinus (vls) is a common feature of archosaur endocasts (Witmer and Ridgely 2008: fig. 2). In ventral view, the center of the prismatic protuberance bears a slight pit caused by the tuber of the basioccipital (Fig. 1D).

The cast of the pituitary fossa (pi) is identified by the cavity of the basisphenoid and accompanied by the paired cerebral (internal) carotid arteries ventrally (Fig. 1B). The pituitary is ovoid-shaped with a tuber projecting caudally and constricting at contact with the cerebral carotid arteries (Fig. 1B, car). Differing from Auroraceratops, Protoceratops, Anchiceratops, and Triceratops (Brown 1914; Brown and Schlaikjer 1940; Forster 1996; Zhang et al. 2020), in Yinlong, the pituitary fossa ends above the level of the ventral margin of the medulla oblongata as in Psittacosaurus (Zhou et al. 2007; Napoli et al. 2019).

Inner Ear

Caudolateral to the cerebellum is the endosseous labyrinth (lab) of the inner ear, represented on both sides, and three semicircular canals on the right side. The right cochlear duct (cd) is only partially complete in our CT data due to the poorly preserved right exoccipital (Fig. 1F). Fortunately, the left cochlear duct is well preserved, although it may be compressed and stretched somewhat due to distortion (Fig. 1C). Hence, the complete morphology of the endosseous labyrinth in Yinlong can be reconstructed.

The upper part of the endosseous labyrinth is composed of the semicircular canals and the vestibule. The anterior semicircular canal (asc) is the largest and is similar to those of Psittacosaurus and Liaoceratops (Zhou et al. 2007: fig. 2; Fig. 1F). The semicircular canals in Yinlong are relatively stocky, and the crus commune (crc) is almost twice as wide as the vertical semicircular canals as in Psittacosaurus, Liaoceratops, and Protoceratops (Brown and Schlaikjer 1940: plate XII; Bullar et al. 2019: fig. 13; Yang et al. 2023: fig. 3). In ceratopsids, the semicircular canals are nearly as wide as the crus commune (Brown 1914: plate XXXIV; Brown and Schlaikjer 1940: plate XII; Witmer and Ridgely 2008: fig. 3; Sakagami and Kawabe 2020: fig. 3). The anterior semicircular canal and posterior semicircular canal (psc) both ascend from the dorsal side of the crus commune and then diverge at nearly a right angle (Fig. 1G), which also occurs in Psittacosaurus and Liaoceratops (Bullar et al. 2019: fig. 13; Yang et al. 2023: fig. 3). This angle is usually obtuse in Protoceratops and ceratopsids (Brown 1914; Brown and Schlaikjer 1940; Witmer and Ridgely 2008; Sakagami and Kawabe 2020). In addition, the anterior semicircular canal ascends significantly higher than the posterior semicircular canal, similar to the situation in other early-diverging ceratopsians. The lateral semicircular canal (lsc) is relatively stocky and widens at both ends (Fig. 1G).

The lower part of the endosseous labyrinth comprises the cochlear duct (cd) and the fenestra ovalis (fv) (Fig. 1E). In lateral view, the relatively long cochlear duct projects well below the fenestra ovalis (Fig. 1E). In Yinlong, the cochlear duct lies next to the metotic fissure (mt), as in Psittacosaurus and Liaoceratops (Figs. 1B, 2B and 3A). The basilar papilla housed within the cochlear duct is closely correlated with hearing frequency sensitivity (Manley 1972).

Figure 2. The 3D reconstructions of cranial endocast and endosseous labyrinth of Liaoceratops (IVPP V12738). Cranial endocast in dorsal (A), right lateral (B), and left lateral (C) views. Left endosseous labyrinth (D) and right endosseous labyrinth (E) in lateral view. Right endosseous labyrinth in dorsal view (F). Abbreviations: asca, ampulla of the anterior semicircular canal; cvcm, caudal middle cerebral vein; ocv, orbitocerebral vein canal; CN IV, trochlear nerve canal; CN V, trigeminal nerve canal. See Fig. 1 caption for other abbreviations.

Figure 3. 3D reconstructions of cranial endocast and endosseous labyrinth of Psittacosaurus (CUGW VH104). Cranial endocast in right lateral (A) and dorsal (C) views. Right endosseous labyrinth in lateral (B) and dorsal (D) views. The dotted line represents the possible morphology of the damaged right olfactory bulb. See Fig. 1 and Fig. 2 captions for abbreviations.

Cranial Nerves

The optic nerve (CN II) canal exits the braincase through the optic foramen immediately rostrodorsal to the pituitary fossa and passes rostroventrally through the laterosphenoid (Fig. 1B). The abducens nerve (CN VI) canal is narrow and surrounded by the basisphenoid. CN VI extends from the brain cavity toward the pituitary and is oriented rostrodorsally (Fig. 1B). In Pachyrhinosaurus, Triceratops, and Anchiceratops, CN VI passes rostroventrally from the rostroventral end of the medulla through both sides of the pituitary (Brown 1914: plate XXXIV; Witmer and Ridgely 2008: fig. 2; Sakagami and Kawabe 2020: fig. 2). The facial nerve (CN VII) canal is observed on both sides of IVPP V18637 and situated rostral to the endosseous labyrinth as in Psittacosaurus and Pachyrhinosaurus (Witmer and Ridgely 2008; Napoli et al. 2019). The glossopharyngeal nerve (CN IX) canal exits the braincase from the metotic fissure on the left side. The vagus and accessory nerve (CN X–XI) canals are posterior to CN IX and exit the braincase from a common canal on the left side. The hypoglossal nerve (CN XII) canal bears two branches of CN XII on both sides (Fig. 1D), whereas only one branch is preserved in Protoceratops and ceratopsids (Brown 1914; Brown and Schlaikjer 1940; Witmer and Ridgely 2008; Sakagami and Kawabe 2020). The rostral branch on the right side may be reconstructed in the otic capsule because of a broken bony boundary.

Endocranial Vasculature

As for the intraneurocranial vasculature of Yinlong, only the cerebral carotid arteries could be reconstructed entirely (Fig. 1D). The rostral part of cerebral carotid arteries is wrapped by the basisphenoid and the caudal part by the exoccipital. The left and right cerebral carotid arteries join as they enter the pituitary fossa. They curve caudolaterally at 120° from each other immediately below the pituitary fossa, as in other ceratopsians.

Additional Endocranial Features of Liaoceratops yanzigouensis

Recently, the endocranial morphology of Liaoceratops has been studied by Yang et al. (2023). The endocast of the holotype of Liaoceratops (IVPP V12738) reconstructed here provides some new information, with most structures preserved (Fig. 2). Here, we describe some previously undescribed neuroanatomical characteristics.

In dorsal view, the olfactory bulbs are well preserved and symmetric, appearing oval in shape. The olfactory bulbs are separate and diverge from the midline at an acute angle (Fig. 2A). The olfactory tract of Liaoceratops is short and wide with a longitudinal ridge, as in PMoL-AD00097 (reported by Yang et al. 2023) on the dorsal surface, but differing from the olfactory tracts of Yinlong, Psittacosaurus, and Auroraceratops (Zhang et al. 2020: fig. 3; Fig. 2A). Due to compression, the right part of the olfactory tract appears fused with the cerebrum (Fig. 2B). Likewise, the ventral part of the olfactory region cannot be reconstructed, as it is not enclosed by bone.

Additional features of IVPP V12738 were previously unknown or unlike the condition in PMoL-AD00097 (Yang et al. 2023). Caudal to the olfactory tract, the cerebrum occurs as a slightly pointed swelling (Fig. 2B). In dorsal view, the right surface of the cerebrum shows the orbitocerebral vein (Fig. 2A, ocv) projecting laterodorsally. The left orbitocerebral vein is not preserved, possibly due to the deformation of the braincase. In left lateral view, the optic lobe is located caudoventral to the cerebrum and visible as a large swelling immediately above the labyrinth (Fig. 2C). The right flocculus is a wedge-shaped tuber projecting caudally and extends laterally past the ring formed by the anterior semicircular canal and the crus commune (Fig. 2B). It is relatively larger than the flocculus in PMoL-AD00097 (Yang et al. 2023: fig. 2). The right caudal middle cerebral vein (cvcm) is located immediately below the dural peak and projects caudally (Fig. 2B).

The endosseous labyrinth of the inner ear on both sides is situated lateral to the cerebellum and rostral to the medulla oblongata (Fig. 2D,F). The anterior semicircular canal and posterior semicircular canal are siphonate and form a slightly acute angle to one another in dorsal view (Fig. 2F). In dorsal view, the lateral semicircular canal widens to the junction with the posterior semicircular canal (Fig. 2F). The crus commune is almost twice as wide as the vertical semicircular canals. The lower part of the inner ear, the cochlear duct, is elongated and relatively long.

The facial nerve (CN VII) is present in IVPP V12738. CN VII is situated below the anterior semicircular canal and extends laterally from the brain cavity (Fig. 2B).

Additional Endocranial Features of Psittacosaurus lujiatunensis

The endocast of Psittacosaurus has been well studied in previous works (Zhou et al. 2007; Bullar et al. 2019; Napoli et al. 2019; Sakagami et al. 2023). However, the endocast of Psittacosaurus (CUGW VH104) reconstructed here provides additional new information about neuroanatomical characteristics (Fig. 3), including the relatively complete cochlear duct for the first time.

In CUGW VH104, the olfactory bulbs are separate and oval (Fig. 3C). In lateral view, medial to the endosseous labyrinth, lies the flocculus (Fig. 3A). The flocculus is relatively wide and does not exceed the plane composed of the anterior semicircular canal and the crus commune and is smaller than those in ZMNH M12414 and ZMNH M12423 (Sakagami et al. 2023: fig. 5). The size of the flocculus is larger than that of Yinlong and smaller than that of Liaoceratops. In the upper part of the inner ear, the semicircular canals are elongated, as in previous studies (Fig. 3B,C). The fenestra ovalis is situated rostral to the metotic fissure and these two are separated by a thin ridge. The cochlear duct of Psittacosaurus is relatively longer than those of late-diverging ceratopsids.

Sense of Smell

Olfactory acuity was explored by Sakagami and Kawabe (2020) in a scatter plot showing the relationship between body mass and olfactory ratio. The ceratopsid Triceratops plotted well below the regression line, indicating that its olfactory acuity was lower than that in both ornithopods and theropods with similar body mass. When the early-diverging ceratopsians Yinlong and Psittacosaurus are added to this dataset, they plot above the regression line, indicating higher olfactory acuity than theropods with similar body size (Fig. 4A). In addition, the olfactory acuity of Liaoceratops (53.24%) is higher than that of both Psittacosaurus (43.35%) and Yinlong (44.6%).

Figure 4. Comparison of sensorineural attributes and endocranial capabilities based on the quantitative evaluations of the endocasts in amniotes. (A) Relationship between olfactory ratio and body mass for dinosaurs and crocodilians (data from Zelenitsky et al. 2009; Sakagami and Kawabe 2020); (B) reptile encephalization quotient (REQ) of three dinosaur clades and extant reptiles; (C) the estimated mean hearing sensitivities and (D) the estimated hearing range inferred from the cochlear length based on interpolation of three ceratopsians into the plots of Walsh et al. (2009) and Hanson et al. (2021); (E) best frequencies of hearing and high frequency of hearing limit for selected dinosaurs (modified from Sakagami and Kawabe 2020). The figure 4 is produced from Supplementary data S1–S4.

Reptile Encephalization Quotient (REQ)

The volume of the brain endocast only includes the forebrain, the midbrain, and the hindbrain. Based on the smallest femoral circumference (about 100 mm in IVPP V18637), the body mass is about 66.834 kg. Therefore, REQ using 73% of brain volume is 1.49 (Table 1). As mentioned earlier, the braincase of IVPP V18637 shows some gentle degree of compression. As a result, the REQ for Yinlong might have been slightly higher. In the same way, we calculated the REQ of Psittacosaurus to be about 2.43 (Table 1). Theropods are characterized by relatively high REQ values (0.83–7.91) (Fig. 4B).

Table 1. Endocranial volume, body mass, brain-to-endocast (BEC) index, and reptile encephalization quotient (REQ) of known ceratopsians. Abbreviations: EV, endocranial volume (expressed as the mass in Protoceratops); M _bd, body mass; V _br, brain volume; V _bd, body volume. The data for Protoceratops and Triceratops (A and B) come from Hurlburt (1996).

Hearing Ability

Following Gleich et al. (2005), the optimal hearing frequency for Yinlong is estimated at 913 Hz. The highest hearing frequency limit of Yinlong is 2726 Hz. Psittacosaurus has a similar range of hearing frequency as Yinlong, and Liaoceratops is estimated to have the highest hearing frequency and the widest range of hearing frequencies (Fig. 4E). Compared with ceratopsids, early-diverging ceratopsians have relatively higher hearing frequencies and wider optimal hearing frequencies (Fig. 4E). Similar to Lambeosaurus and theropods, early-diverging ceratopsians appear to have been adapted to hear higher-frequency sounds (Fig. 4E).

The cochlear duct (cd) length of IVPP V18637 is 11.06 mm, and its basicranial length is 36.29 mm. Following the equation of Walsh et al. (2009), the estimated mean hearing sensitivity and the estimated hearing range are about 2292.46 Hz and 3825.91 Hz in Yinlong, 3262.19 Hz and 5613 Hz in Psittacosaurus, and 2683.18 Hz and 4546.19 Hz in Liaoceratops, respectively (Fig. 4C,D).

Discussion

Head Posture

When animals are alert and in a habitual stance, they tend to hold their heads in an orientation that places the lateral semicircular canal roughly in the horizontal plane (Witmer et al. 2003, 2008; Coutier et al. 2017). Therefore, the typical head posture of extinct animals can be evaluated by orienting the lateral semicircular canal horizontally. However, Taylor et al. (2009) pointed out that in living tetrapods the lateral semicircular canal is not always held horizontally. Despite these flaws, the orientation of the lateral semicircular canal is often used as a proxy for the head posture, especially in paleontological studies (Witmer et al. 2003; Coutier et al. 2017).

Previous researchers have demonstrated that the head postures of dinosaurs and other archosaurs were rather variable, even within a clade (Witmer et al. 2003, 2008). We adjusted our 3D models to place the lateral semicircular canal horizontally (dotted line) and parallel to the ground (Fig. 5). Uniquely, Yinlong has a head posture in which the palatal plane is inclined approximately 10° above the horizontal plane (Fig. 5A). In Psittacosaurus, the palatal plane is usually below the horizontal plane, and the angles between the lateral semicircular canal and the palatal plane are variable in the ontogenetic series of Psittacosaurus but show a clear trend of convergence (Bullar et al. 2019). The angle changes from 38° in the hatchling stage, to 25° in the juvenile, and to 15° in the adult (Fig. 5B), which is consistent with the suggested posture shift from quadrupedal to bipedal during growth (Zhao et al. 2013a; Bullar et al. 2019). Liaoceratops has a head posture similar to Psittacosaurus with an angle of about 19° (Fig. 5C). In Triceratops, orienting the lateral semicircular canal horizontally results in a head posture such that the palatal plane was inclined approximately 45° below the horizontal plane (Sakagami and Kawabe 2020; Fig. 5D).

Figure 5. Comparison of the alert head posture in Ceratopsia when the lateral semicircular canal is horizontal and parallel to the ground. (A) IVPP V18637; (B) IVPP V12617; (C) IVPP V12738; (D) FPDM-V-9775. The 3D model of Yinlong is flipped horizontally, because the right side of the skull is deformed. Psittacosaurus is modified from Bullar et al. (2019) and flipped horizontally. Triceratops is modified from Sakagami and Kawabe (2020) and flipped horizontally.

In general, skulls in ceratopsians displays a transition from the upward palatal posture in Yinlong, to a slight downward posture in Psittacosaurus and Liaoceratops, to 45° below the horizontal plane in Triceratops. Browsers (upward or horizontal) are expected to hold their heads higher than grazers (downward foragers) and mixed feeding species (closer to horizontal than a downward inclination) (Schellhorn 2018). Therefore, Yinlong is possibly a browser, whereas Psittacosaurus, Liaoceratops, and Triceratops, which bear a downward head posture, are possibly mixed feeders or grazers.

In late-diverging ceratopsians, frills and horns are large and exaggerated. Knapp et al. (2021) suggested that the frill functioned as a sociosexual signal in ceratopsian dinosaurs. It is advantageous for ceratopsids and other late-diverging ceratopsians with exaggerated ornamentation to hold their heads at an efficient angle for displaying frills and horns, such as at 45° below the horizontal plane, as in Triceratops (Fig. 5D). This suggests that feeding style is not the only factor that affect head posture.

The Evolution of Endocast Morphology in Cerapoda

The endocasts of Yinlong, Psittacosaurus, and Liaoceratops described here offer an excellent opportunity for comparing the endocasts of different ceratopsians in order to understand the early evolution of the endocast in Ceratopsia. In addition, the endocasts of ceratopsians can be compared with those of known ornithopods to investigate the evolution of ornithischian neuroanatomy.

Cranial Endocast Comparative Anatomy

In Yinlong, Psittacosaurus, Liaoceratops, and Auroraceratops (Zhang et al. 2020), the olfactory bulbs are oval and separate from one another along the midline, which differs from ceratopsids, in which the olfactory bulbs contact each other (Witmer and Ridgely 2008; Sakagami and Kawabe 2020). Among Ceratopsia, Liaoceratops has the relatively largest olfactory bulbs (olfactory ratio: 53.24%) followed by Yinlong and Psittacosaurus (44.60% and 43.35%, respectively). Ceratopsids (Pachyrhinosaurus and Triceratops) have the smallest olfactory bulbs (34.34% and 40.98%, respectively). Additionally, the olfactory tracts of Yinlong and Psittacosaurus develop longitudinal compression on the dorsal surface (Figs. 1C, 3C), while Liaoceratops and ceratopsids develop a longitudinal ridge.

In Cerapoda, most groups have olfactory bulbs that diverge from the midline at an acute angle, with the exception of late-diverging ceratopsids and hadrosaurids (Fig. 6). We adjusted the endocast of these taxa to place the lateral semicircular canal horizontally and found that the orientation of the olfactory region also changes in ceratopsians and ornithopods. In Ceratopsia, the angle between the olfactory region and horizontal plane changes from 15° downward (Yinlong) to 10° upward (Triceratops) (Fig. 6). In Ornithopoda, the angle changes from 30° downward (Jeholosaurus) to 50° upward (Hypacrosaurus) (Fig. 6). In general, the olfactory region in Ceratopsia and Ornithopoda underwent similar transformations, including the fusing of the olfactory bulbs, the raising of the olfactory region, and the shortening of the olfactory tracts. Ceratopsians are different from Ornithopoda in that their olfactory bulbs narrow during their evolution. Large olfactory bulbs typically also occur in some predatory theropods such as Allosaurus (Rogers 1999) and Tyrannosaurus (Brochu 2000), which may indicate an adaptation for seeking out prey. Large olfactory bulbs in early-diverging ceratopsians no doubt play a very different role. In Pachyrhinosaurus and Triceratops, the olfactory bulbs became narrower than the midbrain. Similarly, the acuity of the sense of smell also was lower than in theropods (Fig. 4A).

Figure 6. Comparison of cranial endocasts and endosseous labyrinth in Ornithischia, displayed on a cladogram. Hypsilophodon, Thescelosaurus, and Dryosaurus from Galton (1989) and Button and Zanno (2023). Dysalotosaurus from Lautenschlager and Hübner (2013). Tenontosaurus from Thomas (2015). Hypacrosaurus from Evans et al. (2009). Auroraceratops from Zhang et al. (2020). Protoceratops from Brown and Schlaikjer (1940). Pachyrhinosaurus from Witmer and Ridgely (2008). Anchiceratops from Brown (1914). Triceratops from Sakagami and Kawabe (2020). Parts of endocasts and inner ears have been rotated or flipped to maintain the same right lateral orientation. The inner ear of Yinlong consisted of the left lagena and the right semicircular canals. The black outlines represent the possible morphologies of the endosseous ducts. The angles signify the angle formed between the olfactory region and the horizontal line. Positive values indicate an upward direction, while negative values indicate a downward direction.

In Yinlong, Liaoceratops, and Psittacosaurus, the cerebrum extends laterally and forms two angular processes resulting in a nearly triangular dorsal surface of the brain cavity (Figs. 1C, 2A, 3C). In Auroraceratops and ceratopsids, the cerebrum also extends laterally but forms a nearly circular lateral surface, resulting in an elliptic dorsal surface (Witmer and Ridgely 2008; Sakagami and Kawabe 2020; Zhang et al. 2020). In lateral view, the cerebrum in Yinlong is flat dorsally, whereas that in Psittacosaurus is slightly convex (Fig. 3A). In Liaoceratops, the cerebrum bears obvious swelling dorsally similar to late-diverging ceratopsians (Brown 1914; Brown and Schlaikjer 1940; Witmer and Ridgely 2008; Sakagami and Kawabe 2020; Zhang et al. 2020). The cerebrum thus appears to transform from a flat dorsal surface to a rounded swelling during the evolution of Ceratopsia (Fig. 6). In Jeholosaurus and Hypsilophondon, the dorsal surface of the cerebrum is similar to that of early-diverging ceratopsians (Fig. 6). Therefore, the dorsal flat cerebrum may be the primitive character of Cerapoda.

Zhou et al. (2007) proposed that the orientation of the pituitary fossa housed in the basisphenoid exhibits two patterns in Ceratopsia. The pituitary fossa is caudoventrally directed in Psittacosaurus and Ceratopsidae (Witmer and Ridgely 2008; Sakagami and Kawabe 2020), but is vertically oriented in Protoceratops (Brown and Schlaikjer 1940). In Liaoceratops, the pituitary fossa is also directed vertically (Fig. 6). In Ornithopoda, the pituitary fossa is directed caudoventrally (Fig. 6). Differing from Protoceratops and Anchiceratops, in which the cerebral carotid artery canals enter the lateral sides of the pituitary fossa (Brown and Schlaikjer 1940), in other ornithischians, the cerebral carotid artery canals all enter the posteroventral side of the pituitary fossa.

Before this study, the flocculus had only been reported in Liaoceratops and Psittacosaurus among Ceratopsia (Sakagami et al. 2023; Yang et al. 2023). Here, three additional specimens also preserve the flocculus. In Yinlong, the flocculus is a small pyramidal process. In Psittacosaurus, the flocculus is an elliptic process and does not extend beyond the ring formed by the anterior semicircular canal and the crus commune (Fig. 3A). Liaoceratops has the largest flocculus among Ceratopsia; it is a wedge-like process and extends beyond the anterior semicircular canal–crus commune ring. No flocculus is present in ceratopsids (Fig. 6). The disappearance of the flocculus in ceratopsians may indicate reduced head movement capability or a great reduction in size. Among Ornithopoda, early-diverging ornithopods Jeholosaurus and iguanodontids have a relatively larger flocculus (Fig. 6). During the evolution of Ornithopoda, the flocculus existed for a long period but disappeared in late-diverging hadrosaurids. Overall, the flocculus went through a similar pattern of change in Ceratopsia and Ornithopoda, which may correspond to the similar transformation in the locomotion mode and increased body size.

Inner Ear

The functional morphology and evolution of the endosseous labyrinth are important for understanding the sensory evolution and success of Ceratopsia and Ornithopoda. As in Psittacosaurus, Auroraceratops, Liaoceratops, and Protoceratops (Brown and Schlaikjer 1940; Zhou et al. 2007; Zhang et al. 2020), the posterior semicircular canal in Yinlong is positioned near the foramen magnum, which differs from the posterior semicircular canal in Anchiceratops, Pachyrhinosaurus, and Triceratops, in which a relatively longer medulla oblongata results in the canals being located far rostral to the foramen magnum (Brown 1914; Sakagami and Kawabe 2020). As in more late-diverging ceratopsians (Brown 1914; Brown and Schlaikjer 1940; Sakagami and Kawabe 2020), the semicircular canals in Yinlong are relatively short and thick, differing from those of Psittacosaurus and Liaoceratops, which have elongate and thin semicircular canals (Fig. 6). The anterior semicircular canal in early-diverging ceratopsians (Yinlong, Psittacosaurus, and Liaoceratops) ascends significantly higher than the posterior semicircular canal, while the anterior semicircular canal reaches a similar height in ceratopsids (Fig. 6). In addition, the angle between the anterior semicircular canal and posterior semicircular canal increased through the evolution of Ceratopsia.

In Ornithopoda, the endosseous labyrinth followed similar evolutionary patterns, in that early-diverging taxa have a higher and longer anterior semicircular canal and the angle between the anterior semicircular canal and posterior semicircular canal in late-diverging hadrosaurs becomes obtuse (Fig. 6). The semicircular canals of early-diverging ornithopods (e.g., Jeholosaurus) are flattened in contrast to the cylindrical condition in ceratopsians. The morphological difference of the vertical semicircular canals indicates that the sensory input for the reflexive stabilization of gaze and posture in later-diverging ceratopsians and hadrosaurids was lower than in early-diverging taxa. It has been suggested that bipedal dinosaurs exhibit a dorsoventrally tall anterior semicircular canal, while quadrupedal dinosaurs exhibit a horizontally broad anterior semicircular canal and the posterior semicircular canal (Hanson et al. 2021). However, iguanodontians (e.g., Tenontosaurus), which may be partially quadrupedal, also bear a tall anterior semicircular canal and orthogonal vertical semicircular canals (asc and posterior semicircular canal). Therefore, we conclude that the morphology of the semicircular canals may be influenced by indicators such as phylogeny.

The cochlear ducts of late-diverging ceratopsians relative to their heads are comparatively shorter than in early-diverging ceratopsians (Fig. 6). In Ornithopoda, early-diverging taxa also presented relatively elongated cochlear duct. Therefore, it can be postulated that the relatively elongated cochlear duct is a plesiomorphic trait that gradually shortens in late-diverging ceratopsians. However, late-diverging hadrosaurs also relatively retained long cochlear ducts and ornithopodans showed a different evolutionary pattern. As a ventrally extended cochlear duct has been linked to increased auditory acuity, perhaps the retention of this trait marked a significant increase in auditory ability, vocality, sociality, and cognition, all of which are thought to be integrally linked (Brown et al. 2020). Subsequently, the relatively elongated cochlear duct in early-diverging ceratopsians and ornithopods may contribute to complex social behaviors related to vocalization, as discussed earlier.

Behavior

The endocasts of three early-diverging ceratopsians provide associated structures as a basis for a realistic interpretation of paleobiology and behavior. Parental care or social behavior has been reported in some hadrosaurs (Horner and Makela 1979; Horner 2000). In lambeosaurines, the nasal crests have been hypothesized to function in vocalization (Weishampel 1997) and visual display (Hopson 1975). In Psittacosaurus, juvenile-only clusters suggest that these juveniles may have associated as a close-knit, mixed-age herd either for protection, to enhance their foraging, or as putative helpers at the parental nest (Zhao et al. 2013b). These complex behaviors are related, to some extent, to their high degree of encephalization. In addition, some complex behaviors have been well documented in theropods, such as avian-type nesting behavior (Norell et al. 1995, 2018). These fossil taxa with complex behaviors all have selected for larger brain sizes and higher REQs, as mentioned earlier. Furthermore, a consistent association between total brain size and sociality has been found in ungulates (Pérez-Barbería and Gordon 2005; Shultz and Dunbar 2006). To a certain extent, it is possible to associate complex behaviors with higher REQ.

Previous studies have shown that the REQs of Triceratops and Protoceratops are 0.832 and 2.115, respectively (Jerison 1973; Hurlburt 1996). The recalculated REQ value of Psittacosaurus is 2.43. The assessment of REQ for Yinlong and Psittacosaurus suggests that these species were more encephalized than most extant reptiles (0.4016–2.4035) (Hurlburt 1996) and known sauropods (0.23–1.42) (Fig. 4B). In general, the early-diverging ceratopsians have relatively higher REQs than ceratopsids, indicating a high degree of encephalization. As noted earlier, we suggest that early-diverging ceratopsians had relatively higher hearing ranges and greater sensitivities than most theropods along the lineage leading to birds. Relatively higher hearing frequency is also observed in ecological analogous modern ungulates than in their predators (Heffner and Heffner 1983, 1992). Psittacosaurus has the highest mean hearing sensitivity and hearing range among known non-avian dinosaurs (Fig. 4C,D). It is possible that a high degree of encephalization and high hearing frequency existed in Psittacosaurus to support its complex social behaviors (Zhao et al. 2013b). Therefore, it is likely that some complex social behaviors also exist in Yinlong with its relatively high hearing frequency and REQ. In Protoceratops, the well-developed frill may have played a sociosexual signaling role (Knapp et al. 2021), and the presence of intraspecific communication may account for the high degree of encephalization to some extent. Why the high degree of encephalization is reduced in Triceratops, which also has a well-developed frill, is worthy of further investigation and suggests a decoupling between sociosexual signaling and hearing and encephalization.

Conclusion

The examination of three early-diverging ceratopsians not only filled an important morphological gap but also provided the data for ecological interpretations of the endocast morphology. Early-diverging ceratopsians exhibit some unique characteristics, including large and diverging olfactory bulbs, the triangular dorsal surface of the endocast, the relatively large flocculus, thin and elongated semicircular canals, long and high anterior semicircular canals, and a long cochlear duct.

Based on our interpretations of the endocranial anatomy, we suggest that the sense of smell in early-diverging ceratopsians is more sensitive than in Protoceratops and late-diverging ceratopsids. Early-diverging ceratopsians had higher hearing frequencies than ceratopsids and non-avian theropods. Early-diverging ceratopsians are more encephalized. Yinlong has a unique head posture, in which the palatal line is above the horizontal plane, while Psittacosaurus and Liaoceratops hold their palatal lines at a relatively low angle below the horizontal plane.

Comparisons with other ornithischians suggest a number of evolutionary patterns in the endocasts in Ceratopsia and Ornithopoda. In the evolution of these clades, the olfactory bulbs gradually narrowed and merged, and the flocculus disappeared. Several morphological adaptations also occurred during the evolution of the inner ear, including the reduction of the anterior semicircular canal, the increasing angle between the anterior semicircular canal and the posterior semicircular canal, and the shortening of the cochlear duct. These changes are significant for understanding the evolution of locomotion and behavior in dinosaurs and extant tetrapods.

Materials and Methods

Materials

Among all specimens of Yinlong, only IVPP V18637 preserved the relatively complete braincase. IVPP V18637 is a nearly complete skull lacking a mandible with an incomplete postcranial skeleton. It is the largest preserved specimen, with a skull length (rostral to quadrate condyle) of approximately 23 cm.

The skull and mandible of Liaoceratops were collected from the Yixian Formation of Liaoning, China. IVPP V12738 has a nearly complete skull and a detached mandible with dorsolateral compression. The skull length is approximately 11.29 cm.

A new specimen of Psittacosaurus (CUGW VH104) housed at the China University of Geosciences (Wuhan) is examined for comparison. This specimen includes a complete skull, mandible, and postcranial skeleton. The skull length is about 13.98 cm.

CT Scanning

Nondestructive high-resolution X-ray micro-CT was used to visualize the internal space and anatomical features of the braincase. IVPP V18637 was scanned using a 450 kV micro-CT instrument (450 ICT, Institute of High Energy Physics, Chinese Academy of Sciences) at the Key Laboratory of Vertebrate Evolution and Human Origins, Beijing, China. Scanning parameters are listed in Supplementary Table A1. IVPP V12738 was scanned using a 225 kV micro-CT instrument at the Key Laboratory of Vertebrate Evolution and Human Origins. CUGW VH104 was scanned using a 300 kV micro-CT instrument (Phoenix Vtomex M) and the detector (Dynamic41-100) at Yinghua Testing (Shanghai). CT datasets were input in Mimics (v. 15.0 and 16.0, Materialise Corporation, Leuven, Belgium) to render 3D models of the braincase and the brain cavity. CT scans of the three specimens (all cropped to the braincase regions) have been uploaded in Dryad in .BMP file format (http://doi.org/10.5061/dryad.h70rxwdr8).

Cognitive Capabilities

The digital endocast is a 3D cast of any internal space (Balanoff et al. 2016). However, the soft tissue forming the interface with the surrounding skeleton is not the brain but the superficial surface of the dural meninges and vasculature enveloping the brain (Balanoff et al. 2016). Therefore, there is a variance between the endocast and the brain in size and shape. It can vary widely between lineages (Balanoff et al. 2016) and over ontogeny (Watanabe et al. 2019). We followed the study about the brain-to-endocast index (BEC) of two extant archosaur taxa, Alligator mississippiensis and Gallus gallus, with divergent ontogenetic trends (Watanabe et al. 2019). The value midway between the mean BEC in the oldest individuals of Alligator and Gallus is about 73% (Watanabe et al. 2019; Knoll et al. 2021). It is rational to estimate that the proportion of the endocast occupied by the brain tissues was about 73% for IVPP V18637, the largest specimen.

Although it remains controversial to link “brain” and “cognition,” both represent a set of particularly complex variables (Willemet 2013), the encephalization quotient is one of the very few approaches available to obtain insight, as rudimentary as it may be into the complexity of extinct taxa (Knoll et al. 2021). The REQ is always a dimensionless constant, especially in estimates for extant reptiles. However, recent studies generally calculate REQ by the ratio of the volume of the brain to the estimated mass of the body in extinct species (Paulina-Carabajal and Currie 2017; Müller et al. 2021). In this case, the REQ of an extinct species is difficult to compare with those of extant species. Therefore, we followed the expected brain volume equation modified from the equation calculating REQ by the body and brain mass from Hurlburt (1996):

(1)$${\rm REQ\ } = V_{{\rm br}}/( {0.0155^\ast V_{{\rm bd}}^{0.553} } ) $$

where V _br and V _bd correspond to brain and body volume, respectively.

This equation reflects brain–body relationships in extant non-avian reptiles. We calculated the volume of the body in extinct species following the body density of an adult individual Tarentola mauritanica that Knoll et al. (2021) measured by water displacement, and ρ is 1.035 g/cm³. Finally, we collected previous datasets about REQ in extinct species to compare with extant reptiles (Supplementary Data S1).

Body-Mass Estimations

The body mass was estimated using an estimation method based on the relation between body mass and femoral circumference (Campione and Evans 2012; Campione et al. 2014). Yinlong was usually identified as a bipedal animal (Xu et al. 2006). Therefore, we adopted the equation of Campione et al. (2014):

(2)$${\rm lo}{\rm g}_{10}{\rm B}{\rm M}_{{\rm bip}} = 2.754^\ast {\rm lo}{\rm g}_{10}( {C_{{\rm femur}}} ) - 0.683$$

where BM_bip is the body mass of the biped, and C _femur is the femoral circumference.

Hearing Ability

In extant reptiles and birds, the length and other dimensions of the basilar papilla (the hearing organ of the inner ear) are closely correlated with hearing sensitivity (Manley 1972). To estimate the hearing range, we followed the method described by Walsh et al. (2009). We measured the cochlear duct length and basicranial length (measured from the midventral border of the foramen magnum to the basisphenoid–presphenoid suture). The cochlear duct length is scaled to basicranial length to reduce size effects (Supplementary Data S2). Values were log transformed to meet the assumptions of multiple regression. The best hearing range is calculated by the equation:

(3)$$y = 6104.3x + 6975.2 ( {R^2 = 0.547; \;p < 0.01} ) $$

where x is the scaled and log-transformed cochlear duct length, and y is the best hearing range (in Hz) (Walsh et al. 2009).

The mean best hearing frequency is calculated by the equation:

(4)$$y = 3311.3x + 4000.8 ( {R^2 = 0.566; \;p < 0.01} ) $$

where y is the mean best hearing frequency (in Hz) (Walsh et al. 2009).

To compare hearing ability within Ceratopsia, we adopted the dataset of Sakagami and Kawabe (2020), which includes Triceratops and Pachyrhinosaurus. This study followed the method presented by Gleich et al. (2005), who calculated the best hearing frequency based on the basilar papilla length defined by two-thirds of the cochlear duct length (Supplementary Data S3). The best hearing frequency is calculated by the equation:

(5)$$y = 5.7705e^{{-}0.25x}( {\,p < 0.001} ) $$

where x is the basilar papilla length, and y is the best hearing frequency (in kHz).

The high-frequency hearing limit is calculated by the equation:

(6)$$y = 1.8436x + 1.0426^{}( {\,p < 0.001} ) $$

where x is the best hearing frequency, and y is the high-frequency hearing limit.

The method Gleich et al. (2005) presented is calculated based on the correlation analysis between the basilar papilla length and the hearing frequency in extant archosaurs (mainly composed of birds). However, it cannot reduce size effects. Therefore, we chose another method, the one Walsh et al. (2009) presented, which standardizes the cochlear duct length using basicranial length; this method is based on more reptile groups, including Testudines, Crocodylia, and Squamata. In addition, the results calculated by these two methods are different. The results based on Gleich et al. (2005) are about the highest hearing frequency and best frequency of hearing, while the results based on Walsh et al. (2009) are about the mean hearing and hearing range. We dedicate more discussion to the results based on the method of Gleich et al. (2005) and presented in the text, because there are more data on hearing frequencies of ceratopsids and hadrosaurids. We prefer to retain two methods to provide the comparison with extant groups and an example for future analysis about the hearing ability of ceratopsians.

Olfactory Acuity

Olfactory bulb size has been used as an indicator of the ability to discriminate different odors and sensitivity to odors in extant archosaurs and mammals (Zelenitsky et al. 2009). Zelenitsky et al. (2009) presented the first quantitative evaluation of olfactory acuity in extinct theropod dinosaurs and a phylogenetically corrected least-squares regression of olfactory ratio to body mass in theropods:

(7)$$y = 0.1237x + 1.316 ( {r = 0.87, \;p < 0.05} ) $$

where x is the log-transformed body mass (kg), and y is the log-transformed olfactory ratio (%).

The olfactory ratio is calculated as the ratio of the greatest diameter of the olfactory bulb to the greatest diameter of the cerebral hemisphere. The olfactory ratios of Yinlong, Liaoceratops, and Psittacosaurus are calculated. In addition, we collected the datasets of Triceratops and the other three ornithischians provided by Sakagami and Kawabe (2020) and interpolated these data into the datasets of Zelenitsky et al. (2009) to compare the olfactory acuity with other theropods (Supplementary Data S4).