Jeholornis, Zhou & Zhang, 2002

Hu, Han, Wang, Yan, Fabbri, Matteo, O, Jingmai K., Connor, Mcdonald, Paul G., Wroe, Stephen, Yin, Xuwei, Zheng, Xiaoting, Zhou, Zhonghe & Benson, Roger B. J., 2023, Cranial osteology and palaeobiology of the Early Cretaceous bird Jeholornis prima (Aves: Jeholornithiformes), Zoological Journal of the Linnean Society 198 (1), pp. 93-112 : 95-107

CRANIAL OSTEOLOGY OF JEHOLORNIS

Premaxilla

The premaxillae are complete in STM 3-8 ( Fig. 2A–F ). They are edentulous, as reported in previous publications ( Zhou & Zhang, 2002, 2003; Lefèvre et al., 2014), and their external surfaces are marked by several nutrient foramina. No pits are present to receive the dentary teeth, which is different from the condition present in Ichthyornis Marsh, 1873 ( Field et al., 2018). The tip of the corpus forms a ventral projection, suggesting that the tip of the beak might have been slightly hooked. Caudal to the rostral ‘hook’, the ventral margin of the premaxillary corpus is straight. The relative level of the ventral projection of this rostral ‘hook’ varies among previously reported specimens, being absent in Jeholornis YFGP-yb2 and exaggerated relative to STM 3-8 in Kompsornis AGB-6997 ( Lefèvre et al., 2014; Wang et al., 2020a). This is interpreted here as attributable to variation in preservation, but further studies are needed to exclude the possibility of intraspecific variation confidently. The premaxillary corpora are fused, whereas the frontal processes are separated ( Hu et al., 2022).

The frontal (nasal) process of the premaxilla is relatively short and therefore does not contact the frontal but instead articulates distally with the dorsal surface of the nasal, as in other non-ornithothoracine stem birds (e.g. Archaeopteryx and Sapeornis ; Rauhut, 2014; Kundrát et al., 2018; Hu et al., 2020a) apart from Confuciusornis Hou et al., 1995 ( Chiappe et al., 1999; Elżanowski et al., 2018; Wang et al., 2019a). This is evidenced by the extension level of the articular facet present in the frontal process of the premaxilla. The maxillary process of the premaxilla is short and articulates medially with the premaxillary process of the maxilla ( Fig. 2A, B ). It appears to be step-like, with the dorsal margin extending farther caudally than the ventral margin. The palatal process is crushed mediolaterally but could still be distinguished from the maxillary process in the 3D reconstruction.

Maxilla

The maxillae in Jeholornis STM 3-8 are well preserved, being almost in articulation with the premaxillae and the lacrimals ( Fig. 2A, B, J–L ). The premaxillary process bears a lateral depression that receives the dorsal part of the maxillary process of premaxilla. The jugal process is slender, being half of the dorsoventral height of the premaxillary process and more than twice its length. A shallow groove is present along the medial surface of the jugal process, indicating extensive contact area with the palatine, similar to Archaeopteryx ( Mayr et al., 2007) . The medial surface of the maxilla is rarely visible among Mesozoic bird specimens, precluding further comparisons. Although the palatal process is mostly crushed, it seems that it was sheet-like and well developed, and therefore most probably contacted the vomer, similar to the Late Cretaceous enantiornithine Gobipteryx Elżanowski, 1974 and the ornithuromorph Ichthyornis ( Chiappe et al., 2001; Field et al., 2018), the only other Mesozoic birds in which the morphology of the palatal process of the maxilla is known so far. A dorsoventrally elongate oval fenestra is present between the jugal process and the ascending process in Jeholornis , being enclosed caudally by a thin, bony bar. We identify this tentatively as the maxillary fenestra ( Witmer, 1997). However, it is unclear whether it is homologous with the maxillary fenestra or promaxillary fenestra of non-avian theropods and Archaeopteryx ( Witmer, 1997; Barsbold & Osmólska, 1999; Xu & Wu, 2001; Mayr et al., 2007; Rauhut, 2014; Rauhut et al., 2018) or with the accessory fenestra present in the enantiornithine bird Pengornis Zhou, Clarke & Zhang, 2008 (O’Connor & Chiappe, 2011).

Two alveoli are present in the maxilla. Two teeth are preserved in the left maxilla, and another two similar-sized teeth are dislocated beside the right maxilla. The maxillary teeth are straight and subconical, with blunt crowns and an expanded root.

Lacrimal

Both lacrimals are well preserved in articulation with the maxillae in Jeholornis STM 3-8 ( Fig. 2A, B, G–I ). The rostrodorsal ramus is remarkably short, approximately one-quarter the length of the long caudodorsal ramus. This differs from most other Early Cretaceous birds and non-avian theropods (e.g. Archaeopteryx and Sinornithosaurus Xu, Wang & Wu, 1999 ; Xu & Wu, 2001; Rauhut, 2014; Kundrát et al., 2018; Rauhut et al., 2018), in which the rostrodorsal ramus is long and the caudodorsal ramus short. The ventral ramus is caudally recurved in Jeholornis , such that the caudal margin formed by the ventral and caudodorsal processes is concave, forming the rostral/ rostrodorsal margin of the orbit. This is similar to the morphology in more crownward birds (e.g. the Late Cretaceous ornithurine bird Ichthyornis ), although the rostrodorsal ramus is even more strongly reduced in Ichthyornis and does not contact the maxilla, unlike in Jeholornis ( Field et al., 2018) . The lacrimal morphology of Jeholornis also contrasts with the morphology of most other Early Cretaceous birds and non-avian theropods, in which the ventral ramus is almost perpendicular to the ventral margin of the skull or is inclined cranially ( Wang et al., 2021). The lacrimal of the confuciusornithiforms appears to be slender and reduced, also presenting a short or totally absent rostrodorsal ramus and slightly caudally recurved ventral ramus ( Elżanowski et al., 2018; Wang & Zhou, 2018; Wang et al., 2019a), potentially similar to Jeholornis . However, owing to the potential uncertainty from 2D preservation of currently published skulls of confuciusornithiforms, 3D data are needed to confirm this in future analyses.

Disarticulation prevents detailed reconstruction of articulations between the lacrimal, the nasal and the preorbital ossification in Jeholornis STM 3-8. The ventral ramus of the lacrimal is interpreted as contacting the jugal process of the maxilla and, potentially, might have contacted the rostral tip of the jugal, whereas the lacrimal contacts the jugal in other theropods ( Xu & Wu, 2001; Rauhut, 2014; Kundrát et al., 2018; Rauhut et al., 2018). The caudal margin of the lacrimal, which forms the cranial margin of the orbit, is remarkably excavated, and the excavation extends across both the ventral and caudodorsal processes. A lacrimal foramen lies within the centre of the excavation, entering medially into the main body of the lacrimal at around its mid-height, at the junction of the ventral ramus and the caudodorsal ramus. The size and central position of this foramen resemble the condition in Ichthyornis ( Field et al., 2018) , although the lacrimal of Ichthyornis differs in lacking the rostrodorsal ramus and thus the contact with the maxilla. In contrast, this foramen is much smaller and penetrates lateromedially in enantiornithine IVPP V12707, which also lacks any excavation on the lacrimal on the rostral orbit margin of the lacrimal ( Wang et al. 2021). This contrasts with the craniocaudal extension of the lacrimal foramen in Jeholornis .

Nasal

The left nasal is well preserved ( Fig. 3A, B ). The nasal corpus is mediolaterally broad, similar to Sapeornis ( Hu et al., 2019, 2020a) but unlike the more elongated condition present in Archaeopteryx ( Mayr et al., 2007; Rauhut, 2014; Kundrát et al., 2018), confuciusornithiforms ( Elżanowski et al., 2018; Wang et al., 2019a) and enantiornithines (O’Connor & Chiappe, 2011). Both the premaxillary and the maxillary processes are delicate and sharply tapered rostrally. The premaxillary process is slightly longer than the maxillary process, and the deflections of both processes in the left nasal are taphonomic, resulting from crushing between the right nasal and left lacrimal. The premaxillary process does not extend to the base of the frontal process of the premaxilla, therefore leaving the premaxilla to form most of the rostrodorsal margin of the external naris. This is different from the condition in Archaeopteryx , in which the premaxillary process is substantially longer than the maxillary process and forms part of the dorsal–rostrodorsal margin of the external naris ( Rauhut, 2014; Kundrát et al., 2018). The maxillary process of the premaxilla of Jeholornis is also relatively short and does not extend to the base of the ascending process of the maxilla, therefore not contacting the premaxilla. This leaves the maxilla to form the caudoventral margin of the external naris, similar to Archaeopteryx ( Rauhut, 2014) .

Preorbital ossification

Specimen STM 3-8 preserves a mysterious pair of sheet-like elements previously referred to as ‘preorbital ossifications’ ( Fig. 3C, D ; Hu et al., 2022), which might represent prefrontals based on their overall shape and location. This is supported by their location almost parallel to the craniodorsal process of the lacrimal, which rules out identification as the ectethmoid, especially considering that other rostral elements are mostly preserved in situ . If this element is the prefrontal, it differs from the prefrontals of all other pennaraptorans so far, which are strongly reduced or absent, being typically smaller than the nasal (e.g. in Archaeopteryx and Sinornithosaurus ; Xu & Wu, 2001; Rauhut et al., 2018). This could suggest that an unfused, expanded prefrontal might be a derived feature of Jeholornis and challenges the hypothesis based on the embryonic observations that the prefrontal fused to form the caudodorsal ramus of the lacrimal in all birds ( Smith-Paredes et al., 2018). If correctly identified, this suggests that a broad prefrontal co-exists with a lacrimal with a well-developed caudodorsal process in Jeholornis . However, owing to the lack of available comparisons of any similar ossifications among non-avian dinosaurs and birds, we cannot exclude the possibility that this bone represents some other element that is rarely preserved or developed in Mesozoic birds. For example, the preorbital ossification described here could be a palpebral, although we consider this to be much less likely owing to its preserved location, close to the midline of the skull.

Jugal

Only the left jugal is preserved in STM 3-8 ( Fig. 3G, H ). The maxillary process is as slender as the jugal process of the maxilla, similar to the jugal of Archaeopteryx ( Elżanowski & Wellnhofer, 1996; Kundrát et al., 2018; Rauhut et al., 2018) but in contrast to the relatively more robust condition in Sapeornis ( Hu et al., 2020a) . The rostral quarter of the maxillary process is slightly constricted and bears a depression in the distal end of the dorsal margin, defining the articulation with the maxilla. The articulation between the jugal and the maxilla is much shorter than in Sapeornis , in which the maxilla extends caudally almost to the base of the postorbital bar ( Hu et al., 2020a). An oval concavity is present centrally on the lateral surface of the maxillary process. A similar depression is also present in Archaeopteryx , although in a more rostral position ( Mayr et al., 2007; Rauhut, 2014), but is absent in most other Mesozoic birds (e.g. Sapeornis , Ichthyornis and enantiornithines; Wang & Hu, 2017; Field et al., 2018; Hu et al., 2020a). The quadratojugal process of the jugal of Jeholornis lacks the notch present in Sapeornis and many non-avian theropods, which is also absent in known enantiornithines but possibly present in Ichthyornis ( Rauhut, 2003; Xu et al., 2015; Wang & Hu, 2017; Field et al., 2018; Hu et al., 2020b). Because of this, the quadratojugal of Jeholornis articulates with the dorsolateral surface of the quadratojugal process of the jugal, differing from the wedge-like articulation seen in Sapeornis and other Mesozoic theropods (e.g. Linheraptor Xu et al., 2015 ; Xu et al., 2015; Hu et al., 2020a). The postorbital process of the jugal of Jeholornis is triangular with a broad base and is dorsally oriented. This contrasts with the caudodorsal orientation seen in Archaeopteryx and Sapeornis ( Mayr et al., 2007; Rauhut, 2014; Kundrát et al., 2018; Hu et al., 2020a). A shallow impression on the rostrolateral surface of the postorbital process defines the articulation with the postorbital, indicating the presence of a complete postorbital bar in Jeholornis .

Quadratojugal

The left quadratojugal is complete, but slightly disarticulated from the jugal ( Fig. 3E, F ). The jugal process is twice as long as the squamosal process and is more slender; both are bluntly tapered. The ventromedial surface of the jugal process contacts the jugal, in contrast to the inserting articulation with the caudal notch of the jugal in Sapeornis and most non-avian theropods ( Xu et al., 2015; Hu et al., 2020a). The squamosal process is reduced and does not contact the squamosal dorsally, similar to the condition in other Mesozoic birds, including Archaeopteryx , Sapeornis and various others (e.g. RapaxaƲis pani Morschhauser et al., 2009 and Cruralispennia multidonta Wange et al., 2017 ; Mayr et al., 2007; O’Connor et al., 2011; Rauhut, 2014; Wang et al., 2017b; Hu et al., 2020a).

Postorbital

The left postorbital is completely preserved and the right is broken in STM 3-8 ( Fig. 3I–K ). The postorbital is triradiate and more robust than that of Archaeopteryx ( Kundrát et al., 2018; Rauhut et al., 2018; Hu et al., 2020a). The jugal process is long and tapers ventrally, extending most of the skull height ventrally, and therefore forming most of the postorbital bar. In contrast, one specimen of Archaeopteryx preserves a slightly longer jugal process ( Rauhut et al., 2018), whereas others preserve a jugal process almost equal in length to the other processes ( Kundrát et al., 2018). The elongate jugal process of Jeholornis more closely resembles the condition in some enantiornithines (e.g. Longusunguis Wang et al., 2014 and enantiornithines LP4450 and IVPP V12707). However, it is much more robust than that of some other enantiornithines (Sanz et al., 1997; Hu et al., 2020b; Zhou et al., 2008). The squamosal process of the postorbital of Jeholornis is short, less than half the length of the frontal process, and has a sharply tapered end, whereas this process is longer in Sapeornis and Archaeopteryx ( Rauhut et al., 2018; Hu et al., 2020a). The dorsal surface of the squamosal process bears a concave facet for articulation with the squamosal.

Squamosal

Both squamosals are preserved, although only the right is complete in STM 3-8 ( Fig. 3L, M ). The squamosal is not fused to the braincase, similar to the condition in non-avian theropods, Archaeopteryx and enantiornithines (e.g. LP4450 and IVPP V12707; Elżanowski & Wellnhofer, 1996; Sanz et al., 1997; Rauhut, 2003; Norman et al., 2004; Xu et al., 2015; Rauhut et al., 2018; Wang et al., 2021). The rarity with which squamosals are preserved in other stem birds complicates interpretation of the morphology seen in Jeholornis . However, the concavity in the medial surface of this bone fits the otic process of the quadrate, and thus could be interpreted as the quadrate cotyle of the squamosal, supporting our identification of this element as a squamosal. The triangular, sharply tapered, rostroventrally directed process is identified as the postorbital process, resembling that in enantiornithine IVPP V12707 ( Wang et al., 2021), and contrasts with the forked condition in Archaeopteryx ( Elżanowski & Wellnhofer, 1996; Kundrát et al., 2018). The ventrally oriented quadratojugal process is short, with a blunt ventral margin, not contacting the quadratojugal. This suggests that the loss of the quadratojugal–squamosal contact might have evolved independently in Jeholornis and in ornithurines, but remained present in at least some enantiornithines (e.g. IVPP 12707) [although it also could have been regained secondarily as a derived feature ( Wang et al., 2021)]. It cannot be determined whether the dorsal portion is complete, hence the shape of the parietal and the paroccipital processes of the squamosal remain uncertain.

Quadrate

Both quadrates are almost completely preserved ( Fig. 4A–C ). The shaft of the quadrate extends from the otic process dorsally to the lateral condyle caudoventrally. The orbital process is broad and lateromedially thin, resembling that of Archaeopteryx ( Rauhut et al., 2018) , Sapeornis ( Hu et al., 2020a) and known enantiornithines [e.g. Zhouornis Zhang et al., 2013 ( Zhang et al., 2013) and Pterygornis Wang et al., 2014 ( Wang et al., 2015)], and differs from the narrow and rostrally projecting condition in Ichthyornis and crown birds, including the Late Cretaceous Asteriornis Field et al., 2020 ( Elżanowski & Stidham, 2010; Field et al., 2018, 2020). The otic process is plesiomorphically single headed, as in Sapeornis and enantiornithines ( Wang et al., 2015, 2021; Hu et al., 2020a), but differing from the divided otic capitulum and squamosal capitulum in neognaths, including Asteriornis ( Field et al., 2020) . A dorsoventrally oriented longitudinal ridge is present caudally on the medial surface of the otic process, similar to the condition in Sapeornis and enantiornithines ( Zhang et al., 2013; Wang et al., 2015; Hu et al., 2020a), defining the caudal margin of a gentle excavation on the medial surface of the orbital process. The lateral surface is also excavated by a similar dorsoventrally oriented longitudinal ridge, but it cannot be determined whether this is attributable to the lateromedially crushed preservation of the orbital process. No pneumatic foramen is observed, different from the condition in most modern birds and Late Cretaceous ornithurines (e.g. Ichthyornis and Asteriornis ; Elżanowski & Stidham, 2010; Field et al., 2018, 2020). However, two potential pneumatic recesses could be identified on the lateral surface of the quadrate in Jeholornis . Both the lateral and medial condyles are of a similar size and project caudally, defining a concave caudal margin for the quadrate.

Frontal

The frontals are tightly articulated with each other in STM 3-8, but not entirely fused, with the interfrontal suture clearly visible ( Fig. 4D, E ), similar to the condition observed in other Jeholornis specimens ( Lefèvre et al., 2014). The rostral half is narrow, half the mediolateral width of the expanded caudal half, along which a supraorbital rim is well developed laterally. This rim, together with the orbital margin of the frontal, form the dorsal and the medial walls of the orbit.

Parietal and braincase

The parietals and the braincase are crushed together, preventing segmentation of the individual elements ( Fig. 4F–H ). The parietals are in tight contact medially, similar to the frontals. Owing to poor preservation, it cannot be determined whether they are entirely fused, as in some non-avian theropods ( Clark et al., 2002; Yin et al., 2018), or if the interparietal suture still exists, as in Archaeopteryx ( Rauhut, 2014; Kundrát et al., 2018; Rauhut et al., 2018). The nuchal crest is straight and forms a well-developed ridge. The sagittal crest is absent in the conjoined parietals, but a supratemporal fossa reaching the dorsal portion of the parietal is still present. It is probable that the elements forming the occiput are either fused or sutured with each other, although this area is strongly crushed, preventing detailed observations. The paroccipital process is lateroventrally and somewhat caudally oriented and similar in length to Archaeopteryx ( Kundrát et al., 2018; Rauhut et al., 2018). The occipital condyle is rounded and bears a shallow depression on the caudoventral surface. Two rostrocaudally elongate fossae are identified as the basisphenoid recesses rostroventral to the occipital condyle visible in ventral view. The cultriform process is dorsally concave and slightly restricted in the distal end. We refrain from describing other features owing to the crushed preservation, which makes the braincase morphology difficult to interpret.

Endocast

The braincase of Jeholornis STM 3-8 is partly crushed and deformed, preventing a full endocranial reconstruction. Nevertheless, a mosaic of derived and plesiomorphic characters can be inferred from the dorsal surface of the endocast, defined by the ventral surface of the skull roof elements ( Fig. 4 ) ( Fabbri et al., 2017; Beyrand et al., 2019; Watanabe et al., 2019).

Ventrolateral shift of the midbrain endocast is clearly present ( Fig. 4 ), as observed in Archaeopteryx and other non-avian paravians ( Alonso et al., 2004; Balanoff et al., 2013). The forebrain endocast is expanded, which is a plesiomorphic condition shared among paravians ( Balanoff et al., 2013; Fabbri et al., 2017; Beyrand et al., 2019). The optic bulbs are laterally placed in comparison to more basal theropods, but not ventrally shifted as in modern birds ( Fabbri et al., 2017; Beyrand et al., 2019). The ventral surfaces of the frontals indicate broad, well-developed olfactory bulbs, as in non-avian dinosaurs and Archaeopteryx . The olfactory tracts are mediolaterally wide and cranially elongated, contrary to modern birds ( Balanoff et al., 2013; Fabbri et al., 2017; Beyrand et al., 2019), although they are relatively shorter in comparison to other non-avian coelurosaurs, such as tyrannosauroids, dromaeosaurids and troodontids ( Balanoff et al., 2013; Fabbri et al., 2017; Beyrand et al., 2019).

Vomer

The vomer is completely preserved ( Figs 5A, B , 6 ), although the pterygoid rami are slightly broken and distorted. The vomer is partly fused along approximately the one-quarter rostral portion, where the fused part is dorsoventrally arched such that its ventral surface is convex. Shallow impressions are located cranially on the ventral surface and are identified as the articular facets with the maxilla and, possibly, also the premaxilla. This indicates a dorsoventrally overlapping contact between the vomer and maxilla/premaxilla, resulting in limited cranial kinesis, similar to Sapeornis , enantiornithine IVPP V12707, non-avian dinosaurs and palaeognathous birds, but different from neognaths, in which the vomer contacts the maxillae more laterally ( Tsuihiji et al., 2014; Hu et al., 2019, 2020a; Wang et al., 2021). The unfused caudal portion of the vomers diverges into two caudal flanges (pterygoid rami) that are much longer than the fused rostral portion ( Hu et al., 2022) and proportionally much longer than the pterygoid rami of Archaeopteryx ( Elżanowski & Wellnhofer, 1996) . The vomers of Jeholornis lack the leaf-like caudodorsal process seen at the end of the pterygoid ramus in Sapeornis and some non-avian theropods (e.g. SinoƲenator ; Hu et al., 2020a), which is also absent in Archaeopteryx and enantiornithines (e.g. Gobipteryx ; Elżanowski & Wellnhofer, 1996; Chiappe et al., 2001; Hu et al., 2020a). Instead, the middle portions of the pterygoid rami are dorsoventrally expanded, possibly to accept the pterygoids or the parasphenoid rostrum. A shallow impression on the lateral surface of the dorsal expansion is interpreted as the palatine articulation, similar to the condition in the Late Cretaceous enantiornithine Gobipteryx ( Chiappe et al., 2001) . The only Early Cretaceous enantiornithine bird preserving relatively complete vomer morphology, IVPP V12707, is immature and lacks cranial fusion, preventing unambiguous comparisons ( Wang et al., 2021).

Palatine

Both palatines are complete in STM 3-8 ( Figs 5E, F , 6 ), preserved between the maxillae, and their morphology confirms identification of the palatine in the previously described J. palmapenis SDM 20090109.1/2 ( O’Connor et al., 2012; Fig. 7 ). The palatine resembles that of Archaeopteryx , being unreduced and mediolaterally broad between the jugal process and the choanal process ( Mayr et al., 2007). In SDM 20090109.1/2, the base of the choanal process is perforated by a foramen ( O’Connor et al., 2012), although whether this derives from preservation cannot be determined. The margin between the maxillary process and jugal process is straight, thus the jugal process is almost triangular. The jugal process connects the palatal complex and the lateral cranium and is therefore important for evaluating the degree of possible cranial kinesis. This process is well developed in non-avian dinosaurs ( Witmer, 1997; Tsuihiji et al., 2014; Hu et al., 2020a) and entirely reduced in crown birds and in known Late Cretaceous avian taxa, including the ornithuromorphs Hesperornis Marsh, 1872 and Ichthyornis and the enantiornithine Gobipteryx ( Elżanowski, 1991; Chiappe et al., 2001; Field et al., 2018). The condition in the Early Cretaceous enantiornithines and ornithuromorphs is unclear because only few, ambiguously identified palatines are preserved ( Xu et al., 2021). Based on the condition in Archaeopteryx and Jeholornis , we tentatively interpret the lack of a jugal process in the palatine of Sapeornis as a preservational artefact ( Hu et al., 2019, 2020a) and suggest that all non-ornithothoracine birds are likely to have possessed a plesiomorphic, nearly akinetic palate that contacted the lateral cranium tightly. The jugal process of the palatine mostly contacts the maxilla but also contacts the rostral tip of the jugal in Archaeopteryx ( Mayr et al., 2007) . However, based on the relative preserved position of the palatine, maxilla and jugal in STM 3-8, it is highly possible that the jugal process of the palatine only contacted the maxilla in Jeholornis ( Hu et al., 2022) . This interpretation is also consistent with the fact that the maxilla extends farther caudally in Jeholornis than it does in Archaeopteryx ( Mayr et al., 2007) . The medially hooked choanal process of the palatine articulates with the vomer medially, forming the caudal margin of the internal naris. The pterygoid wing of the palatine is long and slender, approximately half the maximum width of the palatine and forming more than half of its total length. Its caudal end is bluntly tapered, with straight lateral and medial margins. However, its caudal articulation with the pterygoid is unclear owing to the disarticulation.

Pterygoid

Both pterygoids are preserved in STM 3-8 but strongly crushed ( Figs 5C, D , 6 ). Although most morphologies are ambiguous, this element is elongate, and a flange ventrally developed along most of the palatine process indicates the presence of an ectopterygoid ( Hu et al., 2022). The palatine ramus is slender and might contact the parasphenoid rostrum and/or the pterygoid ramus of vomer rostrally. The quadrate ramus of the pterygoid is well developed and extrudes dorsally from the main body, indicating a large, lateromedial contact with the quadrate. The presence of a basipterygoid process like that present in Archaeopteryx and enantiornithine IVPP V12707 cannot be determined owing to mediolateral crushing in STM 3-8 ( Elżanowski & Wellnhofer, 1996; Wang et al., 2021).

Scleral ring

Specimen STM 3-8 preserves both scleral rings ( Fig. 5G, H ), although some scleral ossicles might have been lost during preservation. Each individual ossicle is nearly rectangular, with subequal margins. The articular facet for the adjacent ossicle forms approximately one-quarter of the surface, except for overplates (scleral ossicles that overlap both adjacent ossicles). The 3D models of the best-preserved and the most in situ articulated plates were used to reconstruct the complete scleral ring, suggesting that it consisted of a total of 15 ossicles.

Dentary

Both mandibles are completely preserved, with the right one slightly disarticulated, in STM 3-8 ( Fig. 8 ). The dentaries are not fused at the symphysis and curve rostroventrally along their entire length ( Fig. 8C, D ). Nevertheless, the dorsal margin of the dentary is only weakly convex. The symphysis is expanded ventrally such that the rostral one-third of the ventral margin of the dentary is more strongly convex, giving a rostrally downturned appearance. The rostral margin of the symphysis is inclined caudoventrally. The lateral surface of the dentary is marked by several neurovascular foramina rostrally. The alveolar shelf is well developed along the dorsomedial margin.

The caudal margin of the dentary of Jeholornis is forked, as in Sapeornis but with a much longer ventral ramus, whereas the dorsal ramus is longer in Sapeornis ( Hu et al., 2020a) . However, the weakly forked condition in both Sapeornis and Jeholornis , in which one process is much shorter than the other, contrasts with the strongly forked condition in confuciusornithiforms and neornithines ( Chiappe et al., 1999; Elżanowski et al., 2018; Wang et al., 2019a) and the unforked condition reported in most enantiornithines (O’Connor & Chiappe, 2011). The caudal half of the dentary is excavated laterally by a large, concave fossa. Meckel’s groove, which is located on the medial surface of dentary, is either shallow or absent, whereas it is well developed in Archaeopteryx and non-avialan theropods ( Elżanowski & Wellnhofer, 1996; Rauhut, 2003; Norman et al., 2004). Meckel’s groove is also present in Ichthyornis and some enantiornithines ( O’Connor et al., 2013b; Field et al., 2018; Wang et al., 2021), although comparisons are limited by the paucity of medially preserved dentaries among Mesozoic birds.

Three alveoli are located rostrally in the dentary and house small teeth, with crown heights ~80% that of maxillary teeth. The dentary teeth are different from the maxillary teeth, being smaller in size and less expanded at the root, strongly resembling the dentary teeth of Sapeornis , which are interpreted as having reduced functionality ( Wang et al., 2017a; Hu et al., 2020a). The alveoli are separated by interdental bone, resulting in interdental spaces of similar lengths to the tooth diameters. Another similar-sized tooth is dislocated and preserved close to both the right maxilla and the right dentary ( Fig. 8F ). This tooth is additional to the count pertaining to the combined maxillae and dentaries (ten alveoli in total, but 11 teeth are present) and might represent a disarticulated replacement tooth, consistent with the absence of a root. Despite being more similar to the size of the maxillary teeth, owing to the incomplete preservation of the roots of both maxillary and dentary teeth, it cannot be determined confidently which tooth is the functional tooth related to this replacement tooth. The rostral-most portion of the dentary, rostral to the first dentary tooth, is excavated by a shallow concavity; a similar structure was also reported in some Mesozoic ornithuromorphs ( Dumont et al., 2016; O’Connor et al., 2022).

Postdentary bones

The prearticular and the articular are entirely fused to the surangular ( Fig. 8G–I ). The surangular is shorter than the dentary and has a well-developed dorsally projecting coronoid process on its caudal half. Cranial to the coronoid process the surangular tapers rostrally, articulating with the lateral surface of the dentary. In medial view, a shallow mandibular fossa excavates the surangular above the prearticular. A narrow rostral mandibular fenestra is visible, defined ventrally by the prearticular and dorsally by the surangular, and a caudal mandibular fenestra is absent, whereas in Confuciusornis both the rostral and caudal mandibular fenestrae are well developed ( Chiappe et al., 1999; Elżanowski et al., 2018; Wang et al., 2019a).

The angular is small and positioned underneath the surangular and dentary ( Fig. 8L, M ). The articular lacks the dorsal deflection present in Sapeornis , more resembling that of Archaeopteryx ( Hu et al., 2020a) . No pneumatic foramen is present in the articular. The medial cotyle is broader than the lateral cotyle, divided by the intercotylar crest. The medial process is not tapered. The retroarticular process is weakly developed.

The splenial is triangular in lateral profile ( Fig. 8J, K ). The rostral wing is short and ends abruptly; the caudal wing is long, slender and sharply tapered, and the dorsal wing is broadly convex, defining an obtuse angle.

DENTITION

To determine the accurate dentition of Jeholornis and whether the presence or absence of maxillary teeth is diagnostic among species ( O’Connor et al., 2012), 83 jeholornithiform specimens were examined based on the STM and IVPP collections (main slab and counter slab are counted as one specimen). All specimens were preliminarily referred to Jeholornis , based on their large body size, long bony tail, sternum with lateral trabecula absent and expanded caudalmost pair of sternal ribs ( Zheng et al., 2014). Of these 83 specimens, teeth are preserved in 24 specimens (Supporting Information, Table S2), among which the presence and number of dentary teeth could be evaluated by either dentary teeth in situ or clearly observable alveoli in the dentary in 17 specimens. Of these 17 specimens, three dentary teeth are preserved in four specimens (STM 2-3 counter slab, 2-38, 2-49 and 3-8), and no more than three dentary teeth are preserved in any specimen (Supporting Information, Table S2). Identified by their proportionately small size, disarticulated dentary teeth were identified in three specimens (STM 2-40, 3-5 counter slab and 3-6). Only four specimens exhibit a laterally exposed and unbroken dentary with no visible dentary teeth (STM 2-15, 2-46, 3-13 and 3-16). However, the presence of teeth cannot be ruled out, because the alveolar margin is not exposed and disarticulated teeth in other specimens indicate that teeth can fall free of their alveoli post-mortem. No specimen preserves unequivocal evidence of the absence of dentary teeth.

In situ maxillary teeth or clearly observable alveoli in the maxilla are preserved in 14 specimens. Among them, two maxillary teeth are preserved in eight specimens (STM 2-3 counter slab, 2-4, 2-30, 2-37, 2-50, 2-54, 3-8 and 3-32), and no more than two maxillary teeth are preserved in any specimen. One maxillary tooth is visible in another six specimens (STM 2-2 main and counter slabs, 2-7, 2-14, 2-45, 2-55 and 3-33 counter slab). Most of these teeth represent a single, similar morphology and are larger than the dentary teeth. However, in STM 2-45, 2-55 and 3-33 counter slab, the single visible maxillary tooth is remarkably smaller than those in other specimens, indicating that they are in the early stages of erupting. Displaced maxillary teeth are also identified based on their relatively large size in another four specimens (STM 2-2 main and counter slabs, 2-13, 2-40 and 2-49). Four specimens exhibit an unbroken, intact ventral margin of the maxilla and an absence of maxillary teeth (STM 2-15, 2-24, 2-38 and 2-39).

Maxillary teeth are not preserved in STM 2-37, but two empty alveoli are clearly visible, and the teeth of the right maxilla of STM 2-37, which is buried in the matrix, are also displaced, with two empty alveoli. Combined with the presence of the displaced maxillary/dentary teeth in other specimens, this could suggest that the maxillary/dentary teeth of the jeholornithiforms easily fall out of the alveoli and are lost during the preservation process, resulting in an artificial variance in the dentition number across individuals. Although the ventral margins of the maxilla and the dorsal margins of the dentary are intact but lacking teeth or alveoli in a few specimens, it should be noted that if the teeth were displaced and the alveoli filled by the matrix, it would be difficult to observe the alveoli in those 2D slabs that lack professional preparation.

Together, these observations strongly suggest that the dentition in all jeholornithiforms is 0–2–3 (number of premaxillary teeth–number of maxillary teeth– number of dentary teeth).

BIOLOGY OF JEHOLORNIS

The olfactory bulb-to-forebrain ratio suggests that olfaction was important in Jeholornis , indicated by a higher value in comparison to closely related stem birds, such as Archaeopteryx and Confuciusornis (Supporting Information, Table S3). The proportions seen in Jeholornis are more similar to more basal coelurosaurs, such as Bambiraptor Burnham et al., 2000 , Garudimimus Barsbold, 1981 (an ornithomimosaur) and Dilong Xu et al., 2004 (an early-diverging tyrannosauroid) ( Zelenitsky et al., 2011).

To test quantitatively another important sensory ability (vision) of Jeholornis , we measured the orbital length and the scleral ring outer/inner diameters of STM 3-8, using the protocol described by Choiniere et al. (2021; Fig. 9 ). With the addition of these data, we reran ‘Analysis 5’ in the phylogenetic flexible discriminant analysis of the scleral ring and orbit morphology described by Choiniere et al. (2021), because their analysis has the most accurate posterior classification rates among all other versions of similar analyses ( Choiniere et al., 2021). The results indicate an inference of diurnal habits in Jeholornis (posterior probability of being nocturnal <0.01; Supporting Information, Table S4).

Figures

Figure 2. Three-dimensional snapshots of rostral cranial elements of Jeholornis STM 3-8: lateral (A) and medial (B) views of the articulated left premaxilla, maxilla and lacrimal; lateral view (C), lateral line drawing (D), dorsal view (E) and ventral view (F) of the fused premaxillae; lateral (G), caudolateral (H) and medial (I) views of the right lacrimal; lateral view (J), lateral line drawing (K) and medial view (L) of the left maxilla. Scale bar: 5 mm. Arrows indicate the rostral direction. Labels for articular facets are in red. Abbreviations: a.j, jugal articulation; a.l, lacrimal articulation; a.m, maxilla articulation; a.n, nasal articulation; aof, antorbital fenestra; a.p, premaxilla articulation; a.pt, palatine articulation; apm, ascending process of maxilla; cdr; caudodorsal ramus of lacrimal; fpp, frontal process of premaxilla; jpm, jugal process of maxilla; ld, lateral depression of maxilla to receive the maxillary process of premaxilla; lf, lacrimal foramen; mf?, potential maxillary fenestra; mpp, maxillary process of premaxilla; mt, maxillary teeth; nf, nutrient foramina; pam, palatal process of maxilla; ppm, premaxillary process of maxilla; ppp, palatal process of premaxilla; rdr, rostrodorsal ramus of lacrimal; vr, ventral ramus of lacrimal.

Figure 3. Three-dimensional snapshots of lateral and dorsal cranial elements of Jeholornis STM 3-8: dorsal lateral (A) and ventromedial (B) views of the left nasal; dorsal views of the right (C) and left (D) preorbital ossification; lateral (E) and medial (F) views of the left quadratojugal; lateral (G) and medial (H) views of the left jugal; lateral (I), laterocaudal (J) and medial (K) views of the left postorbital; lateral (L) and medial (M) views of the right squamosal. Scale bar: 5 mm. Arrows indicate the rostral direction. Labels for articular facets are in red. Dashed lines indicate broken margins. Abbreviations: a.m, maxilla articulation; a.po, postorbital articulation; a.q, quadrate articulation; a.qj, quadratojugal articulation; a.s, squamosal articulation; fpo, frontal process of postorbital; jpo, jugal process of postorbital; jpq, jugal process of quadratojugal; mpj, maxillary process of jugal; mpn, maxillary process of nasal; ocj, oval concavity of jugal; poj, postorbital process of jugal; pos, postorbital process of squamosal; ppn, premaxillary process of nasal; qjp, quadratojugal process of jugal; qjs, quadratojugal process of squamosal; spo, squamosal process of postorbital; spq, squamosal process of quadratojugal.

Figure 4. Three-dimensional snapshots of the quadrate and skull roof of Jeholornis STM 3-8: lateral (A), medial (B) and caudal (C) views of the right quadrate; dorsal (D) and ventral (E) views of the frontals; dorsal (F), ventral (G) and caudal (H) views of the braincase. Scale bar: 5 mm. Arrows indicate the rostral direction. Labels for articular facets are in red. Dashed lines indicate broken margins. Abbreviations: a.n, nasal articulation; a.pa, parietal articulation; bsr, basisphenoid recess; cup, cultriform process; lc, lateral condyle of quadrate; mc, medial condyle of quadrate; nc, nuchal crest; oc, occipital condyle; op, orbital process of quadrate; om, obital margin; otp, otic process of quadrate; prp, paroccipital process; sr, supraorbital rim; stf, supratemporal fossa. Labels of neurocranial elements are not abbreviated.

Figure 5. Three-dimensional snapshots of the palatal elements and scleral ring of Jeholornis STM 3-8: ventral (A) and dorsal (B) views of the vomer; medial (C) and lateral (D) views of the right pterygoid; ventral (E) and dorsal (F) views of the right palatine; left view of the scleral ring (G) and the scleral bone (H). Scale bar: 5 mm. Arrows indicate the rostral direction. Labels for articular facets are in red. Abbreviations: a.m, maxilla articulation; a.pa, palatine articulation; a.psr, parasphenoid rostrum articulation; a.pt, pterygoid articulation; a.v, vomer articulation; cp, choanal process; mpa, maxillary process of palatine; op, overplate; par, palatine ramus of pterygoid; pgw, pterygoid wing; pj, jugal process of palatine; pmr, premaxillary ramus of vomer; ptr, pterygoid ramus of vomer; qur, quadrate ramus of pterygoid.

Figure 6. Comparisons of postorbital and palatal complex in Aves. Taxa without the postorbital present indicate that this element has been entirely reduced and fused to the frontal to be the postorbital process in these taxa. Dashed lines indicate uncertain structures. Scale bar: 5 mm only for the postorbitals.

Figure 7. Photograph and reidentified cranial elements of Jeholornis palmapenis. Scale bar: 5 mm. Revised from O’Connor et al. (2012). Abbreviations: pt, palatine; qj, quadratojugal.

Figure 8. Three-dimensional snapshots of mandibular elements of Jeholornis STM 3-8: lateral (A) and medial (B) views of the left mandible; lateral (C), medial (D) and dorsal (E) views of the left dentary; medial view of the left dentary with maxillary teeth (F); lateral (G), medial (H) and dorsal (I) views of the left surangular; lateral (J) and medial (K) views of the right splenial; lateral (L) and medial (M) views of the angular. Scale bar: 5 mm. Arrows indicate the rostral direction. Abbreviations: cnp, coronoid process; dt, dentary teeth; lct, lateral condyle; mct, medial condyle; mf, mandibular fenestra; mfs, mandibular fossa; mp, medial process; mt, maxillary teeth; pra, prearticular; rcd, rostral concavity of dentary; rt, replacement tooth; rtp, retroarticular process.

Figure 9. Scleral ring measurements of Jeholornis STM 3-8. Reassembled three-dimensional cranial model of Jeholornis STM 3-8 used for measurements, following Hu et al. (2022). Scale bar: 5 mm.Abbreviations: EXT, scleral ring outer diameter; INT, scleral ring inner diameter; OL, orbital length.