Sella turcica.

A dorsally projected indentation of the ventral endocast slightly anterior to the level of the pineal foramen (endocast impression: feature 11 in Fig 2) results from the contact between the basioccipital and the basisphenoid forming an incipient yet poorly preserved sella turcica on the basioccipital [22]. In TW480000375, the sella turcica is not anteriorly delimited by an anterior clinoid process but continues smoothly onto the parasphenoid. Its limited preservation does not allow for a detailed assessment of this important braincase structure, nor of the perforations known to irrigate it with blood or innervate it, as was demonstrated for the placodont sauropterygian Placodus gigas [35]. Well-defined projections of the hypophysis or pituitary lobes are absent on the ventral aspect of the endocast. Such an undifferentiated hypophysis also characterizes the cranial endocast of crocodylians and turtles [56–60]. The complex division of the “cavité hypophysaire” that was described by Gorce [61] cannot be recognized in TW480000375, likely due to the damage sustained in this domain. Nevertheless, the generic configuration of the sella turcica in Nothosaurus has been described in detail (e.g. [22]). A pituitary fossa was initially considered absent [47]. However, when it was eventually identified, its morphology was explained to have potentially accommodated an underdeveloped hypophysis relative to those of Sphenodon and lizards [56], where the hypophysis occupies a deep recess. In Sphenodon, Varanus and Lacerta, the lateral periphery of the sella turcica around the crista trabecularis serves as the origin for important oculomotor musculature, notably the m. bursalis and the retractor bulbi group [62]. The ocular muscle configuration of Crocodylia is similar to that of Sphenodon [63]. Furthermore, a low dorsum sellae and shallow sella turcica have been correlated with a reduction of the eyes and ocular muscles in reptiles ([64] and references therein). Nothosaurus exhibits a low dorsum sellae that is positioned quite posteriorly and well behind the intertemporal constriction of the braincase (see also Fig 10 in [22]). Ocular muscles attaching to the basisphenoidal region must thus have reached a considerable distance to reach the eyes (further even than the “extremely long” optic nerves [21]), including the passage of the laterally constricted section of the braincase in between the temporal fenestrae. Where considerable relative orbital size in Nothosaurus marchicus may indicate substantial visual acuity, this indirect pathway for ocular musculature also suggests restricted oculomotor dexterity and thus a relatively fixed ocular orientation in the cranium. However, the positioning of the orbits and relatively elongated neck in Nothosaurus could have granted partial binocular overlap and sufficient cervical agility (possibly even head saccades, as birds [65]) to compensate for poor intracranial ocular mobility. In other words, the constrained ocular musculature was probably compensated for by cervical mobility and a refined vestibulocollic reflex. Basal members of the genus appear to exhibit a relatively larger orbital size associated with shorter temporal fenestrae than more advanced forms that exhibit significant relative cranial elongation (e.g. [6]).

Cava epipterica.

In the cerebral domain immediately posterior to the pineal foramen, the central to ventral aspects of the cranial endocast are laterally constricted and consequentially well defined (endocast impression: feature 10 in Fig 2) through the presence of the medially bulging epipterygoids. The ventrolateral margins of this domain are formed by distinct, anteroposteriorly trending ventral sulci that gradually even out onto the pterygoidal cranial floor anteriorly. More posteriorly, these sulci give rise to paired semitubular passages (feature 12 in Fig 2) that posterolaterally depart from the ventrolateralmost margins of this endocast domain at a small angle. The epipterygoids form the lateral walls of the cava epipterica; enclosed, paired cerebral compartments that are roofed by the palatines, floored by the pterygoids, and separated by a cartilaginous median septum in vivo [22]. Although the straight braincase endocast of TW480000375 exhibits a nearly flat ventral margin that prevents recognition of the transition from the forebrain to the inconspicuously expressed midbrain through a distinct flexure, the cava epipterica provides a more general indication for the position of this transition.

“Foramen eustachii” of Koken (1893).

The basioccipital forms the occipital condyle and the ventral margin of the foramen magnum. It meets the exoccipitals dorsolaterally along a smooth suture and shares an irregular and locally strongly interdigitating suture with the paired pterygoids ventrally. More anteriorly up to the level of the epipterygoids, the basioccipital-pterygoidal contact is deeply depressed. It continues anteriorly along the unossified lateral walls of the braincase where it gives rise to a paired, dorsally open trough (interosseous casts: feature 18 in Fig 2). More anteriorly still, these troughs pass ventral to the otic domain and join the cranial endocast at the level of the basisphenoid. Posteriorly, these troughs develop into the large, posteroventrally widening occipital gaps that result from the natural separation between the lateral basioccipital tuber and the pterygoids. These occipital corridors flanking the basioccipital have been argued to represent the passages of the eustachian tube in Nothosaurus [47], but have also been interpreted to (partly) accommodate the internal carotid arteries [22]. Furthermore, their variable configuration in Nothosaurus has been argued to indicate closure of these passages at a late ontogenetic stage [22]. In TW480000375, the trajectory of these “eustachian foramina” does not involve the internal carotids, nor does it link the pharynx with a potential middle ear cavity, as is the case for a true eustachian passage [66]. In addition, the depressed basioccipital-pterygoidal suture that distally forms the lateral troughs does not correspond to a conventional cerebral compartment but represents a dorsally unossified recess of the basicranium with an unknown function.

Occipital foramina associated with the glossopharyngeal, vagoaccessory and hypoglossal cranial nerves.

Paired ventral troughs depart from the posteroventral braincase endocast and initially trend posterolaterally (interosseous casts: feature 19 in Fig 2). They originate as depressions of the dorsal basioccipitals and continue posteriorly between the medial margin of the opisthotics and the lateral borders of the exoccipitals. Posterior to a mediolateral constriction, these passages depart the cranium through a pair of occipital openings that flank the foramen magnum. These occipital openings are largely defined by the exoccipitals and are only laterally bordered by the opisthotics. Additionally, minute paired conical canals with a subequal diameter of 0.5–0.6 mm branch off near the posterior boundary of cranial endocast at the level of the foramen magnum (feature 20 in Fig 2). These canals project ventrolaterally where they pierce the exoccipital and join the recess of the aforementioned occipital foramina (Fig 4F and 4G). The occipital foramina between the exoccipitals and opisthotics have been explained to be homologous to the jugular foramina (metotic foramina) of extant amniotes that accommodate the pathways of the glossopharyngeal and vagoaccessory nerves [22]. Cranial nerves (CN) IX and X typically branch off closely together [56, 67] and depart the cranium through the jugular foramina in tandem with CN XI [22, 47]. The paired perforation of the exoccipital more posteriorly to the medial endocast accommodated the root of hypoglossal nerve XII that joined cranial nerves IX-XI slightly anterior to their collective cranial departure through the jugular foramina.

Posttemporal foramina.

In TW480000375, an inconspicuous but paired perforation of the occiput resides in the sutural junction of the ventral supraoccipital, dorsolateral exoccipital and medial expansion of the opisthotic (feature 22 in Fig 2). Dorsolaterally to the foramen magnum, these concealed apertures are partially obscured by the lateral projections of the exoccipitals, rendering them nearly invisible in posterior view. They do not proceed into well-defined osseous canals anteriorly but access the non-ossified and therefore poorly resolved otic domain. Small posttemporal foramina have been identified in Nothosaurus mirabilis crania from the Early Ladinian of Germany and it has been proposed that reduced to obliterated posttemporal foramina are synapomorphic for Nothosauridae [22]. Their topographical position agrees with the proportionally even smaller occipital perforations recognized in TW480000375. Well-developed posttemporal fenestrae have been reported for Placodus, but were placed more laterally in the occiput and are delimited by the squamosal, parietal, opisthotic and possibly the pterygoid (Fig 2B in [68]). Notably, that contribution does depict potential occipital perforations of much smaller size at topographical locations that corresponds to those of the posttemporal fenestrae in Nothosaurus ([22]; Fig 2B in [68]). Although the posttemporal fenestrae in Nothosaurus are substantially smaller, they appear to be homologous to the wide apertures found in some plesiosaurs [69]. In crocodylians, the posttemporal fenestrae are small and covered over with cartilage during life [70]. The posttemporal foramina of most squamates are large vacuities, but they are completely absent in scincids [71]. No major vessel passes through the posttemporal fenestrae of extant diapsids [70, 72]. In TW480000375, they appear to communicate with the weakly defined void at or near the region that accommodated the otic capsule in TW480000375. Despite the external expression of the posttemporal openings in TW480000375, their internal morphology is inconsistent with an innervation function in Nothosaurus.

Brain macrostructure.

Overall, the osseous braincase of Nothosaurus marchicus is dorsoventrally constricted by the strongly flattened cranial architecture and laterally by the extremely large temporal fenestrae and associated temporal domain that accommodated the hypertrophied jaw adductor muscle complex [22]. The cranial endocast of TW480000375 consequentially takes the shape of a straight, anteriorly tapering cone, the dorsal margin of which trends anteroposteriorly (Fig 3). Its ventral margin is slightly tilted in an anterodorsal-posteroventral direction. Various additional morphological features can be resolved on the cranial endocast, which will be described and interpreted through osseous landmarks and following comparison with extant and extinct taxa.

Forebrain–olfactory lobes.

The portion of the cranial endocast delimited by the frontal dorsally and posterior vomers and pterygoids ventrally exhibits a longitudinal, V-shaped dorsal cleft resulting from a ventrally projecting sagittal ridge on the frontal. Near the posterior orbits and lateral to the V-shaped sagittal trough, two shallow but clearly paired and distinctly ovoid recesses in the ventral frontal mark the location of the olfactory lobes (Fig 3). Slightly further posterior, a sagittal ridge dorsally terminating in a minor bifurcation rises from the interdigitating suture between the paired pterygoids. This osseous ridge extends from the midpoint between the central orbits up to the frontoparietal suture and causes a confined yet pronounced medial cleft on the ventral endocast. The subtle anteroposteriorly extending bifurcating rim carried by the sagittal upwarp of the interpteryoidal suture defines the support for a cartilaginous interorbital septum (see Fig 4C), as is expected for a tropibasic cranium [22]. During life, this septum carried the planum supraseptale; a cartilaginous trough-shaped structure occurring ventral to the frontal that forms the ventral envelope for the olfactory tracts and bulbs [21, 73–75]. The positions of the two shallow but resolvable recesses in the ventral frontal (see Fig 4C), possibly also referred to by Edinger (“between the orbits”) [21], correspond to those of the osteological correlates of the olfactory lobes in mosasaurs [76], Varanus [77], rhynchosaurs [78], archosaurs such as phytosaurs [73], ichthyosaurs [34], and potentially in elasmosaurid plesiosaurs [79]. This provides sufficient support for a confident recognition of the location of the olfactory lobes in TW480000375.

Forebrain–olfactory tracts.

A medial connection extends posteriorly between the interorbital domain and the cerebral portion of the endocast (Fig 3). Since its lateral to ventrolateral margins represents artificial truncations, it only informs on the dorsal geometry of the corresponding endocranial domain. As such, it describes a gradual dorsal inflexion of the anteroposterior endocast axis at the transition from a laterally and lateroventrally poorly defined element with an incised to flattened dorsal osseous delimitation of its anterior and central domains to a more overarched and better-defined transition to the cerebral domain posteriorly. The olfactory lobes and distal part of the olfactory tracts are bilaterally divided by a sagittal ventral projection of the interfrontal suture. Posteriorly, this ventrally protruding ridge shallows until it progressively disappears, after which the passage accommodating the olfactory tracts continues posteriorly as a dorsally undivided passage [21].

Forebrain–cerebrum.

Moving posteriorly along the confined olfactory tracts, a gradual yet conspicuous lateral divergence of the osseous walls resulting in a pronounced bilateral bulging of the endocranial void marks the anterior margin of the cerebral domain (Fig 3). The dorsal, lateral, and particularly the ventrolateral osseous margins delimiting this median void resolve bilaterally symmetrical yet faint bulges that seems to have accommodated discrete soft-tissue structures in vivo. The braincase appears to be constricted (e.g. [47]) by physiological constraints, which most likely resulted in a strongly sequential division of cerebral domains (see Discussion). Coupled with the presence of additional, non-cerebral tissues in the braincase vault [49], this complicates the delimitation of integer brain domains. The cerebral hemispheres, however, are readily discernible in dorsal and ventral view as a bilateral set of faint lateral bulges following the gradual widening of the olfactory tracts [21].

Forebrain–pineal organ.

On the cranial endocast, the pineal system is expressed as an elliptic cone that is apically truncated at the pineal foramen (Fig 3). Consequentially, the pineal foramen ventrally opens up into the cranial endocast through this large, diverging and initially posteroventrally oriented passage that has an ellipsoid cross-section in the anterolateral plane. The anterior and lateral margins of this passage are subtly concave whereas its posterior margin is straight and oriented posteroventrally. More ventrally, the posterior margin of the passage exhibits an apparent deflection towards a more posteriorly trending dorsal boundary that occurs near the major body of the endocast. The walls of the pineal passage progressively diverge ventrally to meet the cranial endocast, on which it is set dorsally, in a gradual and flush manner. A well-developed, true pineal foramen in Nothosaurus conclusively reflects the presence of a photoreceptive pineal organ [80], also referred to as anterior parietal organ [81], which resembles the lateral eye to a remarkable degree [82]. The relatively large size of the pineal foramen and volume of the corresponding endocranial parietal passage compared to other taxa ([80]; this study) suggests a substantial reliance on the pineal complex during life for Nothosaurus and Placodus [35]. However, these voids also accommodated supportive and/or additional tissues that prevent a detailed reconstruction of the original size and morphology of the receptor system during life (e.g. [80]). Nevertheless, the relatively thick-walled parietal enclosure in TW480000375 strongly suggests that the large dorsal aperture of its pineal foramen conservatively reflects the actual size of the pineal eye, which in turn indicates a strong photoreceptive capability [80]. Although the posterior parietals of more derived Nothosaurus species often exhibit a pronounced lateral constriction inflicted by the enlarged temporal fenestrae, the relative size of the pineal foramen remains unchanged [83], which reinforces the conclusion of a significant dependence on the associated photosensor.

Midbrain and hindbrain.

The posterior portion of the cranial endocast behind the cava epipterica (endocast impression: feature 10 in Fig 2) gradually widens laterally and slightly expands dorsoventrally. This expansion extends up to the level of the parietal-supraoccipital suture dorsally and extends beyond the pterygoids ventrally up to the basisphenoid-basioccipital suture. The portion of the endocast corresponding to the midbrain exhibits a subtle, apparently bilateral swelling on its dorsolateral aspect (Fig 3) that results from corresponding depressions on the ventral parietals, resembles the optic lobes identified in an ichthyosaur [34] with regard to location and orientation, and may thus represent a homologous structure in Nothosaurus. Although morphologically unpronounced, they do occupy one of the broadest cephalic domains, which would be consistent with a certain degree of visual acuity. Nevertheless, although extant crocodylians feature optic systems adapted for their respective niches [84], their optic lobes are poorly delimited due to the locally particularly thick dural envelope [56, 85, 86]. In crocodylians, bulges observed in this region typically reflect a portion of the venous blood system [87]. Furthermore, the lateral boundary of this endocranial domain is not continuously ossified and cannot be accurately identified throughout. These factors warrant caution when interpreting the referred features in Nothosaurus. At the level of the posteriormost part of the parietals, the endocast laterally communicates with paired voids that are formed by the descending flanges of the opisthotics and are excluded from the braincase endocast (Fig 4F—23). However, the referred lateral expansions in the posterior braincase endocast (asterisks in Fig 3) follow topographical contours of the ventral parietals slightly beyond the original delimitation of the brain into these voids. They thus result from artificial truncation and do not reflect the local lateral morphology of the cerebral envelope. The anterior hindbrain exhibits only a non-informative lateral expression in the domain where the floccular lobes may be expected to have resided, if present. This is comparable to the condition of extant lepidosaurian and crocodylian reptiles, which also share a small (yet distinct) flocculus [88], and disagrees with the well-developed floccular expression of birds, many other dinosaurs, and pterosaurs [89]. The floccular complex has an important role in visual stabilization and may potentially aid the reconstruction of habitual activity patterns (see Discussion, but see also [90]). Although the transition from the midbrain to the hindbrain is poorly resolved, the transition from the pons to the medulla oblongata within the hindbrain is discernable through a faint flexure and marked by dorsal and ventral “steps” in the endocast that result from a constriction associated with the parietal-supraoccipital suture (Fig 3). However, since the aforementioned artificial truncations contribute particularly to ventrolateral aspects of these “steps”, the appearance of this transition is artefactually pronounced in the endocast presented here. Between the parietal-supraoccipital constriction and the foramen magnum, the medulla oblongata tapers posteriorly, is dorsally confined by the supraoccipital and ventrally by the basioccipital, and carries a pronounced dorsal sagittal rim that extends under the (dorsal) supraoccipital crest (Fig 3).

Vomeropremaxillary foramina and premaxillary lumen.

Two slit-like passages perforate the palate at the level of the vomeropremaxillary suture and close to the sagittal midline of the skull (feature 2 in Fig 2) to anterodorsally access a medially located lumen (feature 1 in Fig 2). This elliptic-cylindrical cavity resides completely within the premaxilla and measures circa 1.25 mm in anteroposterior diameter, 0.75 mm in width, and circa 1.30 mm in dorsoventral height. The posterodorsal margin of the lumen bifurcates into a paired and generally posteriorly trending tubular passage that can be traced up to the internarial domain (feature 3 in Fig 2). The lumen and posterior tubules connect to smaller peripheral passages that originate in surrounding rostral hollows, such as the alveoli, and at irregular grooves at the dorsal surface of the premaxillae (not depicted). An unpaired “foramen incisivum” has been previously reported for the eusauropterygian genera Simosaurus, Nothosaurus, Cymatosaurus and Pistosaurus ([91]; [7] and references therein). Conversely, a paired vomeropremaxillary foramen, such as present in TW480000375, has also been depicted in other specimens of Nothosaurus marchicus (Fig 11 in [42]; Figs 39 and 61 in [83]; Fig 60 in [6]; Fig 1B in [41]), as well as in Simosaurus (Fig 91C in [92]). Assuming this discrepancy does not represent ontogenetic disparity or a preservational, preparatory and/or visualization artefact, it illustrates that both paired and unpaired vomeropremaxillary foramina may occur within one eusauropterygian genus and thereby suggests that the contrast between singular and paired vomeropremaxillary foramina is less conservatively distributed among nothosaurian genera than presently understood. The vomeronasal fenestrae of squamates typically reside in the vomeromaxillary suture [93], whereas the antrochoanal palatal foramina of Eosauropterygia perforate the vomeropremaxillary suture. The latter condition partially agrees with that of the anterior choanae accommodating the vomeronasal ducts in Sphenodon, which separate the premaxillae from the anterolateral vomers [93].

We interpret that the sauropterygian antrochoanal foramina, which may be either paired or unpaired within a single genus or be absent altogether, are conservatively located in the vomeropremaxillary suture. Such foramina in plesiosaurs, when present, have been proposed to represent vomeronasal fenestrae that communicated with the Jacobson’s or vomeronasal organ [94] that appears to be plesiomorphic for Tetrapoda. Notably, the large unpaired premaxillary foramina of crocodylians exist completely within the interpremaxillary suture [92] and do not house a functional vomeronasal organ, rendering them heterologous to the palatal foramina discussed here, contra [95]. Whereas an ichthian vomeronasal system may exist in some form [96], possibly including a hybrid olfactory and vomeronasal epithelium [67], a distinct vomeronasal chemoreceptor accommodated in a nasal diverticulum makes its first appearance in non-amniote tetrapods [97–99]. Although turtles enjoy chemoreception through a vomeronasal epithelium located in the nasal cavity [100], a discrete vomeronasal system separated from the nasal cavity is commonly present in many amniotes, such as in synapsids [98] and squamates [93, 101, 102]. The tubular Jacobson’s organs of the rhynchocephalian Sphenodon communicate with the oral cavity through the anterior choanae [103]. They reside on the vomer and remain separated from the osseous nasal cavity only by the septomaxillae and cartilaginous elements [104]. Within extant Archosauria, discrete extra-nasal vomeronasal organs do form in the embryonic crocodylian rostrum but subsequently disappear during early development while accessory olfactory lobes never form [105]. Nevertheless, the main olfactory epithelium in the nasal cavity of Alligator does incorporate solitary chemosensory cells [106]. Birds lack accessory olfactory lobes and vomeronasal nerves [107], and both zebra finch and chicken genomes have been demonstrated to lack genes encoding vomeronasal receptors altogether [108].

A discrete adult amniote vomeronasal system consistently comprises three osteologically discriminable components. Those are: 1) dorsoventrally trending ducts that originate at recesses in the dorsal oral cavity, 2) a pair of lumina that typically remains separated from the more dorsolaterally residing nasal cavities, and 3) posteriorly trending vomeronasal (terminal) nerves that ultimately penetrate the cribriform plate to terminate in the accessory olfactory lobes [98, 100, 102, 109]. In TW480000375, the anteriormost palatal foramina were found to independently connect to a single, small and discrete endosseous semi-cylindrical chamber positioned anteroventrally to the narial passage. The endosseous morphology of the vomeronasal organ in Varanus exanthematicus [93, 110] demonstrates that the bilateral separation of its vomeronasal lumina, which are positioned directly dorsal to the vomeronasal fenestrae (Panels A-C in S1 Fig; [110]), is largely achieved through a median internasal septum [111, 112] and most ventrally through a membranous wall (see also Fig 1 in [77]). Although the vomeronasal organs in the rostrum of Sphenodon punctatus [93, 113] are less constrained, the corresponding topography of the narial chamber floor exhibits clear bilateral separation ([104]; Panels D-F in S1 Fig; [113]).

The single median lumen in TW480000375 lacks any morphological indication for a median partition during life. It communicates peripherally with a premaxillary infrastructure arguably homologous to the rostral canal system resolved in an Upper Jurassic pliosaur [114] and at least superficially resembling those encountered in crocodylians and some theropod dinosaurs [115]. Those structures were tentatively interpreted to represent the peripheral neurovascular plexus of a dermal sensor innervated by the trigeminal nerve [114] and may very well have facilitated a corresponding function in Nothosaurus. Notably, a superficially similar maxillary infrastructure was interpreted to accommodate nutritive canals in a second Upper Jurassic pliosaur [116]. Although the vomeropremaxillary domain in TW480000375 has been demonstrated to crucially deviate from the conventional tetrapodal vomeronasal architecture, the bilateral nature of the vomeropremaxillary fenestrae, ducts, and associated posterior innervation in TW480000375 lends some support for an origin through median fusion of a bilateral system. The elongated yet slender rostral morphology of nothosaurs supports a deeply rooted dentition to form a piscivorous “trapping basket” [83, 117] that may conceivably have prevented the accommodation of extensive intraosseous systems and could justify an adaptive modification of the vomeronasal system.

Bony labyrinth.

The dorsalmost portion of the endosseous labyrinth, represented by the dorsal apex of the crus communis and the medialmost aspects of the diverging posterior and anterior semicircular canals, are preserved in TW480000375 at the dorsolateral margins of the braincase endocast posterior to the osseous constriction that defines the transition to the posteriormost braincase endocast domain (feature 16 in Fig 2). These bifurcating tubular projections arise from a descending flange of the supraoccipital lateral to the braincase endocast and the nuchal ridge. Lateral to their origin on the braincase endocast, these voids anteroposteriorly expand into a bifurcation before their abrupt lateral truncation against matrix infilling. However, if the general curvature of the posterior branches of the bifurcations is continued through the opisthotics, additional tubular sections can be discerned along this trajectory (feature 21 in Fig 2). Except for these minor tubular sections of the posterior semicircular canals, no other portions of the semicircular canals, the lagenae, and the vestibules could be reconstructed due to the absence of ossified correlates. The proximodorsal portion of the endosseous labyrinth is abruptly truncated at the level of the original sutures between (probably the epiotic contributions to) the supraoccipitals and the prootics, suggesting that both prootics were either lost or completely unossified. The anterior portion of the posterior semicircular canal and the posterior portion of the anterior semicircular canal are ellipsoidal in cross-section and measure ~0.66 mm along their major axis and ~0.45 mm along their minor axis, which renders them quite robust with respect to the size of the cranium. Furthermore, the spatial relation between the dorsal divergence of the anterior and posterior semicircular canal and the posteriormost expression of the posterior semicircular canal in the opisthotics demonstrates an anteroposterior expansion of the vestibular apparatus with respect to its dorsoventral height. This condition was also described for Placodus [35] and has been associated with an adaptation to an aquatic lifestyle [38, 118]. The discontinuously resolved endosseous labyrinth prevents the reconstruction of the “alert position” (see [51] for discussion).

Uppermost respiratory tract.

The somewhat posteroventrally tilted external nares continue posteromedially into the nasal cavity that extends between the paired nasals and the vomers. The posterior portion of the nasal cavity constricts dorsally while ventrally opening up into the oral cavity through the internal nares (choanae). These are situated behind the external nares at the junction of the vomers, maxillae and palatines, and appear to be tilted somewhat posteroventrally with respect to the palate. Slightly anterior to the posterodorsal constriction of each narial passage, the nasal is perforated by a minute and dorsally branching aperture (feature 7 in Fig 2). The dorsally bulging chambers that continue immediately ventral and posterior the external nares represent nasal vestibules (feature 4 in Fig 2) and, more posteriorly, the cava nasi (feature 5 in Fig 2). Towards the internal nares, the narial passage slightly constricts into the nasopharyngeal duct (feature 6 in Fig 2). The geometry of the narial passage between the external and the internal nares offers reduced airway resistance, which is selected upon in the narial respiratory tracts in numerous marine tetrapods [119]. We conclude that this complex represents the air passage during nasal respiration. The maxilla and vomer share a well-developed suture that separates the vomeropremaxillary foramen from the internal naris, which represents the neochoanate condition (sensu [120]) for sauropsids [93].

Salt glands.

The lateral and ventrolateral margins of the central antorbital void are mainly defined by the paired pterygoids. Around the anterior margin of the orbits, the inconspicuous prefrontals contribute to the lateral delimitation of the medial endocranial cavity. Posterolateral to the narial passages and internal choanae, and dorsomedially and dorsolaterally confined by respectively the lateral nasals and the medial maxillae, the cast of the nasal cavities reflects the presence of two pronounced and anterolaterally—posteromedially trending oblate, bilaterally paired, and partially constrained ellipsoid recesses that align with the anteromedial margin of the orbits (feature 8 in Fig 2). They correspond with voids that are in open communication with the posterolateral narial passages but remain excluded from the posteroventral narial passages and internal nares by a low bony ridge. All extant marine diapsids require extrarenal mechanisms for salt excretion, which are universally derived from cephalic glands (e.g. [121]). This advocates the presence of cephalic salt glands in Nothosaurus marchicus. The developmental affinity of cephalic salt glands is variable between groups and ranges from orbital glands in sea turtles to exclusively lateral nasal glands in lizards and (accessory) sublingual glands in sea snakes, to orbital, lingual and nasal glands in crocodylians [121–127]. It has been argued that nasally derived salt glands constitute the primitive diapsid condition that arose through selection on maintaining ionic balance in particularly arid terrestrial habitats at the end of the Paleozoic [128]. The lacertid genus Acanthodactylus exhibits a “typical” squamate nasal architecture in which the nasal glands that reside in conchal spaces lateral to the cava nasa act as functional salt glands [126]. Fluid secreted by the lateral nasal glands passes into the nasal cavity through a secretory pore, collects in the vestibules, and is subsequently expelled by “sneezing” or allowed to dry out [126]. The paranasal recesses identified in TW480000375 are positioned posterolateral to the nasal cavity rather than purely lateral, but exhibit a similar communication with the nasal cavity as Acanthodactylus, which supports the inference that the paranasal recesses likely accommodated lateral nasal glands in TW480000375. The antorbital recesses in TW480000375 share their geometry, orientation and topographical position of the referred paranasal sinuses ventral to the nasomaxillary suture with the paired antorbital protuberant structures described in the marine fossil crocodylomorph Geosaurus araucanensis and in various other marine reptiles (e.g. [128] and references therein). Those antorbital structures have been explained to have accommodated hypertrophied nasal salt glands [128]. Such salt glands have furthermore been preliminarily reported in the Late Cretaceous polycotylid plesiosaur Pahasapasaurus haasi [129, 130] at a homologous location, although a detailed description has thus far been lacking. Their shared location under the medial maxillary suture posterior to the premaxillae and shared oblate spheroid geometry (indicated as “nc” in Fig 4 of [130]) corroborate the presence of homologously developed salt glands in TW480000375.

Infraparietal canals.

Pronounced anteroposteriorly trending paired tubules were encountered in the fused parietals (feature 9 in Fig 2). Each parietal contains a posteroventral perforation near its anterior margin that gives rise to a posteriorly trending tubular channel. These channels gradually verge out into the secondary tubular mesh around the pineal foramen. The dorsal head vein (vena capitis dorsalis; vcd) enters the cranial endocast posterior to the pineal foramen in Sphenodon [131] and is associated with the dorsal sagittal sinus in squamates [59, 72]. In the dicynodont Niassodon mfumukasi, the pineal foramen perforates the medial aspect of the frontoparietal suture [101], comparable to the condition of Sphenodon [132], whereas it occupies a medial position in the posterior part of the compound parietal in TW480000375. This disparity in osseous configuration may have originated to accommodate the extreme elongation of the nothosaurian temporal domain [83]. Slightly posterior to the frontoparietal suture in Niassodon mfumukasi, the local trajectories of the bilateral branches of the vcd were resolved as posteromedially trending dorsal ridges that appear to discharge the frontal anterolaterally to the pineal foramen [133]. Intrafrontal veins supplying these endocranial intervals of the vcd were not described, but may be expected to be present anterior to the pineal foramen. The aforementioned tubules in TW480000375 were encountered at a corresponding location relative to the pineal foramen in, albeit in the parietal rather than in the frontal, and verge out before actually reaching the level of the pineal foramen. As such, these intraparietal tubules in TW480000375 would roughly continue into the inferred endocranial course for the vcd of Niassodon mfumukasi if the endocasts were superimposed. This lends support for the interpretation that the intraparietal pathways of TW480000375 accommodated the branches of the vcd discharging the vasculature of the dorsal cranium.

Internal carotid branches.

Paired longitudinal cylindrical passages enter the cranium from the direction of the neck through occipital foramina distinctly lateral to the foramen magnum. They subsequently penetrate the pterygoid and ultimately merge with the braincase endocast near the cava epipterica. In their trajectories, these canals gradually arc from an anteromedial course near their occipital origin to an anterior course close to their contact with the braincase endocast. A posterolateral bifurcation branches off at the level of the anterior squamosal suture with the parietal. This marks the posteromedial arrival of a second duct that lacks an osseous enclosure posteriorly to its penetration of the pterygoid. The internal carotid arteries are among the major cranial vessels in squamates [72] and crocodylians [134, 135]. The common carotid arteries depart from the dorsal branch of the aortic arch and bifurcate into an external and an internal carotid ramus before reaching the skull [72, 134, 135]. The internal carotid ramus gives rise to the cerebral carotid where the stapedial artery branches off [72, 134, 135]. After the departure of the stapedial artery, the cerebral carotid advances through the carotid canal where it eventually reaches the sella turcica. It continues anteriorly as the sphenopalatine artery [72, 134, 135]. More anteriorly still, the sphenopalatine artery runs through the vidian canal that also carries the palatine branch of the facial nerve [22, 71, 72, 136].

Most of the endocranial path of the internal carotid artery and its branches could be resolved in TW480000375, where they were found to broadly resemble the condition of Nothosaurus mirabilis [22]. The internal carotid penetrates the pterygoidal flange and remains entirely enclosed by bone during its lateral bypass of the hindbrain behind the cavum epiptericum in the carotid canal (feature 15 in Fig 2) and remains entirely excluded from the paracondylar interstices (feature 18 in Fig 2; see Discussion). The important connection to the sella turcica is not preserved in TW480000375, but the sphenopalatine artery that branches off simultaneously with the arch irrigating the pituitary gland must be located where the vidian canal shallows at the cavum epiptericum (feature 12 in Fig 2). Here, the corresponding artery passes onto the dorsal surface of the pterygoid [72] as it enters the basicranium at the level of the midbrain. This has previously been considered a derived condition in Nothosauroidea [22].

The encountered bifurcation may represent the departure of the stapedial artery (feature 14 in Fig 2) from the internal carotid, which would also imply that the internal carotid continues into the cerebral branch of the internal carotid (feature 13 of Fig 2) at this level. This interpretation is supported by the observations that the internal carotid artery bifurcates into the cerebral branch and the stapedial artery in the posteriormost region of the cranium in both Iguana [72] and Alligator [134]. In those taxa, the cerebral branch proceeds anteriorly to cross the sella turcica and irrigate the pituitary gland, which is partially reflected in the course of the cerebral branch in TW480000375. This is also consistent with nothosaurian material from Tunisia where the ventrolateral arrival of large carotid canals at the pituitary fossa was recorded [61]. The poorly resolved short branch of the referred bifurcation is directed posterolaterally in TW480000375, which is where the otic capsule resides. In both Iguana and Alligator, the stapedial artery proceeds dorsally towards the otic domain after its departure from the internal carotid [72, 134], where it irrigates the middle ear [100, 118]. Alternatively, this bifurcation could represent the arrival of a vein or nerve at the pterygoidal canal on its course towards the postcranium independent from the internal carotid but sharing the trans-pterygoidal osseous passage. In iguanids, both the internal jugular vein and the vagus nerve traverse the posterior cervical domain in close association with the internal carotid artery [72, 75].

Middle cerebral vein or paratympanic sinus.

In the posteriormost segment of the braincase endocast, a bilateral pair of curved canals departs from the posterodorsal aspect of the endocranial cavity into the supraoccipitals while curving downwards and inflecting ventrolaterally into the otic capsule (feature 17 of Fig 2). These tubules arise at the suture shared by the supraoccipital and the prootic and overarch the dorsolateral bifurcation of the crus communis into the anterior and posterior semicircular canal (feature 16 of Fig 2) before opening up into the void of the otic domain ventrally. Based on comparison with extant crocodylians, the position and geometry of these passages suggest they either accommodated the middle cerebral vein [135] or allowed for communication between the paratympanic air sinus system and the paired auditory system by overarching the brain [57]. A similar structure has recently been described in a highly secondarily-adapted crocodylomorph as the dorsal dural venous sinus [36] and was recognized as unnamed cavity in a related marine crocodylomorph taxon [37]. In squamates however, this region is formed by the dorsal longitudinal sinus or the parietal sinus [72], which would be expected to have corresponded with a clear osteological correlate. These foramina are unlikely to have accommodated cranial nerves, as they arise from a dorsal region of the endocast.

Braincase heterochrony.

The absence of (ossified) prootics suggests that TW480000375 had not yet reached skeletal maturity, which is inconsistent with the observation that the external cranial sutures appear well ossified. Prolonged or truncated physical and osteohistological maturation is a well-known trend in organisms secondarily adapted to an aquatic lifestyle [6, 9, 17, 92, 138–144]. This predominantly affects the postcranium in secondarily-adapted Mesozoic marine reptiles [9], but the specimen described here exhibits important cranial paedomorphism as well. Skeletal paedomorphosis in secondarily-adapted aquatic taxa is a common trait [144–146]. It typically results in decreased ossification during bone development, as with the extremely deferred prootic ossification in Nothosaurus, but it occasionally leads to deletion of cranial elements altogether [147, 148].

In TW480000375, the prootic is absent and only the occipital portion of the opisthotic is preserved. In the generalized developmental ossification sequence across reptiles, elements of the otic capsule are among the last to ossify [149, 150]. This is a common trait shared with synapsids [151], thus emphasizing the conservative nature of this pattern. The delayed, reduced or lacking ossification of components of the otic capsule is not an isolated preservational issue, as it was also observed in other nothosaurid specimens (e.g. [47, 148]; PGIMUH K3881, SMNS 16363; TR pers. obs.), as well as in plesiosaurs [139, 148]. We interpret this selective delay in ossification to result from paedomorphosis [152] through which the development of the bones forming the otic capsule was considerably delayed relative to the ancestral state of basal neodiapsids, whose osseous otic capsule is conspicuously ossified (e.g. [153, 154]).

In nothosaurids, the delayed development of bones forming and supporting parts of the braincase and the otic capsule is reflected in the absence of various structures that are commonly preserved surrounding the brain in other taxa. For example, placodonts exhibit an osteologically matured braincase that permits the extraction of several well-defined cranial nerve passages and the osseous labyrinth [35], which is also the case for the Late Cretaceous plesiosaur Libonectes morgani [155]. Because the lateralmost braincase walls can only be interpolated in Nothosaurus, definition of the topology of most cranial nerves, the vestibular organ, and several brain structures is rendered impossible. The eosauropterygian pterygoid, on the other hand, is exceptionally large [9, 156–158] and remarkably well ossified [159]. It typically extends posteriorly from the preorbital region towards its contribution to the occiput where it supports the basicranial axis bones ventrally. Although the pterygoid is one of the first bones to ossify in reptiles [149, 150], the pronounced development of the eosauropterygian pterygoid relative to the ancestral state, conversely to the condition of the otic capsule, might be the result of peramorphosis [152]. Although the postcranial anatomy of eusauropterygians appears to predominantly reflect a paedomorphic trend, the cranial anatomy seems to exhibit a mosaic of heterochronic effects. This may have important implications for phylogenetic coding, as ossification of the external cranial elements is typically considered a reliable indicator for skeletal maturity in Nothosaurus where alpha taxonomy is predominantly founded on cranial characters (e.g. [16, 83, 160, 161]).

“Foramina eustachii” of Koken (1893).

The ontogenetic plasticity [22], relatively large size, posteroventrally oriented occipital eruption, and absence of main vascular passages or potentially homologous foramina in other diapsids indicate that “foramina eustachii” probably represent taxon-specific and ontogeny-dependent morphological expressions of the junction between the basioccipital and the pterygoids rather than functional cranial foramina. Therefore, the geometry of these paracondylar interstices appears to be an architectural byproduct of the unique cranial construction of nothosauroids. Because these occipital gaps do not represent cranial foramina in the strict sense of the term, we suggest avoiding the interpretative term “eustachian foramina” and instead propose to adopt the descriptive term “paracondylar interstices” when referring to these phenomena,.

Endocast macrostructure.

The specialized cranial condition in Nothosaurus that combines dorsoventral flattening with strong lateral constriction of the braincase by the temporal fenestrae is associated with an extremely elongated linear brain morphology that exhibits an overall sequential zonation of the brain along its anteroposterior axis, including a strongly extended olfactory tract [21, 47, 56]. Although the hindbrain is broader than the forebrain, which represents the conventional reptilian condition, the forebrain and hindbrain reside at the same dorsoventral level in the unusually straight and strongly anteroposteriorly orientated cranial endocast. A generally corresponding but less pronounced linear endocranial morphology has been reported for other marine reptiles, including the closely-related placodont Placodus gigas [35] and a lower Jurassic ichthyosaur [34]. A similar reconstructed brain morphology of thalattosuchians [36] was recently proposed to result from the large, laterally placed orbits [85], which would have provided comparable constraints on cerebral morphology as the temporal fenestra in Nothosaurus marchicus.

The total volume of the endocranial cavity, excluding the poorly defined olfactory tract and olfactory lobes, amounts to circa 810 mm³. In reptiles, the brain itself does not fill the endocranial cavity, which in fact mirrors the external surface of the dural envelope [49]. Therefore, the brain of crocodiles and certain dinosaurs is believed to occupy only circa 50–60% of the endocranial cavity [55, 162, 163]. The relation between condylobasal skull length and full body length in Nothosaurus marchicus, not corrected for ontogenetic allometry, was recently described [14], and yields a reconstructed body length of circa 650 mm for TW480000375. Assuming TW480000375 exhibited a body length to body mass ratio comparable with that of Varanus keithhornei, its total body mass would have been circa 270 g [164]. Similarly, comparison with a juvenile alligator (610 mm [165]) yields a reconstructed body mass of 306 g for TW480000375. A corresponding brain mass (assuming a brain density of 1.036 g/cm³ [89]) between 0.4 (circa 50% of the braincase vault) and maximally 0.8 g (nearly 100% of the braincase vault) corresponds with a reptilian encephalization quotient (REQ; [166]) between 0.15 and 0.35. This places Nothosaurus marchicus within the typical range of the relation between body weight and brain weight followed by extant reptilian taxa [167].

The complex division of the brain in cerebral compartments is particularly challenging to recognize in extinct forms due to the presence of additional structures (e.g. meningeal layers) in the endocranial vault during life that usually do not preserve during fossilization [49, 56, 168]. Identification of discrete cerebral domains in fossils represented by exclusively osseous remains can only be successfully achieved through recognition and conservative application of well-understood osteological correlates of such soft-tissue structures in the cranium. For example, in the elongated brain of Nothosaurus with an associated sequential brain zonation, this implies that the suture between the prootic and the opisthotic, which ventrally terminates in the fenestra vestibuli, accounts for the anteriormost possible extent of the medulla oblongata, as the prootic represents an indisputable posterolateral delimitation of the cerebral domain, corroborated by a faint flexure. Particularly for taxa exhibiting a sequential zonation of the brain, the developmentally conservative relations between osseous markers and associated soft-tissue structures, such as those between the vestibular system and the cerebellum [169], are to be respected when delimiting and identifying individual brain compartments.

Pineal organ.

Both the pineal eye and the associated pineal gland (also termed posterior parietal organ or epiphysis), which is retained as a neuroendocrine gland in numerous vertebrates lacking the pineal eye [170, 171]; see also [172]) have photosensitive capabilities [81], possibly because they represent the bilateral remnants of an originally paired structure [82, 173]. Together, these organs comprise the pineal complex [174] that, in modern lizards, influences behavior, body temperature regulation and reproductive synchrony on circadian to annual timescales through sensory stimulation of the endocrine pineal system that, for example, regulates thyroid activation [82, 170, 174–177]. Notably, contribution of the pineal eye to spatial orientation by means of a “time-compensated sun compass” when negotiating a water maze [178] and to the expression of aggressive display through an interaction between pineal endocrinal cues and thermoregulation [179] have also been documented. Finally, parietal foramen size in mosasaurs has been suggested to correlate positively with diving depth, but a hypothesized positive correlation with latitude could not be substantiated [180].

In most modern lizards, however, the size of the pineal organ strongly correlates with latitude and diurnality [181, 182], and the pineal eye may be absent altogether in equatorial species ([183] and references therein). A particular role for the pineal organ in “fine-tuning” thermoregulation in ectotherms was described and it was hypothesized that the reduction and loss of the pineal eye in Eucynodontia reflects the transition from exo- and mesothermia to endothermia [184]. Although plesiosaurs have been proposed to be homeothermic and capable of maintaining a body temperature substantially higher than that of ambient waters [185], a pineal foramen is variably present [155], and a somewhat enlarged pineal foramen was recognized in a Cretaceous plesiosaur recovered from a high paleo-latitude [186]. Pistosaurs exhibit a cortical microstructure in their long bones indicative for high growth rate and high metabolic rate [18, 138] but also retain a well-developed pineal foramen (e.g. [6]). This suggests that, in Sauropterygia, an elevated metabolic rate does not preclude the presence of a well-developed pineal foramen, as was also noted for Eucynodontia [183]. Also, the pineal organ has been described as an adaptation to terrestriality [186], which is at odds with the large size of the pineal foramen in nothosaurs.

Nothosaurus exhibits a cortical microstructure that indicates a moderately low growth rate [138] and suggests a comparably low metabolic rate [18, 138]. The conclusion that Nothosaurus marchicus relied substantially on the pineal organ (see also [80]) is consistent with an exo- to marginally mesothermic thermoregulatory strategy inferred from osteohistological proxies. Nothosaurus has thus far only been reported from warm epicontinental seas [138] at tropical and subtropical paleolatitudes (Western and Central Europe, Israel, Tunisia, Saudi Arabia and South China; [6]), suggesting a distribution governed by isotherms. Exothermic sea snakes, notably lacking a photoreceptor in their pineal organ [82, 177], are incapable of elevating their body temperature substantially above the ambient water temperature, even when floating at the surface. They are therefore restricted within specific surface isotherms of circa 18–20° C [187–189]. Furthermore, most sea snakes and all sea kraits are bottom foragers that rapidly equilibrate to the lower temperatures at depths up to 100 m [189]. Although Pelamis platurus can stand a water temperature of 5° for about an hour of time and can tolerate circa 11° C for about 36 hours, substantially longer periods endured at even 17° C are lethal [188]. Paleotemperature proxies applied to stacked Triassic marine deposits across Europe, including Lower Muschelkalk deposits from Germany, have recovered a paleotemperature range between circa 18–32° C [190]. This demonstrates that a thermoregulatory strategy similar to that of modern sea snakes (e.g. [191]) appears to have been available to Nothosaurus without the necessity for a photosensitive pineal organ. This, in turn, suggests that an alternative or contributing accessory function may have provided the pineal organ with a functional advantage that warranted its proportionally large size in several non-pistosauroid sauropterygians. Interestingly, the pineal organ is also known to perform a neurosecretory function governing melatonin secretion and distribution [192]. Melatonin is responsible for regulating skin pigmentation, locomotion, somnia and reproduction [193]. Melatonin controls the activity of melanophores and thereby melanin concentrations in tissues [194, 195]. Furthermore, it has been observed that the admission of pineal gland extracts to various tetrapods causes lightening of skin color [196–199]. Countershading, or Thayer's Law, describes a mode of camouflage in which an animal's coloration is darker on the upper side and lighter on the underside of the body. This pattern is found in many species of mammals, reptiles, birds, fish, and insects, both predators and prey, and has occurred in marine reptiles at least since the Cretaceous period [200]. In marine environments, countershading appears to offer camouflage particularly for mid-level dwellers towards both deeper and shallower vantage points in the water column. In a shallow marine exotherm foraging at the sea floor over a light substrate, such as Nothosaurus marchicus, skin coloration is intuitively subjected to the conflicting demands of promoting heat absorption while afloat at the surface (conventional countershading) and ensuring sufficient camouflage while sojourning at the sea bottom. Involvement of the remarkably large pineal foramen and potentially enlarged pineal gland could offer a functional solution for this dilemma in Nothosaurus. Such a hypothesis, although speculative, proposes the consideration of a mode of adaptive skin pigmentation (metachrosis) through pigment translocation that adjusts skin tone on seasonal, diurnal or even sub-diurnal timescales towards optimally balancing thermoregulation and crypsis

Floccular complex.

Relative size of the floccular complex has been argued to proportionally reflect the requirement for image stabilization during rapid, agile locomotion in archosaurs through the vestibulocular and vestibulocollic reflexes [89], which grant a steady gaze during, for example, optical guidance while pursuing prey. This claim was recently challenged in a study assessing the relation between the size of the floccular fossa and ecology and behaviour across mammals and birds [90]. No floccular complex was recognized in TW480000375, although poor ossification of the prootic prevents reconstruction of this domain with absolute certainty. However, a floccular expression is also lacking in the endocast of the strongly ossified braincase of Placodus gigas [35]. Irrespective of the absolute predictive capabilities of relative floccular size and the contribution of other cerebellar regions to visual stabilization, the virtual absence of a floccular expression in the endocast of TW480000375 may be expected to correlate with reduced processing capacity and corresponding decreased performance [55], and may be linked with reduced oculomotor performance. This is intuitively more consistent with ambush predation than with endured, agile, high-speed pursuit of prey. Aquatic ambush predation guided by visual cues characterizes crocodylians (e.g. [84] and references therein), whereas vision also represents a crucial sense in short-distance ambush predation by various aquatic snakes and sea kraits [201, 202].

Chemoperception.

While the vomeronasal organ is present in many tetrapods, only selected squamates possess a deeply cleft or forked tongue of which the tips individually communicate (either directly or indirectly through sublingual plicae [203]) with paired palatal fenestra that lead to the paired lumen accommodating sensory epithelium [101, 204]. Such an arrangement enables tropotaxis: the immediate perception of directional gradients in the concentration of certain chemical compounds after tongue flicking, which permits sampling and directional comprehension of chemical prey cues prior to attack [101, 204]. Other squamates, including ambush-hunting iguanian lizards, but also the rhynchocephalian Sphenodon, have a less or non-bifurcated tongue that does not permit true tropotaxis but can aid in mediating chemosensory evaluation of prey during capture by lingual pretensions or after oral contact is established [204], which does not rely on perception of a directional component. Among squamates that engage in tongue flicking, actively hunting taxa tongue-flick regularly throughout their forage, whereas ambush hunting taxa only tongue-flick when moving between ambush sites [205].

Although N. marchicus possess paired vomeropremaxillary foramina that appear homologous to the vomeronasal fenestrae of extant tetrapods, these are spaced closely together and continue dorsally into a single, shared lumen. This pattern is most consistent with the absence of a forked tongue in Nothosaurus and, consequently, the inability of a “squamate” mode of chemical gradient detection by means of tropotaxis. Retention of chemosensory capability in Nothosaurus, however, cannot be ruled out. Non-directional applications of chemoperception, such as the evaluation of potential prey, do not rely on spatial separation of sampling locations. Furthermore, in a nectic marine predator, the capability of sequentially sampling ambient water currents rather than obtaining instantaneous directional information should be sufficient to engage in tracking biochemical gradients using klinotaxis. Notably, klinotaxis represents the plesiomorphic mode of spatial chemoperception for lepidosaurs [205].

Vomeronasal chemoreception has been suggested to contribute to the array of sensory perceptions available to Sauropterygia [94]. We have observed that sauropterygian foramina incisiva occur as either one single or as two paired vomeropremaxillary foramina (e.g. [7]). Furthermore, the paired vomeropremaxillary foramina of TW480000375 communicate with a single, medially positioned lumen. The endorostral morphology of TW480000375 thereby departs from the conventional architecture of vomeronasal chemoreceptors in tetrapods but may have originated through median fusion of an initially bilaterally developed chemosensor under the morphological and ecological constraints imposed by the specialized rostral morphology of nothosaurs. It is well established that secondarily marine amniotes exhibit a reduced or even completely lost vomeronasal organ [206, 207], although a functional tropotactic vomeronasal system has been argued to have been present in mosasaurs [208]. The unique reduced endorostral morphology of TW480000375 reflects that of an atrophied vomeronasal organ with respect to those typically present in ground-dwelling tetrapods. Such an atrophied vomeronasal organ does not preclude the presence of an eusauropterygian rostral sensory system such as that proposed to enable mechano- or electrosensory perception in a pliosaur [114]. Notably, these two systems are not intrinsically linked, as chemoperception by the vomeronasal organ is achieved through innervation by the vomeronasal nerves connecting to the accessory olfactory lobes [209], whereas mechanoelectric perception is typically enabled through the facial nerve of the trigeminal nerve [210].

Olfaction.

The limited osseous expression of the olfactory lobes on the ventral frontal prevents a confident reconstruction of their size and of the inferred neurosensory dependence on olfaction during life. Although the main olfactory system of extant marine reptiles performs poorly or not at all during submersion [100], and accommodation space for nasal olfactory epithelium may have been restricted through the presence of well-developed salt glands in Nothosaurus (this contribution), poorly defined osseous expressions of the olfactory lobes cannot be construed to conclusively support reduced olfactory performance. However, olfaction was evidently inhibited and will not have contributed considerably to successful subaqueous foraging. This pattern is convergent with many other secondarily aquatic tetrapods (e.g. [31, 211, 212]).

Nasal cavity.

A previous study argued against a primarily respiratory function for the narial passage in Eusauropterygia in general and for Nothosauria in particular [116]. Exclusive nasal respiration is inconsistent with an oral morphology supporting relatively irregularly protruding fangs that would likely have prevented a watertight seal of the oral cavity (see also [213]), for example during surfacing with a partially submerged cranium. However, in absence of a secondary palate [116], we conclude that the narial passage between the external and the internal nares did provide an air passage accessory to oral respiration, providing the choanae were not continuously covered with a palatal tissue. Size reduction of the internal nares from pistosauroids to Plesiosauria has been reported [7, 156], and the resulting size discrepancy between the external and the internal nares in plesiosaurs has been deemed a critical obstruction during respiration [119]. More importantly, the referred evolutionary trajectory is also accompanied by a gradual rearrangement of the internal nares to a location anterior to the external nares (e.g. [94, 213]), which has been proposed to coincide with the development of a functional secondary palate [116, 119]. However, more basal sauropterygian taxa exhibit the more conventional choanal placement posterior to the external nares (e.g. Nothosaurus) or approximately at the same level (e.g. pachypleurosaurs [214]), which is consistent with a more traditional respiratory function in these groups.

Chelonioidea accommodate a designated and strongly domed olfactory chamber (the “upper chamber” of [98]) in their nasal architecture. It has been speculated that an air bubble trapped in this dome during submersion receives volatile chemicals from water pulsed through the sinuses, which permits olfactory access to waterborne chemicals in marine turtles [100]. Compared to the more strongly domed compartments in sea turtles, the dorsoventrally flattened cranium of Nothosaurus only houses shallow nasal chambers in which air retention during cranial tilting or in turbulence appears highly unlikely.

Salt excretion.

The “upper” olfactory chambers of marine Cryptodira share their topographical location and geometry with the recesses housing the postulated salt glands in Nothosaurus marchicus (consider the morphology of the “nasal cavity” of Plesiochelys etalloni; [59]). This suggests that the cryptodiran “upper olfactory chambers” and the nothosaurian antorbital recesses could represent homologous cranial voids in which glandular tissue may have exaptively replaced olfactory epithelium towards accommodating the increased need for salt excretion in the secondarily marine Sauropterygia. In extant diapsids, the secreted hypertonic solution is discharged into the nasal cavity and secreted through the nostrils [128]. In Nothosaurus marchicus, the postulated salt glands are contained within the posterior nasal domain and appear to remain partially separated from the narial airway by non-ossified septa, as indicated by bony ridges that supported such structures in vivo. Glandular discharge of hypertonic fluid likely proceeded into the cava nasa, where this solution could subsequently be expelled through the external nares, possibly aided by partial nasal respiration (as in some modern lizards [126]). Furthermore, choanal expulsion of saline discharge has been proposed specifically for Pistosauria [116]. However, since analogous modes of choanal discharge do not exist in modern taxa, phylogenetic support for this secondary mode of saline fluid emission is lacking. Additional discrete antorbital fenestra or neomorphic preorbital fenestra, such as present in Geosaurus and there suggested to potentially contribute to salt excretion [128], are absent in Nothosaurus. Lacrimal salt excretion as in sea turtles [123] appears highly unlikely for Nothosaurus, since sufficient interorbital accommodation space and the corresponding large interorbital foramen [214] that support lacrimal salt glands larger than the brain in sea turtles [215, 216] are absent in TW480000375.

An inferred hydrodynamically optimized “ram jet” configuration of the plesiosaurian narial passage has been proposed to benefit a speculative subaqueous operation of the main olfactory system or a vomeronasal chemosensor [213]. The main olfactory system probably does not function during submersion in extant non-chelonian reptiles [100] and we inferred that the chamber ancestrally lined with olfactory epithelium was likely at least partially occupied by the salt glands in Nothosaurus, which suggests a shared diminished dependence on the main olfactory system in plesiosaurs. Although the presence of hydrodynamically aided olfaction is therefore unsupported, the described forced flushing mechanism in plesiosaurs would be consistent with drainage of the hypertonic solution through the external nares rather than into the oral cavity.