Modified skulls but conservative brains? The palaeoneurology and endocranial anatomy of baryonychine dinosaurs (Theropoda: Spinosauridae)

Abstract The digital reconstruction of neurocranial endocasts has elucidated the gross brain structure and potential ecological attributes of many fossil taxa, including Irritator, a spinosaurine spinosaurid from the “mid” Cretaceous (Aptian) of Brazil. With unexceptional hearing capabilities, this taxon was inferred to integrate rapid and controlled pitch‐down movements of the head that perhaps aided in the predation of small and agile prey such as fish. However, the neuroanatomy of baryonychine spinosaurids remains to be described, and potentially informs on the condition of early spinosaurids. Using micro‐computed tomographic scanning (μCT), we reconstruct the braincase endocasts of Baryonyx walkeri and Ceratosuchops inferodios from the Wealden Supergroup (Lower Cretaceous) of England. We show that the gross endocranial morphology is similar to other non‐maniraptoriform theropods, and corroborates previous observations of overall endocranial conservatism amongst more basal theropods. Several differences of unknown taxonomic utility are noted between the pair. Baryonychine neurosensory capabilities include low‐frequency hearing and unexceptional olfaction, whilst the differing morphology of the floccular lobe tentatively suggests less developed gaze stabilisation mechanisms relative to spinosaurines. Given the morphological similarities observed with other basal tetanurans, baryonychines likely possessed comparable behavioural sophistication, suggesting that the transition from terrestrial hypercarnivorous ancestors to semi‐aquatic “generalists” during the evolution of Spinosauridae did not require substantial modification of the brain and sensory systems.

Given the strong evidence for aquatic behaviour in the skeletal (and particularly cranial) anatomy of these dinosaurs, it follows that the brain and nervous system may be expected to exhibit specialisations for aquatic foraging or swimming. Schade et al. (2020) described the endocranial anatomy of the South American spinosaurine Irritator challengeri (SMNS 58022) from the Romualdo Formation (Lower Cretaceous: Aptian) of Brazil (Arai & Assine, 2020;Martill et al., 1996;Sues et al., 2002). Irritator shares similarities with the endocasts of other non-maniraptoran theropods and possesses unexceptional hearing capabilities; more interesting is the tentative evidence from the inner ear that suggest an ability to rapidly move and tightly control ventral movements of the head that may have aided in the capture of small, agile prey such as fish (Schade et al., 2020).
However, to date, no study has examined the endocranial morphology or neurosensory capabilities of baryonychine spinosaurids.
Here, we fill this knowledge gap via the description and interpretation of X-ray computed tomography (CT) scan data pertaining to Baryonyx walkeri and Ceratosuchops inferodios. Both are from Barremian strata of the Wealden Supergroup (Lower Cretaceous) of southern England, and both possess well-preserved (albeit partially disarticulated) braincases (Barker et al., 2021;Charig & Milner, 1986, 1997. Given the temporal and phylogenetic relationship of these specimens relative to Irritator, the baryonychines Baryonyx and Ceratosuchops may provide some context regarding the evolution of the spinosaurid endocranium and associated neurosensory capabilities. evolution of Spinosauridae did not require substantial modification of the brain and sensory systems.

| Terminology
Unlike that of birds or mammals, the brains of non-avian reptiles such as Alligator (Hurlburt et al., 2013), turtles (Evers et al., 2019) and squamates (Allemand et al., 2022) do not typically fill the endocranial cavity. As such, a reptilian endocast represents the total soft tissues within the braincase and only provides superficial information regarding brain topography (the extent of which depending on the species and neuroanatomical regions) (Allemand et al., 2022;Hopson, 1979;Hu et al., 2021). As such, we follow previous works by referring to the digital casts of the space within the braincase as "endocasts". Similarly, the segmented labyrinths do not truly represent the membranous or osseous features of the inner ear, and we also follow previous authors in adopting "endosseous" throughout when referring to the reconstructed structure (Witmer et al., 2008). Other segmented structures (e.g. neurovascular canals, pituitary fossae) are referred to as if they were the structures themselves.

CT scanning
The preserved braincase consists of three fragments (i.e. a large posterior basicranial fragment, a smaller skull roof fragment with portions of the frontals, parietals and left laterosphenoid, and the left otoccipital with its paroccipital process) that were scanned at OhioHealth O'Bleness Hospital in Athens, Ohio, USA, on a General Electric LightSpeed Ultra at 140 kV and 200-300 mA, with voxel sizes of 432 × 432 × 625 μm and using Extended Hounsfield and bow-tie filtration.

Volume reconstruction and image processing
Scan data were reconstructed using a bone algorithm and were exported as a DICOM image stack, which were then imported into Windows-based workstations running Amira-Avizo (Thermo Fisher Scientific) for analysis and segmentation. To generate 3D PDFs, 3D models were exported as OBJ files and imported into SimLab 3D Composer (Amman, Jordan).

Braincase rearticulation
The three preserved braincase fragments show some amount of plastic deformation and, although close, do not fit together perfectly. Different models were generated based on different optimization criteria: (1) best overall fit considering both bony structure and enclosed soft tissues (e.g. semicircular canals); (2) best fit optimizing bony structure, especially the otoccipital fragment; and (3) best fit optimizing the structure of the labyrinth, which is the most consistently symmetrical aspect of endocranial anatomy (Cerio & Witmer, 2019). The differences between the resulting models are slight, and Figures 1 and 2 presents the result of the second optimization criterion above. It is also worth noting that portions of the bony braincase overlying the left cerebral region were not preserved (or were lost during collection or preparation) but the underlying cerebral endocast is preserved in matrix as a natural endocast. (IWCMS 2014.95.1-3)

CT scanning
The three braincase fragments were scanned at the μ-VIS X-Ray Imaging Centre at the University of Southampton (UK), using a custom 225/450 kVp Hutch dual source walk-in micro-focus CT system (Nikon Metrology, UK). Peak voltage and current were set at 300 kVp and 250 μA respectively. A total of 3142 projections were collected over a 360° rotation, averaging 16 frames per projection with 250 ms exposure time per frame.

Volume reconstruction and image processing
Projection data were reconstructed as 32-bit float raw volumes with an isotropic voxel size using filtered back-projection algo- To improve contrast, the raw image files of the basicranium fragment (IWCMS 2014.95.3) were manipulated in FIJI (Schindelin et al., 2012) using the "sharpen" filter and background subtract function. Additional sharpening of the Ceratosuchops braincase material was conducted within Object Research Systems (ORS) Dragonfly (v. 2022.1, build 1249) via the Unsharp filter (Workflow > Image Filterin g > Operations > Sharpening) using a factor of 4-9 (standard deviation = 1), depending on the specimen/region of interest (ROI).

| Measurements
Morphological measurements were taken digitally using the measuring tools in Autodesk Meshmixer, ORS Dragonfly and FIJI (Schindelin et al., 2012).

| Reptile encephalisation quotient
The reptile encephalisation quotient (REQ) (Hurlburt, 1996) is a measure of relative brain size and gross cognitive capacities. We generated an estimated REQ based on the brain and body mass of the more skeletally complete Baryonyx type specimen. Body mass estimates are unknown for Ceratosuchops, but both taxa are of similar proportions where their known anatomy overlaps (Barker et al. (2021); see also Table 1), and the REQ estimated for Baryonyx likely approximates that of Ceratosuchops. The equation is as follows: F I G U R E 2 Cranial endocast of Baryonyx walkeri (NHMUK PV R9951), reconstructed from CT scans, in (a) right lateral, (b) anterior, (c) ventral, (d) posterior and (e) dorsal views. Vascular structures and endosseous labyrinth also depicted. Abbreviations: c, cochlea; car, cerebral internal carotid artery canal; cer, cerebral hemisphere; de, dorsal expansion; fl, floccular lobe; lab, endosseous labyrinth; otc, olfactory tract; pit, pituitary; pmcv, posterior middle cerebral vein canal; sin, blind dural venous sinus of the hindbrain; V, trigeminal nerve canal; VI, abducens nerve canal; VII, facial nerve canal, VIII co , cochlear ramus of the vestibulocochlear nerve; IX, glossopharyngeal nerve canal; X-XI, shared canal for the vagus, and accessory nerves, and accompanying vessels; XII, hypoglossal nerve canal. Asterisk (*) marks the disarticulated left otoccipital portion of the endocast. Scale bar: 50 mm.
where M Br is brain mass (in grams, excluding the olfactory tract and bulbs) and M Bd is body mass (in grams). Following Hurlburt et al. (2013), we assumed the brain occupied 37% or 50% of the endocast volume.
As the relative density of brain tissues nears unity, brain volume and mass are interchangeable (Hurlburt et al., 2013). Endocast volume was calculated in Meshmixer (Analysis > Stability).
A mass estimate for Baryonyx was generated based on the equation of Campione et al. (2014) implemented in Benson et al. (2018): where FC is minimum femoral circumference (in millimetres). Inputting the available femoral data for Baryonyx (minimum preserved circumference of the incomplete right shaft: 350 mm; S. Maidment, pers. comms., 2022) produced a mass estimate of 2011 kg. However, the measured Baryonyx femur is not only incomplete but also damaged (Charig & Milner, 1997), which affected the ability to collect reliable data (S. Maidment, pers. comms., 2022). Nevertheless, this estimate is comparable to the 1981 kg calculated by Therrien and Henderson (2007) using least-squares regressions involving skull length, and is used herein pending the discovery of better preserved material.

| Hearing frequency and range
The length of the endosseous cochlear duct (lagena of some authors) is often used as a proxy for hearing range and sensitivity (Hanson et al., 2021;Walsh et al., 2009). We employ the methods used by Walsh et al., 2009 in estimating hearing parameters. Two metrics are required: endosseous cochlear duct (ECD; pars cochlearis) length and basicranial length (BCL). The dorsal limit of the former is usually defined by a marked constriction where the ECD meets the saccule (Walsh et al., 2009): preservation does not allow us to determinate this with absolute confidence in Ceratosuchops but we were still able to provide an estimate. The measurement tools in Meshmixer were employed for both spinosaurids.
The constraints of the BCL were not outlined in Walsh et al. (2009) but this measure is defined as the anteroposterior distance between the anterior limit of the basisphenoid (excluding the cultriform process of the parasphenoid) and the posterior margin of the occipital condyle (Dudgeon et al., 2020). In Ceratosuchops, the basisphenoid and parasphenoid are indiscernibly fused together; the anterior limit of the basisphenoid was herein defined by following an obliquely trending line from the anterior margin of the basipterygoid process pedicel towards the region of the pituitary fossa in lateral view.
The ECD was scaled to the BCL and log transformed. For hearing range, we followed the equation: with x being the scaled and log-transformed ECD, and y being the best hearing range in Hz. For mean hearing frequency we followed the equation: with x being the scaled and log-transformed ECD length, and y being the mean best hearing frequency in Hz.
We also calculated the hearing capabilities of several other theropods for further comparisons, using the STL models of theropod inner

| Olfactory acuity
The form and external limits of the olfactory bulbs are difficult to determine in the Ceratosuchops endocast, being located at the anterior end of the expanding olfactory tract and separated by a midline sulcus in dorsal view (see below). An olfactory ratio was calculated by measuring the longest diameters of the bulbs and cerebral hemisphere (regardless of orientation); this is expressed as a percentage following previous studies (Zelenitsky et al., 2009(Zelenitsky et al., , 2011. Raw olfactory ratios should not be directly compared across taxa due to the influence of body size (Zelenitsky et al., 2009 ote: Linear measurements in millimetres (mm). Volume data in cubic centimetres (cm 3 ). Asterisk (*) denotes incomplete data due to preservation.
Ceratosuchops cannot be presently estimated but likely approximates that of Baryonyx (see above), we used the Baryonyx mass estimate employed in our REQ calculations and log-transformed both metrics. These were subsequently plotted onto the graph produced by Zelenitsky et al. (2009: Figure 2). Given the use of both specimens in this analysis, we provisionally consider the latter result representative of the generalised baryonychine condition.
However, our effort to rearticulate the braincase has introduced three morphological artefacts to the endocast. These are as follows: (1) a divot on the posterior dorsal expansion, which was presumably smoother in life given the morphology of other theropods (Sampson & Witmer, 2007;Witmer & Ridgely, 2009); (2) the bilateral, transverse sinus-like ridges posterior to the cerebral hemispheres do not represent cerebral topography but the supraoccipital-parietal contacts; and (3) an artificial "step" between the ventral laterosphenoidprootic contacts which, when combined with the loss of a portion of the right dorsum sellae (exposing some of the right abducens (CN VI) nerve trunk in cross-section), produces an anteriorly projecting ventral "lobe" on earlier iterations of the endocast near the right trigeminal (CN V) trunk. To avoid confusion or interpretation of this artefact as a genuine endocranial feature, this region of the endocast was bevelled in 3D view using the Polygon tool in ORS Dragonfly. This loss of bone and suboptimal laterosphenoid-prootic contact also rendered it difficult to visualise the ventral border of the trigeminal nerve trunk, which was inferred based on the preserved margins of the trigeminal foramen.
Elsewhere, the floccular lobes could not be reconstructed in full: the left lobe exists as two disjointed pieces and the lateral extent of both cannot be determined. Image contrast issues mean that the posterior dorsal head veins were difficult to trace, while the secondary, more ventral passages for the hypoglossal (CN XII) nerve trunks could only be traced over a short distance. The left vestibulocochlear nerve trunks (CN VIII) could not be visualised.
The endosseous labyrinths are partially preserved (Figures 3, 4b and 5e-h). The left side is missing its vestibule, cochlear duct and portions of the anterior and lateral semicircular canals, which could not be visualised due to the lack of contrast and presence of a substantial quantity of radiopaque matrix in the region. The right labyrinth is more complete and forms the basis of the structure's description and comparison, although the midsection of the anterior semicircular canal (within the prootic) could not be segmented (this was also due to the presence of radiopaque matrix obscuring its path). The posterior margin of the cochlear duct could not be reconstructed due to disarticulation of the prootic and otoccipital.

| Endocranial morphology
Baryonychine endocasts (Figures 2 and 3) conform to the general morphology of other non-coelurosaurian theropods, the details of which are discussed below, and corroborate previous suggestions of endocranial conservatism amongst the taxa concerned (Hopson, 1979;Sampson & Witmer, 2007). As is typical for most non-maniraptoriform sauropsids, the brain did not fill the endocranial cavity, with much obscured by a dural envelope and its extensive venous sinuses (Hopson, 1979;Sampson & Witmer, 2007;Sedlmayr, 2002;Witmer et al., 2008;Witmer & Ridgely, 2009). To facilitate interspecific comparisons, the endocasts will be oriented with the lateral semicircular canal assuming a horizontal path.
The olfactory tract is well preserved in Ceratosuchops. It is an elongate, narrow structure that expands anteriorly towards the olfactory bulbs. The anterior portion is indented along its dorsal midline by a shallow sulcus, as in Irritator (Schade et al., 2020), likely marking the position of the olfactory bulbs. The tracts are in line with the forebrain in both these specimens, and their elongate proportions conform to the plesiomorphic archosaurian condition present in many non-avian theropods (Bever et al., 2013;Franzosa, 2004); though incomplete, the base of the olfactory tract is also present in Baryonyx, and a similar morphology is thus inferred. The spinosaurids lack evidence of medial separation of their respective bony olfactory tracts, as in Majungasaurus (Sampson and Witmer, 2007)  Although the brain does not completely occupy the endocranial cavity, the portion of the endocast corresponding to the telencephalon is considered to approximate the underlying brain contours, especially laterally (Sampson & Witmer, 2007;Sedlmayr, 2002;Witmer et al., 2008); the cerebral hemispheres can be clearly observed. The studied baryonychines possess the relatively small and unexpanded ancestral morphology typical of nonmaniraptoriform theropods (Franzosa, 2004;Larsson et al., 2000;Rauhut, 2003). Despite the limited expansion, the widest part of the endocast appears located at the level of the cerebral hemispheres, as is typical for fossil reptiles (Hopson, 1979) and observed in such theropods as Irritator (Schade et al., 2020), abelisaurids (Paulina- (Paulina-Carabajal & Currie, 2017) and allosauroids such as Allosaurus (Rogers, 1998) and Sinraptor (Paulina-Carabajal & Currie, 2012).
The forebrain and hindbrain are generally horizontally oriented in both baryonychines, as is typical for non-coelurosaurian theropods and also early-branching coelurosaurs such as tyrannosaurids (Witmer & Ridgely, 2009). A nuance is that the forebrain is directly slightly anteroventrally along its length in Ceratosuchops.
Instead, these potentially represent anteriorly diverging venous channels emanating from the dorsal expansion.
The dorsal expansion of the endocast -an eminence just posterior to the cerebrum -has been thought to house (at least in part) the pineal apparatus in non-coelurosaurian theropods, although the expansion is absent in more basal theropods (Sampson & Witmer, 2007;Witmer & Ridgely, 2009). However, recent work on extinct and extant turtles suggests that a cartilaginous origin, rather than one relating to the pineal gland also merits consideration (Werneburg et al., 2021), as does a venous interpretation as the dorsal expansion is associated with a variety of clearly venous structures in a variety of extinct and extant diapsids (Porter & Witmer, 2015Witmer et al., 2008;Witmer & Ridgely, 2009).
The dorsal expansion is pronounced and ascends above the dorsal margin of the forebrain in Baryonyx, in contrast to Ceratosuchops.
As such, the dorsal margin of the latter's forebrain assumes a largely linear trend in lateral view, contrasting against the more concave margin in Baryonyx. As in Majungasaurus and Allosaurus (Sampson & Witmer, 2007), the dorsal expansion in Baryonyx is located within the parietals, whereas Ceratosuchops recalls the condition in tyrannosaurids, where the apex is situated at the parietal-supraoccipital suture (Bever et al., 2013). Variation in the development of the dorsal expansion has been noted in Tyrannosaurus; however, it is likely that this structure is non-homologous with that of non-coelurosaurian taxa (Witmer & Ridgely, 2009). Damage to this region in Irritator The dorsal expansion has also been used to approximate the size of the midbrain region in theropods, with many non-coelurosaurian tetanurans and ceratosaurs possessing elongated midbrain regions (Paulina-Carabajal & Currie, 2017). The sampled baryonychines follow this trend (Table 2), although the elongation is less marked compared to such forms as Majungasaurus or Allosaurus.
The optic lobes (optic tecta) in the baryonychines are imperceptible, recalling the plesiomorphic sauropsid condition (Franzosa, 2004;Schade et al., 2020); this is also the case in Irritator (Schade et al., 2020). These lobes are equivocally visible in some Tyrannosaurus specimens, and a trend whereby these structures become increasingly obvious is observed within coelurosaurs on the line to birds (Witmer & Ridgely, 2009). The course of the transverse sinus and middle cerebral vein can be used to identify the gross position of the optic lobes and cerebellum, given that these structures pass between these brain regions in extant sauropsids (Sampson & Witmer, 2007;Witmer & Ridgely, 2009) Abbreviations: c, cochlea; fl, floccular lobe; fo, fenestra ovalis; lab, endosseous labyrinth; pit, pituitary; sin, blind dural venous sinus of the hindbrain; V, trigeminal nerve canal; VI, abducens nerve canal; VII hym , hyomandibular ramus of the facial nerve; VII pal , palatine ramus of the facial nerve; VIII vest , vestibular ramus of the vestibulocochlear nerve; IX, glossopharyngeal nerve canal; X-XI, shared canal for the vagus and accessory nerves, and accompanying vessels; XII, hypoglossal nerve canal; ?, potential accessory hypoglossal nerve or venous canal. Scale bar: 10 mm. Witmer, 2007) but likely contained various other soft tissue structures (Hopson, 1979), is a largely vertically oriented structure located just anterior to the level of the dural peak when viewed laterally. It produces a pair of dorsal and ventral posterior projections in the baryonychines, imparting a "wavy" posterior margin in lateral view. In Irritator, the posterior margin is more uniformly convex (Schade et al., 2020), although variation in the shape of the pituitary fossa is known amongst well sampled theropods such as Tyrannosaurus (Bever et al., 2013). A minor degree of posterior angulation is also noted in the Irritator pituitary, which would follow the conservative theropod pattern (Bever et al., 2013); baryonychines, however, do not appear to deviate substantially from the vertical.
Within the region of the pituitary fossa, Ceratosuchops possesses the plesiomorphic cavernous sinus present in archosaurs and many non-coelurosaurian theropods through which pass the trochlear (CN IV) and abducens (CN VI) nerves, their associated vasculature, and likely the encephalic branches of the cerebral carotid artery (Bever et al., 2011;Sampson & Witmer, 2007;Witmer & Ridgely, 2009). This feature could not be discerned in Baryonyx due to preservation. The floccular lobes (part of the cerebellum) are distinguishable, however, as is typical for theropods (Franzosa, 2004) and dinosaurs more generally (compared to extant non-avian reptiles) (Witmer et al., 2003). The lobes themselves, best visualised on the wellpreserved right side of the Baryonyx endocast, display the mediolaterally thin and somewhat triangular/tabular lateral morphology of many non-maniraptoriform theropods, lacking the bulbous derived condition (Franzosa, 2004;Paulina-Carabajal & Currie, 2017;Witmer & Ridgely, 2009). Like Irritator, these project posterolaterally from the endocast, passing the plane of the anterior semicircular canal to occupy the space delineated by these vestibular structures, as in many theropods other than adult tyrannosaurids (Witmer & Ridgely, 2009), whose floccular lobes barely project into this space Equivalently positioned "blind dural sinus of the hindbrain" have been documented in an indeterminate theropod (Knoll et al., 1999), Tyrannosaurus (Witmer & Ridgely, 2009) and the sauropod Spinophorosaurus (Knoll et al., 2012), and a similar feature was tentatively identified as an "endolymphatic duct" in Murusraptor (Paulina-Carabajal & Currie, 2017). Rounded protuberances in this region are nevertheless noted in many archosaurs, where they have been interpreted as diverticula of the longitudinal sinus (Hopson, 1979).

| Cranial nerve trunks
Cranial nerve trunk organisation is highly conserved within Dinosauria (Hopson, 1979;Witmer & Ridgely, 2009), and the baryonychine endocasts do not deviate from this general pattern (Figures 1-3). As above, however, cranial nerves I-IV are unfortunately not preserved in Baryonyx, although the more posterior nerve trunks can all be readily distinguished. We reiterate previous works in noting that veins, as well as nerves, likely passed through the cranial nerve foramina, as is the case in extant archosaurs (Sampson & Witmer, 2007;Sedlmayr, 2002), such that segmented canals likely contained both soft tissue structures.
In Ceratosuchops, both the olfactory bulbs and tract and optic nerve (CN II) trunk are situated medially and project anteriorly from the endocast. The latter is large and undivided in contrast to the condition present in some abelisaurids, whose optic nerve trunk may be subdivided by the calcified interorbital septum (Sampson & Witmer, 2007). The small, ovate trochlear nerve (CN IV) trunks are located ventral to the cerebral hemispheres. These are usually situated dorsally to the oculomotor nerve (CN III) trunks in archosaurs (Hopson, 1979); however, independent foramina for the latter could not be distinguished. The trochlear nerve trunks identified in Irritator (Schade et al., 2020;Sues et al., 2002) are too dorsally located to correspond to this cranial nerve, which always passes behind and then ventral to the optic lobe in extant taxa (Witmer & Ridgely, 2009).
Instead, these most likely correspond to passages for orbitocerebral veins, as observed in Majungasaurus (Sampson & Witmer, 2007). laterally projecting coelurosaur condition (Bever et al., 2011;Witmer & Ridgely, 2009). This more lateral course results from the loss of the above-mentioned cavernous sinus in members of the latter clade, although it is present in some tyrannosauroids (Bever et al., 2011, Witmer & Ridgely, 2009).
The facial nerve (CN VII) trunk becomes dorsoventrally expanded as it approaches its lateral exit through the prootic. Dorsoventral elongation of the associated external foramen has been recovered as a spinosaurid synapomorphy in some studies (Barker et al., 2021).
The associated palatine and hyomandibular rami most probably di- Baryonyx and Ceratosuchops thus possess the derived theropod condition (Rauhut, 2003;Sampson & Witmer, 2007), which is also observed in Irritator. The subdivision of the cavum metoticum is incomplete in the baryonychines and the crista tuberalis does not meet the lateral endocast wall; the medial aperture of the cavum is thus undivided, as in most archosaurs (Bever et al., 2013).
Only a single hypoglossal nerve (CN XII) trunk is bilaterally present in Baryonyx (Figure 2d): this opens within a common fossa on the occiput alongside the vagal canal. Ceratosuchopsin spinosaurids, in contrast, possess a smaller third external foramen located ventrally to the vagal and hypoglossal foramina (and thus display three external foramina penetrating the occiput) (Barker et al., 2021). The path of the corresponding third canal can only be traced a short distance in Ceratosuchops, and its identity is unclear ( Figure 3d); it may represent an accessory hypoglossal nerve trunk or a venous canal (Witmer & Ridgely, 2009 & Witmer, 2007) and Alioramus (Bever et al., 2013), whilst two are observed in Dilophosaurus (Marsh & Rowe, 2020) and Allosaurus (Hopson, 1979), though we note that Franzosa (2004) scores the latter taxon as possessing only a single foramen in specimen UMNH VP 18055 (previously UUVP 5961). Two hypoglossal canals are also present in various maniraptoriforms (Franzosa, 2004;Lautenschlager et al., 2012).
The presence of three external foramina on the occiput was deemed synapomorphic for megalosauroids in a previous phylogenetic analysis, the spinosaurid condition whereby two are present being regarded as a reversal (Carrano et al., 2012). other saurischians (Rauhut, 2003).
Ventrally, the canals of the internal carotid arteries unite anteriorly prior to entering the posteroventral pituitary fossa, imparting a V-shaped morphology in the baryonychines when viewed ventrally. Such organisation is unusual among non-coelurosaurian theropods: certain ceratosaurs, allosauroids and early neotheropods possess carotid canals that enter the pituitary fossa separately (Franzosa, 2004;Marsh & Rowe, 2020;Sampson & Witmer, 2007). The condition in Irritator cannot be ascertained (Schade et al., 2020), but that in Baryonyx and Ceratosuchops recalls the condition observed in Tyrannosaurus and several other coelurosaurs (Franzosa, 2004), as well as Giganotosaurus (Paulina-Carabajal & Canale, 2010).

| Endosseous labyrinth
The inner ear can be grossly subdivided into two mechanosensory organs that are respectively responsible for the detection of head movements and sound -the vestibular labyrinth (composed of the three semicircular canals, the utriculus and sacculus) and cochlea (which includes the cochlear and perilymphatic ducts) (Bronzati et al., 2021;Hanson et al., 2021;Pfaff et al., 2019;Wever, 1978;Witmer et al., 2003). Despite representing a (slightly) smaller individual, the Baryonyx inner ear is a slightly larger structure compared to the reassembled organ of Ceratosuchops, but the disarticulation experienced by the latter's braincase may have impacted some of its dimensions ( Figure 5). Nevertheless, the semicircular canals are also slightly larger in diameter in Baryonyx.
The baryonychine semicircular canals are generally thin structures organised roughly orthogonally relative to one another, as seen in other theropods (Sampson & Witmer, 2007;Witmer & Ridgely, 2009). The vertical semicircular canals are asymmetricala common feature among theropods including Irritator (Bever et al., 2013;Schade et al., 2020). The anterior semicircular canal is tall -a typical morphotype amongst avemetatarsalians (Hanson et al., 2021) -and like other dinosaurs (Sampson & Witmer, 2007) The shorter posterior semicircular canal in both baryonychines forms a simple, near-vertically oriented arc, generally comparable to the non-coelurosaurian condition (Bever et al., 2013). Its parasagittal orientation, which helps create a sub-triangular space between the three semicircular canals in lateral view, is representative of the plesiomorphic theropodan condition observed amongst various non-maniraptoriform theropods (Franzosa, 2004), including Irritator (Schade et al., 2020). Baryonychines also lack the anterior bowing of the posterior canal observed in tyrannosaurids and most other coelurosaurs (Witmer & Ridgely, 2009).
The lateral semicircular canal in Baryonyx, like that of Irritator, is not particularly bowed laterally in dorsal view: its anterior portion assumes a largely linear initial trajectory. This is again typical of non-coelurosaurian theropods (Bever et al., 2013), and the canal in both spinosaurids only "hooks" towards the secondary common crus in their posterior halves. Allosaurus produces a similar (if not slightly more exaggerated) posterior "hook" (Rogers, 1998;Witmer & Ridgely, 2009). In comparison, the lateral semicircular canal of Ceratosuchops is more uniformly bowed along its length, although the arc through which it sweeps is not as broad as that of various coelurosaurs (Witmer & Ridgely, 2009). Nevertheless, distinction between the posterior and lateral semicircular canals is impossible to determine in posterior view in any of the above-mentioned spinosaurids, a characteristic of non-maniraptoran theropods (Witmer & Ridgely, 2009).
The vestibule is typically archosaurian (some exceptions not withstanding) in failing to extend dorsally above the level of the lateral semicircular canal (Bever et al., 2013;Sampson & Witmer, 2007;Witmer & Ridgely, 2009). The ventrally projecting cochlea is relatively long, as in archosaurs generally (Hanson et al., 2021), and accounts for approximately two-thirds of the dorsoventral height of the vestibular labyrinth in both baryonychines. Intriguingly, this is different from the subequal proportions described for Irritator (Schade et al., 2020) and the relatively short cochlea of abelisaurids (Cerroni & Paulina-Carabajal, 2019). A subtle degree of medial inclination is also observed in Baryonyx and Ceratosuchops when viewed anteriorly, as in Irritator (Schade et al., 2020) and tyrannosaurids (Witmer & Ridgely, 2009). In comparison, the medial inclination of the cochlea is more exaggerated in Sinraptor and Murusraptor : Data S2). A small, dorsally situated lateral projection on the cochlear duct in both baryonychines marks the passage from the fenestra ovalis, through which the columella passes.

| Basicranial pneumaticity
Details of the braincase pneumaticity will be presented elsewhere; here, we include brief comments given their effects on hearing capabilities, which are discussed below. Relative to that of some other theropods such as tyrannosaurids (Witmer & Ridgely, 2009) An indentation on the posterior basioccipital and ventral to the occipital condyle is present in both taxa, which varies substantially in form and development. This was incorrectly referred to as a "subcondylar recess" in Ceratosuchops in Barker et al. (2021); however, such a recess usually refers to paired structures and is only present in a few theropod clades such as tyrannosaurids and ornithomimosaurs (Witmer, 1997). It remains unclear whether these indentations pertain to an excavation by a pneumatic system or are simply epiphenomena related to the elevation of the muscular ridges that delimit them laterally.

| DISCUSS ION
The cranial endocasts of the Wealden Supergroup baryonychines provide insight into the early evolution and anatomy of the spinosaurid endocranium. Like Irritator (Schade et al., 2020), both Baryonyx and Ceratosuchops possess neuroanatomical features typical of their phylogenetic position as non-coelurosaurian tetanurans. These include the possession of weakly demarcated brain regions, prominent cranial flexures, mediolaterally thin, "tabular" floccular lobes, and asymmetrical vertical semicircular canals.
With the above-described caveats regarding the body mass estimation used in our assessment of relative encephalisation, the calculated REQ for Baryonyx is approximately between 1.2 (REQ 37% ) and 1.6 (REQ 50% ). Some question exists on the utility of encephalisation quotients (Balanoff & Bever, 2020); nevertheless, we note that the degree of encephalisation calculated for large predatory theropods is generally comparable, the exception being select tyrannosaurids (Table 3) (Hopson, 1979) and Tyrannosaurus (Witmer & Ridgely, 2009) have been described as "very similar" or "[showing] little variation". Variation, when present, is possibly taphonomic or ontogenetic in nature, but studies examining this variation is rare (Lautenschlager & Hübner, 2013) or speculative (McKeown et al., 2020), and our understanding of its influence between and within taxa remains preliminary (Hu et al., 2021).

| Vision and gaze stabilisation
As with many other studied theropods (Witmer & Ridgely, 2009), the imperceptible optic lobes in both baryonychine endocasts provide little information regarding visual acuity or sensitivity. The floccular lobes, however, may provide some information on the visual system.
Integrating information from the latter as well as the vestibular systems of the inner ear (Voogd & Wylie, 2004), the floccular lobes regulate gaze stabilisation via compensatory movements of the eyes in response to rotation of the head (vestibulo-ocular reflex, VOR), help track moving objects within the field of view (smooth pursuit), and (in some taxa) stabilise the head via recruitment of the cervical musculature (vestibulo-collic reflex, VCR) (Ferreira-Cardoso et al., 2017;Ito, 1982;Witmer et al., 2003).
Relative floccular lobe size has been used (in part) to reconstruct theropod ethology, where enlargement has been qualitatively correlated with increased capacity for gaze stabilisation (Cerroni & Paulina-Carabajal, 2019;King et al., 2020;Lautenschlager et al., 2012). The relatively large floccular lobe in Irritator (compared to other non-coelurosaurian theropods but not many coelurosaurs) was regarded as evidence of improved performance of the VOR and VCR systems (Schade et al., 2020).
However, the size of this structure appears to correlate negatively with body size in theropods . Size may also reflect differing ontogenetic status, as has been noted during tyrannosaurid ontogeny, for example, and the small condition in mature individuals may not necessarily reflect of the size TA B L E 3 Reptile Encephalisation Quotients (REQ) calculated for a range of theropods.
We note, however, that the assessment of ontogenetic status in non-coelurosaurian theropods is complicated (Griffin et al., 2020), and as a result, the relative age of the above spinosaurids has not been rigorously determined. Further clouding the interpretation of floccular lobe size is the fact that the reconstructed structures may not be indicative of the neural tissues themselves, since additional tissues were likely also present alongside the floccular lobes (Ferreira-Cardoso et al., 2017;Walsh et al., 2013).
The more complete right floccular lobe of Baryonyx is grossly comparable to other non-coelurosaurian tetanurans in shape and extent. With the above caveats in mind and assuming the preserved lobe contained comparable amounts of neural tissue per unit volume, this similarity crudely implies Baryonyx possessed similarly developed gaze stabilisation mechanisms, and that these were potentially less developed than in Irritator.
Although beyond the scope of this study, some preliminary inferences regarding the visual capabilities of Ceratosuchops can be additionally made via the skeletal elements. Rearticulation of the holotype braincase (IWCMS 2014.95.1-3) and the referred and mirrored postorbital (IWCMS 2014.95.4) shows that the orbits appear more anteriorly facing compared to some other theropods (e.g. Ceratosaurus), which display the plesiomorphic, laterally facing condition (Zelenitsky et al., 2009). If correct, this could indicate a higher degree of stereoscopic vision, which perhaps aided in to the calculation of prey position during hunting. However, we note that stereopsis can evolve in animals with laterally facing orbits, and that it may not necessarily be present in animals with binocular overlap (Nityananda & Read, 2017).

| Hearing
The endosseous cochlear duct (housing the cochlea or basilar papilla), is closely associated with hearing performance (sensitivity and frequency range) Walsh et al., 2009). Indeed, longer cochlear ducts provide improved sensitivity to low-frequency sounds (Gleich et al., 2005;Walsh et al., 2009). The auditory capabilities of the Wealden Supergroup baryonychines, as well as various other saurischians, are presented in Table 4.
Archosaurs display elongate cochlear ducts relative to most other reptiles (Hanson et al., 2021). An increased degree of cochlear ducts elongation is observed in several theropods and is potentially related to adaptations for auditory foraging .
Although some works have critiqued the use of Walsh et al.'s (2009) equations (Knoll et al., 2021), they have been commonly used as an approximation of hearing capabilities in extinct taxa and provide a basis for comparisons (Table 4). Previous works estimated that most saurischians have an optimal hearing frequency between 3500-5500 Hz (akin to modern birds) and best mean hearing frequencies between 2250-3250 Hz (Hanson et al., 2021), although the extinct sample was biased towards smaller maniraptoran taxa. However, Gleich et al. (2005), using a different method, suggested that large dinosaurs were capable of low-frequency perception up to 3000 Hz. The baryonychines share gross hearing capabilities in common with other large theropods such as abelisaurids, allosauroids and tyrannosauroids in having lower-pitched optimal frequency ranges ( Table 4). Baryonyx and Ceratosuchops also both possessed lower estimated auditory capabilities relative to those reported for Irritator (Schade et al., 2020), suggesting differing hearing ecologies between the clades.
The lack of extensive tympanic pneumaticity in either baryonychine also suggests that, like Irritator (Schade et al., 2020), the middle ear system was not as specialised for the reception of low-frequency sounds as those of tyrannosaurids (Witmer & Ridgely, 2009). Extensive pneumaticity (and thus volume of these sinuses) in this region affects the impedance-matching capabilities of the middle ear and permits the emphasis of low-frequency sounds (Witmer and Ridgely (2009), and references therein), which spinosaurids were seemingly less reliant upon.

| Equilibrium and head posture
Semicircular canal shape has been previously used to infer aspects of spinosaurid ecology: Schade et al. (2020) discussed the possible behavioural ecology of Irritator based on the elongate anterior semicircular canal, which they considered to impart enhanced sensitivity to pitch-down movements of the head. The morphology of the vestibular labyrinth observed in both Irritator and the Wealden Supergroup baryonychines (i.e. vertically tall anterior semicircular canals) meanwhile resembles the bipedal locomotor morphotype described in Hanson et al. (2021). While the bipedal nature of baryonychines is now uncontroversial, it is of historic interest that Baryonyx was initially considered to be quadrupedal (Charig & Milner, 1986), a stance that was revised in subsequent works (Charig & Milner, 1997). However, it is becoming increasingly difficult to support correlations between semicircular canal geometry and specific ecological or locomotor functions, which may instead be linked to spatial and developmental constraints (Benson et al., 2017;Bronzati et al., 2021;David et al., 2022;Evers et al., 2022). We note that some quadrupedal dinosaurs, such as Spinophorosaurus (Knoll et al., 2012), also possess vertically tall anterior semicircular canals, as do some pterosaurs such as Anhanguera (Witmer et al., 2003), and increases in relative height of the anterior semicircular canal may not only be affected by bipedal locomotion but also erect limb postures (Bronzati et al., 2021). Taken together, the above inferences regarding spinosaurid locomotion/ecology and semicircular canal geometry are incompatible with the current consensus.
Nevertheless, independent support for a down-turned position of the skull in Irritator was provided by the slight ventral rotation of the occipital condyle (Schade et al., 2020;Sues et al., 2002). Such rotation is absent in Baryonyx and Ceratosuchops, with the condyle projecting largely posteriorly in the baryonychines. Thus, regardless of lateral semicircular canal orientation, the craniocervical articulation suggests differing "standard" head postures within Spinosauridae.

| Implications of baryonychine sensory systems
Consensus exists that spinosaurids exhibit specialisation for some (almost certainly varying) degree of aquatic or semiaquatic adaptation. This view has to be evaluated within the broader debate on which endocranial features are indicative of semi-aquatic or aquatic behaviour amongst archosaurs. Some studies suggest that shape changes in the inner ear of various amniotes correlate with transitions from terrestrial to semi-and fully-aquatic ecologies (Neenan et al., 2017;Schwab et al., 2020;Spoor et al., 2002), for example. Others, as mentioned above, find no evidence for change in labyrinth shape being associated with such transitions and, furthermore, question previous form-function inferences based on vestibular geometry (Bronzati et al., 2021;Evers et al., 2022).
Furthermore, derived ecologies can arise in the absence of significant shifts in endocranial anatomy, as observed in the extant diving bird Cinclus (Passeriformes: Cinclidae), whose endocasts and inner ears are very similar to non-diving passeriform relatives (Smith et al., 2022).
The grossly comparable endocranial morphologies and REQ scores, relative to other non-coelurosaurian tetanurans and ceratosaurs, suggests that baryonychines shared similar cognitive capacities with terrestrial, hypercarnivorous forms. This might indicate that spinosaurids were pre-adapted in terms of neural and sensory adaptations for the exploitation of aquatic prey, or that modification of the rostrum and dentition were sufficient for successful prey capture in aquatic settings and that substantial reorganisation of the endocranium was apparently not required. Further, the unremarkable hearing and olfactory capabilities inferred here also indicates that baryonychines did not deviate substantially from the auditory and olfactory capabilities of "typical" non-coelurosaurian theropods.
The tactile capabilities of the spinosaurid rostrum have also been deemed important in the context of prey detection and capture in some previous works, with neurovascular networks and patterning of the associated external foramina perhaps imparting crocodilelike sensitivity to prey movement in water (Ibrahim et al., 2014).
Tactile sensation of the face and jaws of vertebrates is detected by the trigeminal nerve (CN V) (Schneider et al., 2016;Vermeiren et al., 2020); however, the available endocranial data for spinosaurids suggest, qualitatively at least, that there is little evidence of an emphasised trigeminal nerve system in these theropods, as is consistent with previous observations of other dinosaur endocasts (Porter & Witmer, 2020). Difficulties persist in collecting the relevant quantitative data in theropod dinosaurs, such as trigeminal ganglion volume, a proxy for facial sensation in crocodyliforms and possibly other archosaurs (George & Holliday, 2013;Lessner & Holliday, 2022). Meanwhile, measurements based on neurovascular foramen size alone would be unreliable due to the likely passage of concomitant vasculature (see above). Whilst some tetanurans with clear adaptations for terrestrial predation possessed rostral neurovascular features akin to spinosaurids and suggestive of tactile jaw edges (Barker et al., 2017;Kawabe & Hattori, 2022), there is no a priori reason to assume theropods had enhanced sensitivity around the oral margin (Porter & Witmer, 2020). Nevertheless, spinosaurids may prove to be an exception (Porter & Witmer, 2020), with the degree of neurovascular branching perhaps more developed relative to other tetanurans (Bouabdellah et al., 2022). Further work, perhaps focusing on the branching pattern of canals within the dentary (which may provide the clearest signal for facial tactile sensation) (Lessner, 2021), for instance, is clearly needed before rostral neurovascular specialisations can be inferred for the clade.
Intriguingly, low-frequency hearing in the tyrannosaurid Alioramus has been tentatively associated with the detection of abnormally low juvenile vocalisations, a possible lack of parental care, or an adaptation for the detection of sounds made by large prey (Hanson et al., 2021).
We show that using a larger sample of basal tetanurans and ceratosaurs that low-frequency hearing is not autapomorphic to Alioramus and is seen in a range of large predatory theropod taxa (Table 4). It would be extremely speculative to link this with inferred parental behaviour (or the lack of it), and dietary characteristics related to cochlear duct dimensions failed to produce significant signals (Hanson et al., 2021). Nevertheless, low-frequency hearing in baryonychines as an adaptation for detecting large prey would not corroborate with evidence of spinosaurids as predators of generally small vertebrates (Hone & Holtz Jr, 2017;Therrien et al., 2005), as shown by their relatively weak bite force for instance (Sakamoto, 2022). Indeed, small prey items were likely commonly selected and depredated by carnivorous theropods in general (Hone & Rauhut, 2010).

| CON CLUS IONS
Baryonyx and Ceratosuchops possess well-preserved endocranial features, and provide insight into the evolution of the brain and sensory systems in earlier branching spinosaurids. The overall morphology of the reconstructed endocasts is reflective of their phylogenetic TA B L E 5 Olfactory ratios of select theropods. Murusraptor c 45-50 Ornithomimosauria 28.2-32.5 Oviraptoridae 31.5 Dromaeosauridae

28.5-36
Troodontidae 32.6-33.5 Note: Data from Zelenitsky et al., (2009Zelenitsky et al., ( , 2011, and supplemented by a Cerroni and Paulina-Carabajal (2019) position and follows previous observations of endocranial conservatism in non-maniraptoriform theropods. Similarly, the morphology of the endosseous labyrinth, and in particular the shape of the semicircular canals, is comparable to other non-coelurosaurian tetanurans.
REQ values imply that baryonychines did not deviate substantially in terms of cognitive capacity and behavioural sophistication relative to other basal theropods, and these predators possessed unexceptional hearing and olfactory capabilities on par with several other large-bodied terrestrial theropods. There is no tangible evidence for adaptations to semi-aquatic ecologies in the baryonychine endocrania, suggesting that neural and sensory systems of spinosaurids were pre-adapted for the successful detection and capture of aquatic prey, or that the initial transition to semi-aquatic ecologies simply required modifications of the craniodental apparatus. CTB reconstructed the full endocast and associated neurovasculature. RR and LW reconstructed the Baryonyx endocast and wrote the associated segmentation methodology and scanning protocol.
All authors edited the manuscript.

ACK N O WLE D G E M ENTS
We would like to thank:

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on