A taxonomic revision of the Sinopterus complex (Pterosauria, Tapejaridae) from the Early Cretaceous Jehol Biota, with the new genus Huaxiadraco

Morphological survey

This specimen (Fig. 1) was originally described by Wang & Zhou (2003). It exhibits a relatively slender rostrum (~36% of jaw length), with a very low, incipient premaxillary crest and a low dentary crest. The rostrum is gently downturned at about 14° relative to the posterior occlusal line. The premaxillary crest is parabolical in outline. The nasoantorbital fenestra length/height ratio is not readily clear due to a slight anteroventral displacement of the orbitotemporal region. Still, it can be restored as somewhere between 2.8 and 3.2 (by restoring the position of the orbitotemporal region based on the inferred location of the quadratomandibular joint as indicated by the proportions of the mandible). The orbit has been described as subcircular (e.g., Andres, Clark & Xu, 2014), since its height and length are subequal. However, it may be described as subquadrangular due to the angular corners. This differs from the typical elongated piriform condition (higher than long, with a round dorsal margin and tapered ventral margin) of tapejarids and azhdarchoids in general (e.g., Kellner & Campos, 2007). Still, a tapered shape of the lower orbital margin is still present (in the jugal). The lacrimal process of the jugal is subvertical (only slightly anterodorsally oriented). A pair of slender, anteroventrally directed, and medially placed descending nasal processes is present. The posterior cranial crest processes (the posterior process of the premaxillae, and the frontoparietal crests) curve upwards. The quadrate is posteriorly reclined at ~160° relative to the palatal plane. The observable cervical formula is III < IV > V > VI > VII. The scapula is about 1.30 the length of the coracoid. The coracoid exhibits a clear ventral flange. The humeral deltopectoral crest is tongue-like and its long axis is sub-perpendicular relative to the long axis of the humeral shaft. The pteroid accounts for 43% of ulnar length. Metacarpal I is elongate, reaching the carpal region, while metacarpals II and III are reduced and restricted distally. Metatarsal I is the longest of the metatarsals (Wang & Zhou, 2003; Zhang et al., 2019).

Ontogenetic assessment

This specimen has already been regarded as a juvenile before (Kellner, 2010; Zhang et al., 2019). A large number of skeletal elements remain unfused in this specimen: scapulacoracoid, humeral epiphysis, carpal series, extensor tendon process of the first wing phalanx, and tibiotarsus. Several skull elements also remain unfused. It is clear that this specimen is a juvenile indeed, if compared to more ontogenetically advanced specimens in which the abovementioned elements are fused, such as in the postcranial skeleton of D2525 (File S1, Sheet 6). At a wingspan of 1.2 m, it would be conceivable that it was an advanced juvenile, older than smaller specimens such as the holotypes of S. gui (0.8 m), S. lingyuanensis (~0.85 m) and H. atavismus (0.85 mm, see further below), and younger than larger specimens such as the holotype of S. jii (1.6 m) and D2525 (2 m).

Remarks

This specimen is the holotype of Sinopterus dongi—the first genus and species of tapejarid to be described for the Jiufotang Fm. and Jehol Group as a whole. The validity of this genus and species has never been questioned.

Morphological survey

The specimen (Fig. 2) is unfortunately badly preserved, with quite damaged and crushed bone surfaces (Li, Lü & Zhang, 2003). Still, general outlines of some of the skull and appendicular bones can be discerned. The skull is exposed mostly in left lateral view, except for the posterior region which seems to be broken and exposed in a slightly dorsolateral view. The rostrum accounts for ~39% of total jaw length. It is very slender (RI = 0.33) and crestless, while the dentary symphysis bears a very shallow crest. The nasoantorbital fenestra is very elongate (length/height ratio ~3.2). Quadrate inclination is unclear due to the bad preservation of the posterior region of the skull. Details of the cervical series are unclear due to bad preservation. The coracoid ventral margin bears a flange, similar to other Sinopterus complex specimens (see below). The deltopectoral crest of the humerus is rectangular, proximally placed, and bears a long axis roughly perpendicular relative to the main humeral shaft. The relative length of metacarpals I–III cannot be assessed. Of the wing fingers, only a first phalanx is preserved, thus obscuring wing phalanges proportions. Unfortunately, not much further details can be assessed due to the very limited preservational quality of the specimen.

Ontogenetic assessment

Unfused elements: palatal and posterior skull bones, dorsal centra and neural arches, scapula and coracoid, pelvic elements, tibia and fibula (entirely unfused). Other ontogenetic correlates cannot be assessed. This specimen is clearly a very young juvenile. It is also the second smallest of all Jehol tapejarid specimens (second to the holotype of Nemicolopterus crypticus), with an estimated wingspan of only 64 cm (Kellner & Campos, 2007).

Remarks

This specimen is the holotype of Sinopterus gui—the second species of tapejarid to be described for the Jiufotang Fm. and Jehol Group as a whole (Li, Lü & Zhang, 2003). It was subsequently recognized as a very young juvenile (Kellner & Campos, 2007). The validity of this species has been questioned several times, in all such cases being regarded as a junior synonym of S. dongi even when multiple Jiufotang tapejarid species were accepted, on the basis that it could not be distinguished from S. dongi (Kellner & Campos, 2007; Kellner, 2010; Zhang et al., 2019). This is problematic because recent publications have simply repeated the interpretation of S. gui being indistinguishable from S. dongi while not comparing S. gui to other more recently named species considered as valid, thus not justifying why it is indistinguishable from S. dongi only and not from any further species (e.g., Zhang et al., 2019). First described by Li, Lü & Zhang (2003), these authors recognized it as distinct from Sinopterus dongi at a species-level, yet sufficiently similar to be placed in the same genus. Originally, Li, Lü & Zhang (2003) proposed the following diagnosis for the new species: “[e]leven dorsal vertebrae fused into notarium, and they are nearly equal in length. At least four sacral vertebrae, humerus longer than scapula, wing metacarpal slightly shorter than the first wing phalange, the distal end of the deltopectoral process not expanded, ratio of the femur to the tibia is approximately 0.49” (Li, Lü & Zhang, 2003: p. 445). Later, Kellner & Campos (2007) observed that this specimen does not present a notarium (which is an advanced ontogenetic feature). Instead, it represents a very young, juvenile specimen (Kellner & Campos, 2007; Kellner, 2010). Most authors have, since then, been unable to distinguish S. gui from S. dongi, and thus interpreted the holotype of Sinopterus gui as a juvenile specimen of Sinopterus dongi (e.g., Kellner & Campos, 2007; Zhang et al., 2019), although Kellner (2010) noticed that it could represent a juvenile of some other Jiufotang tapejarid instead, such as Huaxiapterus corollatus (therein referred to as Sinopterus corollatus). The interpretation of the holotype of S. gui as a juvenile of S. dongi (and not any other Jiufotang tapejarid species) has been maintained by Zhang et al. (2019) without further justifications, even though these authors accept the validity of several other Sinopterus species (S. lingyuanensis, S. corollatus, S. benxiensis, and S. atavismus). We maintain here that S. gui is indeed indistinguishable from S. dongi except for the complete absence of a premaxillary crest in the former, which is easily attributed to ontogeny (Witton, 2013; Zhang et al., 2019).

Morphological survey

This almost complete specimen includes a partial skull, although the posterior region is disarticulated and damaged (Fig. 3). The rostrum is ventrally deflected at 14° relative to the posterior palatal plane. The rostrum exhibits a premaxillary crest. It is similar in shape to that of S. dongi (parabolical in outline), despite being larger. It is distinct from the premaxillary crest condition of other proposed species, such as the pointed premaxillary crests of Huaxiapterus atavismus (both specimens, the holotype XHPM 1009 and the referred specimen IVPP V 22338) or the trapezoidal crests of Huaxiapterus corollatus and Huaxiapterus benxiensis, or the crestless conditions seen in Sinopterus gui and Sinopterus lingyuanensis. Most of the posterior region of the skull is badly damaged, except for the left jugal which is partially preserved. The jugal is triradiate, unlike the tetraradiate condition seen in Tapejara wellnhoferi (Wellnhofer & Kellner, 1991), Caiuajara dobruskii (Manzig et al., 2014) and Tupandactylus navigans (Beccari et al., 2021). The lacrimal and postorbital processes of the jugal describe a roughly perpendicular angle. The proportions of the nasoantorbital fenestra cannot be readily measured due to the damaged nature of the posterior region of the skull, but an estimate can still be given based on the location of the lacrimal process of the jugal (about three times as long as high). As with the premaxillary crest, the dentary crest is also larger than in S. dongi. Only two disarticulated cervical vertebrae can be seen, so that the cervical formula cannot be assessed. Pteroid length is equivalent to about 44% of the ulna length. Metacarpal I is elongate, extending for at least 90% the length of metacarpal IV. Wing proportions are closest to the holotype of S. dongi (Fig. 3; File S1, Sheets 1, 3). Pedal elements are entirely disarticulated, so that the metatarsal formula cannot be assessed.

Ontogenetic assessment

The holotype of S. jii has been regarded as a juvenile compatible with the holotype of S. dongi, given their similarity in lacking bone fusion between posterior skull elements, scapulocoracoid, humeral epiphyses, carpals, extensor tendon process of the first wing phalanx, and tibiotarsus (Kellner, 2010). However, it is worth noticing that the dorsal centra and arches of GMN-03-11-001 are partially fused (they bear a visible suture, but are not found disassociated), unlike some entirely unfused and disassociated dorsal centra and arches seen in the holotype of S. dongi. This suggests that GMN-03-11-001 is slightly more ontogenetically developed than the holotype of S. dongi, both as juveniles. Concerning body size, GMN-03-11-001 is larger than the holotype of S. dongi, with a humerus of 79 mm in length and a wingspan of 1,602 mm.

Remarks

This specimen was originally described as representing a new genus and species, Huaxiapterus jii (Lü & Yuan, 2005). Subsequent publications have considered it either as a species of Sinopterus, as S. jii (Kellner & Campos, 2007; Pinheiro et al., 2011; Kellner, 2013), or as a junior synonym of Sinopterus dongi (Wang & Zhou, 2006; Witton, 2013; Zhang et al., 2019), thus invalidating the genus Huaxiapterus. Still, other researchers still considered H. jii as valid and as a distinct taxon, with the genus Huaxiapterus being valid (Andres, Clark & Xu, 2014; Lü et al., 2016).

This taxon was originally diagnosed based on cranial crest development: premaxillary and dentary crests deeper than in Sinopterus dongi and shallower than in Tapejara wellnhoferi (see Lü & Yuan, 2005), though without precise quantitative comparisons. Later, this species has been regarded as a junior synonym of Sinopterus dongi: Wang & Zhou (2006) were unable to find differences between the holotypes of the two species, and thus synonymized them. At the time, these two species (together with Sinopterus gui) were the only named species within the Sinopterus complex. We maintain that the holotypes of S. gui and S. jii are indistinguishable from S. dongi, and further add that S. jii shares with S. dongi the following features: metacarpal I articulating with the carpus, and wing phalanx 4/phalanx 1 length ratio about ~0.30, which distinguish these proposed taxa from other proposed taxa such as H. corollatus and H. benxiensis (see below). Sadly, these features are uncertain in the holotype of S. gui.

Morphological survey

This specimen is almost complete, although some skeletal regions are badly damaged and anatomical details are obliterated, particularly the posterior region of the skull, post-cervical vertebrae, and the pedes (Fig. 4). The skull exhibits a trapezoidal premaxillary crest and a shallow dentary crest. The rostrum is relatively robust, akin to that of the holotype of Huaxiapterus jii and unlike the holotypes of S. dongi or S. gui. The rostrum is ventrally deflected by 21° (contra 14° in the holotypes of S. dongi and H. jii). The nasoantorbital fenestra is relatively short, with an estimated length/height ratio of about 2.2 (based on its length as inferred from the location of the quadratomandibular joint, as indicated by the preserved mandible, as it roughly correlated to the posterior margin of the nasoantorbital fenestra in sinopterines and tapejarids overall; e.g., Kellner & Campos, 2007; Kellner, 2013; Lü et al., 2016). A clear occlusal gap is present between the dentary and the rostrum (as originally indicated, see Lü et al., 2006a), unlike what has been represented in some reconstructions (e.g., Witton, 2013). The cervical series is partially obscured by the radius and ulna, which lay over cervicals IV–V, hindering assessment of their relative lengths. Metacarpals I–III are reduced, and it can be seen that metacarpals I and II do not contact the carpus, reaching only about a third of the length of metacarpal IV. Wing proportions deviate from previously reported specimens in that the fourth wing phalanx is relatively shorter, accounting for only ~20% of the first phalanx (contra ~30% in the holotypes of S. dongi and S. jii).

Ontogenetic assessment

In this specimen, unfused skeletal elements include the posterior skull bones, scapulocoracoid, and extensor tendon process of the first wing phalanx. Unfortunately, fusion of humeral epiphyses cannot be assessed due to poor preservation. The tarsals are fused to the tibia, forming a tibiotarsus, as can be seen from the right hindlimb. The carpals also seem to be fused into distal and proximal syncarpals. Thus, this specimen seems to be relatively more mature than the holotypes of S. dongi, S. gui and H. jii, as a subadult. It is roughly equivalent to the holotype of H. jii, with a humerus of 75 mm in length and a wingspan of 1,560 mm.

Remarks

This specimen was designated as the holotype of Huaxiapterus corollatus by Lü et al. (2006a). The species-level validity of this species (irrespective of its generic status) has been mostly accepted (Pêgas, Leal & Kellner, 2016; Lü et al., 2016; Zhang et al., 2019; Andres, 2021), except for Witton (2013) who preliminarily proposed that all Jiufotang tapejarids were synonymous with S. dongi. It is interesting to note that, although Naish, Witton & Martin-Silverstone (2021) preliminarily corroborated Witton (2013) view, they highlighted that at least the holotype of H. corollatus could potentially represent a new taxon (based on its limb proportions), pending further study.

The taxon Huaxiapterus corollatus was originally diagnosed on the basis of cranial crest features, namely crest shape (“hatchet-shaped”), position (level with the anterior margin of the nasoantorbital fenestra), and orientation (“short axis perpendicular to the anterodorsal margin of the nasoantorbital fenestra”; see Lü et al., 2006b). These conditions differ starkly from what is seen in the holotypes of S. dongi, S. gui and H. jii. However, as noticed by Witton (2013) and Naish, Witton & Martin-Silverstone (2021), cranial crest features used alone make for dangerous taxonomic decisions, as they could rather reflect ontogenetic or sexual variations. Still, the holotype of H. corollatus also differs from the holotypes of S. dongi and S. jii in exhibiting a reduced metacarpal I, and in wing proportions (File S1, Sheet 1). H. corollatus exhibits a reduced wing phalanx 4, which accounts for ~20% of the length of the first wing phalanx, contra ~30% in the previously named S. dongi and H. jii. Naish, Witton & Martin-Silverstone (2021) noticed that the holotype of H. corollatus was an apparent outlier within the Sinopterus complex regarding limb proportions, leading them to propose that it could be a potentially valid taxon pending further study.

Morphological survey

D2525 is an almost complete postcranial skeleton, lacking the skull, part of the anterior cervical series, part of the posterior dorsal series, and the sacral and caudal series (Fig. 5). The preserved cervical vertebrae, as well as shoulder girdle and right humerus, are badly damaged. Although previously unreported, the ?fourth cervical (exposed in ventral view, retaining some tridimensionality) clearly exhibits a pneumatic foramen piercing its lateral surface. The sternum is approximately square, with the posterior margin convex. The left coracoid bears a well-developed ventral flange. The left humerus is exposed in dorsal view, and no dorsal proximal pneumatic foramen can be seen in this specimen, as in IVPP V 23388 (Zhang et al., 2019). The ulnar crest is rounded. The humeral shaft is mostly straight, except for the distal portion which is slightly anteriorly recurved. Metacarpals I–III are tightly appressed to metacarpal IV on the distal metacarpal region on both sides. Metacarpal I extends for only about 40% of the length of metacarpal IV (Fig. 5). Wing proportions are very similar to the holotypes of H. corollatus and H. benxiensis, with the fourth wing phalanx corresponding to ~20% the length of the first wing phalanx (contra ~30% in S. dongi and S. jii). Wing phalanges are exposed in ventral view, and a longitudinal ridge can be seen in phalanges 2 and 3, similarly to H. atavismus (Lü et al., 2016) and IVPP V 23388 (Zhang et al., 2019). In the pedes, metatarsal I is distinctively shorter than metatarsal II, which is the longest.

Ontogenetic assessment

Specimen D2525 is the third largest of all known Jiufotang tapejarids, with a 2-m wingspan (Lü et al., 2006b), and also appears to be one of the most osteologically mature ones. Observable fused elements include dorsal neural arches and centra, the scapulocoracoid, the syncarpals, and the extensor tendon process of the first wing phalanx. Partial fusion (almost complete fusion, with faint indications of sutures) can also be seen in the pelvis, tibiotarsus, and tarsal elements. The presence of a notarium cannot be assessed due to preservational limitations, since the anterior dorsal series is preserved in ventral view and badly crushed.

Remarks

This specimen was originally described as a new specimen of Sinopterus dongi, based on the assertion that the limb proportions of D2525 were most similar to S. dongi than to S. gui, H. jii or H. corollatus, which were the four existing nominal species at the time (Lü et al., 2006b). Such referral has never been contested in the literature. Contrary to previous reports (Lü et al., 2006b), the limb proportions of D2525 are most similar to the holotype of H. corollatus, and not S. dongi (see File S1, Sheets 1, 3). In fact, D2525 is herein considered as indistinguishable from H. corollatus, with which it shares a shortened metacarpal I (about 40% the length of metacarpal IV, contra >90% in S. dongi and S. jii) and a shortened fourth wing phalanx (~20% of first phalanx length, contra ~30% in S. dongi and S. jii). It differs from the holotypes of S. dongi and H. jii in wing proportions and in metatarsals I–II relative length (metatarsal II is the longest one in D2525, instead of metatarsal I as in S. dongi).

Morphological survey

This specimen consists on a virtually complete specimen (Fig. 6). However, some anatomical regions are damaged and/or partially obscured, mainly the torso region (with the post-cervical vertebral series, sternum, ribs, and scapulocoracoid). The rostrum is built similarly to the holotype of H. corollatus, with a downward deflection of 20°. The premaxillary crest is slightly larger than in the holotype of H. corollatus, but it is similar in being distinctively anterodorsally protrusive with abrupt limits, unlike the smoothly-transitioning borders of the parabolical crests of the holotypes of S. dongi and S. jii. Despite broken, the premaxillary crests seems to have been trapezoidal in shape, as in the holotype of H. corollatus. The posterior process of the premaxillae is steeply dorsally recurved. An elongate posterior spine (posterior process of the premaxillae + frontoparietal crest) is present, much larger than in the holotype of S. dongi. The nasoantorbital fenestra is approximately as elongate as in S. dongi, with a length/height ratio of about 2.4. The long axis of the nasal process is very deflected anteriorly, unlike the almost verticalized nasal process seen in the holotype of S. dongi. The shape of the jugal (as seen from the lacrimal and postorbital processes) demonstrates that the orbit was piriform, with a tapered ventral margin, and quite higher than wide, unlike the subquadrangular orbit of S. dongi. The quadrate is posteriorly inclined at about 153°. Not much further detail can be seen due to extensive superficial damage. The observable cervical formula is III < IV < V > VI. Both humeri are badly damaged, with only a section being exposed. The original description reported on an oddly short humerus only 55% the length of the femur (Lü et al., 2007), but this seems to have been based on the fairly incomplete right humerus. We reidentify here the damaged proximal and distal limits of the left humerus, which indicate it was comparable to that of other Jiufotang tapejarids (about 80% of femur length) instead of oddly short (Figs. 6A and 6D). The extension of the pteroid is unclear. Metacarpal I confidently extends for only ~40% the length of metacarpal IV. The proximal extension of metacarpals II and III is unfortunately obscure, since it is unclear if the proximal tips are broken or not. Wing proportions closely match H. corollatus, with relatively short fourth wing phalanges (20% the length of the first phalanx). The relative length of metatarsals I–III overall cannot be assessed due to poor preservation.

Ontogenetic assessment

In this specimen, fused skeletal elements include the humeral epiphyses, syncarpals, the extensor tendon process of the first wing phalanx, and the tibiotarsus. Scapulacoracoids cannot be observed due to damage. Only the posterior skull bones are still unfused to the rest of the skull. Indeed, posterior skull bones are known to be among the last skeletal elements to fuse in pterosaurs (e.g., Kellner, 2015). This specimen thus exhibits a relatively advanced level of skeletal fusion, fitting well with the concept of an advanced subadult among pterosaurs (e.g., Kellner & Tomida, 2000). This specimen is clearly one of the most mature ones in the present sample, along with D2525 (see above), since all of the previously described specimens lack fusion of the extensor tendon process of the first wing phalanx. It is roughly equivalent in body size to the holotypes of H. jii and H. corollatus, with a humerus of 82 mm in length and a wingspan of 1,600 mm.

Remarks

The species H. benxiensis was erected on the basis of BXGM V0011 and attributed to the genus Huaxiapterus, following H. jii and H. corollatus. The validity of this species has been mostly accepted without further comments (Pinheiro et al., 2011; Kellner, 2013; Pêgas, Leal & Kellner, 2016; Zhang et al., 2019; Andres, 2021), except for works that argued for the “restrictive taxonomic scheme” of the Sinopterus complex, which regarded it as most likely a junior synonym of S. dongi along with all other nominal species of Jiufotang tapejarids (Witton, 2013; Naish, Witton & Martin-Silverstone, 2021).

Huaxiapterus benxiensis has been regarded as distinct from H. corollatus on the basis of an “elongate parietal spine”, “well-developed premaxillary crest”, and a shallow groove on the occlusal surface of the dentary symphysis (Lü et al., 2007). Witton (2013) noticed that crest-related features could be influenced by ontogeny rather than interspecific variation. We further note that the “shallow groove” on the anterior end of the symphysis corresponds to the anterior occlusal depression (ubiquitous to tapejarids), interrupted posteriorly by a transverse ridge (similar to the condition seen in Bakonydraco galaczi; see Ősi, Weishampel & Jianu, 2005). This condition can also not set H. benxiensis apart from any other proposed Jehol tapejarid species, since preservation precludes the verification of this feature in other type specimens. H. benxiensis is here considered as indistinguishable from H. corollatus, with which it shares a rostrum deflection of ~20°, a reduced metacarpal I, and a reduced fourth wing phalanx (~20% of first wing phalanx length). Both H. benxiensis and H. corollatus further differ from S. dongi and S. gui in exhibiting a relatively shorter nasoantorbital fenestra (only 2.2–2.4 in height/length ration, contra ~3 in S. dongi and S. gui).

Morphological survey

The holotype of S. lingyuanensis exhibits a relatively fine preservation, comprising an almost complete skeleton lacking only some distal wing phalanges and the tail. Some anterior trunk and appendicular elements, such as posterior cervical vertebrae, some dorsal vertebrae, ribs, sternum, and pectoral girdle, are severely crushed against each other and cannot be discerned (Fig. 7). Other than that, most other skeletal elements are discernible, with decent surface preservation despite crushing. The skull is exposed mainly in left lateral view, and the occipital region is laterally displaced towards the left, thus being visible in a somewhat posterolateral view. The rostrum is entirely crestless and slender, accounting for 44% of total jaw length. The rostrum is gently deflected at 12° relative to the palatal plane. Beneath the anterior level of the nasoantorbital fenestra, a bulge is present on the jaw margin, indicating the presence of a slight lateral palatal expansion similar to what is seen in Tapejara and Caiuajara (Wellnhofer & Kellner, 1991; Manzig et al., 2014). The nasoantorbital fenestra is quite elongate, being 3.25 times longer than high. The nasals exhibit a pair of descending nasal processes, which are subvertical and elongate, similar to S. dongi and unlike the anteriorly directed, short condition seen in H. benxiensis. The orbit is roughly subquadrangular, about as wide as high, similarly to S. dongi. The divergence angle between the lacrimal and postorbital processes of the jugal is about ~90°, similar to S. dongi and H. jii but unlike H. benxiensis (~68°), which exhibits a piriform orbit. The quadrate is reclined at about 160°. A small, short frontoparietal crest is present, extending beyond the occiput. The mandible is exposed in dorsal view. Sadly, the occlusal surface is not well-preserved. Still, it can be seen that a slight lateral expansion occurs at the posterior region of the symphysis, as in Tapejara and Caiuajara (Wellnhofer & Kellner, 1991; Manzig et al., 2014), matching the slight lateral palatal expansion beneath the anterior margin of the nasoantorbital fenestra. The dentary symphysis and the retroarticular process account for, respectively, 53% and 4% of total mandibular length. Atlas and axis cannot be observed. The observable cervical formula is III < IV > V > VI, similar to S. dongi and unlike H. benxiensis in which the fifth cervical is the longest. The mid-cervicals clearly exhibit at least one pneumatic foramen piercing their lateral sides. The pteroid accounts for 47% of ulnar length. Sadly, the distal extensions of metacarpals I–III are obscured by metacarpal IV. The relative length of the fourth wing phalanx is also unknown. In the pedes, the metatarsal formula is I < II > III > IV, similar to D2525 but unlike S. dongi.

Ontogenetic assessment

JPM-2014-005 is small-sized, with a skull length of 112 mm and an estimated totalized wingspan of ~850 mm. Skull elements are mostly unfused, to the exception of the premaxillomaxillae and dentaries. Postcranial unfused elements include the humeral epiphyses, carpals, extensor tendon process of the first wing phalanx, pelvic elements, and tibiotarsus. Fusion (or lack thereof) of further elements cannot be assessed. The available information suggests JPM-2014-005 is a young juvenile, as the holotypes of S. gui and S. dongi.

Remarks

This specimen was originally designated as the holotype of a new species, S. lingyuanensis, by Lü et al. (2016). This was subsequent to Witton (2013) proposition that all Jiufotang tapejarids formed an ontogenetic continuum of S. dongi, which was not accepted by Lü et al. (2016). Later, Zhang et al. (2019) expressed their approval over the validity of S. lingyuanensis, without further comments. Later, Naish, Witton & Martin-Silverstone (2021) echoed the proposition of Witton (2013) that all proposed Jiufotang tapejarids most likely represented a single species (to the potential exclusion of H. corollatus), including S. lingyuanensis.

The species Sinopterus lingyuanensis was proposed based on the following features: nasoantorbital fenestra length/height ratio 3.2, rostral index 3.03, femur/tibia length ratio 0.66, and wing phalanx 2/wing phalanx 1 length ratio 0.85 (Lü et al., 2016). However, all of these values fit well within the spectrum seen in the Sinopterus complex (File S1, Sheet 1) and cannot set S. lingyuanensis apart from other species, particularly from S. dongi, S. gui and H. jii which also exhibit nasoantorbital fenestra about three times as long as high (distinct in this regard from the holotypes of H. corollatus and H. benxiensis). Still, S. lingyuanensis does differ from S. dongi in metatarsal configuration (I ≈ II, rather than I > II), and also differs from H. benxiensis in orbit shape (subcircular rather than piriform), nasal descending process configuration (subvertical and elongate, rather than anteriorly directed and short), and cervical formula (IV > V, rather than IV < V). It also differs from both H. corollatus and H. benxiensis in exhibiting a gentler rostrum deflection (12° rather than 20°). The significance of these variations will be discussed further below, in the Discussion section.

Morphological survey

Despite virtually complete, many skeletal remains of this specimen are quite jumbled together, preventing the observation of much anatomical data (Fig. 8). The rostrum exhibits a very small, triangular-shaped premaxillary crest, whose apex is anterodorsally oriented and located posterior to the anterior margin of the nasoantorbital fenestra (this configuration is distinct from any other tapejarid specimen previously published, but similar to specimens IVPP V 23388 and D4019). The rostrum is slender, ventrally deflected by 14°, and with a deflection point anteriorly located, similarly to S. lingyuanensis. A small, yet clearly perceivable, occlusal gap is present. The dentary bears a slight dorsal eminence, as well as a low dentary crest. The observable cervical formula is III < IV > V ≅ VI > VII > VIII. Not much can be discerned from the remaining of the axial skeleton, and the same is true for the pectoral girdle. The pteroid accounts for 40% of ulnar length. Unfortunately, the relative lengths of the metacarpals cannot be assessed. Wing phalanx proportions are a close match for S. dongi and S. jii (File S1, Sheet 1), and distinct from H. corollatus, H. benxiensis and D2525 which exhibit a comparatively reduced fourth wing phalanx about 20% the length of the first wing phalanx (File S1, Sheet 1). Metatarsal I is shorter than metatarsal II, which is the longest, unlike S. dongi.

Ontogenetic assessment

XHPM 1009 is a small-sized specimen, with an estimated skull length of ~120 mm and total wingspan of ~850 mm. Unfused skeletal elements include the carpals, extensor tendon process of the first wing phalanx, and tibiotarsus. Sadly, not much else can be discerned. Still, this specimen is compatible with a young juvenile, not much more advanced than the holotype of S. gui.

Remarks

This specimen was originally designated as the holotype of a new species, H. atavismus, by Lü et al. (2016). This was subsequent to Witton (2013) proposition that all Jiufotang tapejarids formed an ontogenetic continuum of S. dongi. Still, Zhang et al. (2019) accepted the validity of this species, which they assigned to the genus Sinopterus, as Sinopterus atavismus. Later, Naish, Witton & Martin-Silverstone (2021) echoed the proposition of Witton (2013) in interpreting all Jiufotang tapejarids as probable synonyms, to the inclusion of S. atavismus.

The species H. atavismus was originally diagnosed based on the presence of a squared premaxillary crest and of a ventral groove on the second wing phalanx. As noticed by Zhang et al. (2019), the crest is actually not squared (Fig. 8), and cranial crest morphology should be viewed with caution when discussing pterosaur diagnoses; while the ventral groove on the second wing phalanx is probably common within tapejarids (see Kellner, 2004; Zhang et al., 2019), although admittedly hard to ascertain in other Sinopterus complex specimens due to heavy crushing. H. atavismus shares with S. dongi and S. lingyuanensis a fourth cervical vertebra longer than the fifth, distinct from H. benxiensis and other tapejarids. H. atavismus differs from the holotype of S. dongi in pedal morphology, showing the typical condition (metatarsal II the longest), and not the unique condition seen in S. dongi (metatarsal I the longest). H. atavismus differs from H. corollatus and H. benxiensis in exhibiting a gentler rostrum deflection and a more elongate fourth wing phalanx (File S1, Sheet 1), and from D2525 in the latter aspect as well.

Morphological survey

This specimen has been described and figured in detail by Zhang et al. (2019). The rostrum is elongate and slender, with a gentle ventral deflection of 14°. The rostrum deflection point lies anterior to the anterior margin of the nasoantorbital fenestra, as in S. lingyuanensis and H. atavismus. The premaxilla produces a small, subtriangular crest, as noted by Zhang et al. (2019), similar to that seen in the holotype of H. atavismus. Despite the incomplete, disarticulated nature of the skull remains, the nasoantorbital fenestra is notoriously elongate, and was confidently over three times as elongate as high (Zhang et al., 2019). The jugal is triradiate, and the angle formed between the lacrimal and postorbital processes is very wide (~90°, similar to S. dongi and S. lingyuanensis), indicating the orbit was probably subquadrangular in shape, and not ventrally tapered (piriform) as in H. benxiensis. The postoccipital extension of the premaxillae is elongate and curved posterodorsally. The observable cervical formula is IV > V ≅ VI > VII > VIII > IX (contra Zhang et al., 2019). The coracoid exhibits a deep ventral flange proximally. Metacarpals II and III are reduced, while the preserved metacarpal I extends for about 85% the length of metacarpal IV. The proximalmost tip of metacarpal I is missing due to a crack in the slab. Sadly, pteroid and wing phalanges 4 are missing. Metatarsal I is shorter than metatarsal II, which is the longest.

Ontogenetic assessment

With a humerus of 79 mm in length and a wingspan of ~1,600 mm, this specimen is similar in size to the holotype of H. benxiensis, and amongst the largest specimens in the Sinopterus complex. Partially fused elements include the carpals, pubis and ischium, ilium and pubosichiadic plate, notarium and synsacrum, and tibiotarsus—these elements are tightly bound, though with faint, visible sutures. Unfused elements include the scapulocoracoid, humeral epiphysis extensor tendon process, and orbitotemporal bones. Despite not an adult, this specimen is clearly more mature than the juvenile holotypes of S. gui, S. dongi, S. lingyuanensis and H. atavismus, and could be considered a subadult.

Remarks

This specimen has been attributed to Sinopterus atavismus (=Huaxiapterus atavismus) by Zhang et al. (2019). No alternative attributions have been given by any other workers, except for Naish, Witton & Martin-Silverstone (2021) who preliminarily considered that all Jiufotang tapejarids were most likely conspecific with S. dongi (to the potential exception of H. corollatus only).

This fairly complete specimen was described recently by Zhang et al. (2019), who were unable to distinguish it from Huaxiapterus atavismus and thus referred the new specimen to this species (using the combination Sinopterus atavismus). Zhang et al. (2019) considered that three features allowed IVPP V 23388 to be identified as H. atavismus: the shape of the premaxillary crest, the shape of the anterodorsal margin of the premaxilla, and the proportions between metatarsals I and II (Zhang et al., 2019). However, the first two features are influenced by the development of the premaxillary crest, which, as discussed above, is prone to sexual and ontogenetic variation, and should be viewed with caution before being utilized in diagnoses, as will be discussed further below in this work.

Furthermore, proportions between metatarsals I and II in IVPP V 23388 (metatarsals I/II = ~0.90) and the holotype of H. atavismus are rather close to those of other specimens such as S. lingyuanensis (File S1, Sheets 1, 3), and thus this condition should be seen with caution. These three specimens also match well in the configuration of the nasoantorbital fenestra (over three times as long and high) and rostrum deflection angle (12°–14°), also matching S. dongi and H. jii in these regards, being all distinct from H. corollatus and H. benxiensis (with nasoantorbital fenestrae about 2.3 times as long as high, and rostrum deflections of 20°–21°). We regard that IVPP V 23388, along with the holotype of H. atavismus, are both indistinguishable from S. lingyuanensis. They are all also undistinguishable from S. dongi except for the metatarsi proportions.

Morphological survey

This specimen has been recently described and figured in detail by Shen et al. (2021). It consists of a partial postcranial skeleton, comprising most of the cervical and dorsal series, the forelimbs, and partial hindlimbs. The observable cervical formula is III < IV > V > VI > VII > VIII > IX. Single pneumatic foramina can be seen piercing the lateral sides of some cervical vertebrae (at least III, IV and V; unclear in others). Metacarpal I is elongate, with a preserved portion accounting for about 90% of metacarpal IV length; the proximal tip is missing and it may have been longer. The first wing phalanx exhibits two pneumatic foramina piercing the ventral side of the proximal region, similar to Keresdrakon vilsoni (see Kellner et al., 2019). The fourth phalanx is relatively large, accounting for 36% the length of the first wing phalanx, approaching more closely the value seen in the holotype of S. dongi and in IVPP V 23388 (30%). In the pedes, metatarsal I is the longest one.

Ontogenetic assessment

As originally indicated by Shen et al. (2021), this specimen is clearly a juvenile as seen from the lack of fusion between many skeletal elements: the humeral epiphyses, scapulocoracoid, the extensor tendon process of the first wing phalanx, the carpal elements, tibia and fibula, tibia and proximal tarsals, and neural arches and centra of most dorsal vertebrae. Only the neural arches and centra of cervical vertebrae and anterior dorsal vertebrae are fused. With a humerus of 55 mm in length and wingspan of 1,135 mm, this specimen is similar in size to the holotype of S. dongi (humerus 58 mm in length and wingspan of 1,200 mm), which is also interpreted as a juvenile.

Remarks

This specimen has been referred to S. dongi by Shen et al. (2021), as accepted by Zhou et al. (2022) and not commented on the literature any further so far. Shen et al. (2021) noticed that D3072 shares with the holotype of S. dongi similar limb proportions as well as a reduced metatarsal I (shorter than metatarsals II and III), which has been considered a diagnostic apomorphy for S. dongi within the expansive taxonomic scheme of the Sinopterus complex (Zhang et al., 2019).

Morphological survey

This specimen comprises an almost complete skeleton, although not very well preserved. Many of the elements are articulated, except for most skull and manual elements (Fig. 9). The rostrum is slender and gently decurved (by 13°) and bears a well-developed, heaped crest. The dorsal margin of the premaxilla is slightly jagged. The jugal-quadratojugal-quadrate complex indicates the quadrate was strongly reclined (by 162°). Unfortunately, the jugal is incompletely preserved and lacks a lacrimal process. A well-developed and posterodorsally inclined frontoparietal crest is present. The cervical vertebrae not very well-preserved and not much can be observed beyond their lengths. The fourth cervical is the longest. The trunk region is very crushed and not much can be observed. Limb elements bear slightly abraded surfaces, precluding observation of much detail. Scapulocoracoid, humeral epiphyses, and carpal elements are unfused. As preserved, metacarpal I reaches 82% the length of metacarpal IV, but its proximal end is unclear and it may have been longer. Both pedes are badly preserved and not much can be discerned.

Ontogenetic assessment

The new specimen D4019 is small-sized, with a humerus length of 64 mm (only slightly larger than the holotype of S. dongi, with a 58 mm humerus). It is too incomplete for a confident wingspan estimate. Based on the lack of ossification between scapulocoracoid elements, humeral epiphyses, and carpal elements, this individual is inferred as a juvenile.

Morphological summary

This specimen includes an almost complete skull (exposed in left lateral view), incomplete cervical series (exposed in dorsal view), incomplete forelimbs, and incomplete hindlimbs (Fig. 10). The rostrum is slender and deflected ventrally at an angle of 20°. A slight ventrolateral tilt of the plane of exposure of the rostrum reveals that the occlusal surface is sulcate, sporting thick tomial edges that emarginate an elongate sagittal excavation. Slit-like neurovascular foramina pierce the lateral surface of the rostrum close to the tomial edge (unclear in the occlusal surface). The premaxillary crest is large and protrusive. The anterior margin is roughly perpendicular to the main dorsal margin of the rostrum, anterodorsally oriented, similar to H. benxiensis and H. corollatus, and thus seems to have been originally trapezoidal in shape. The posterodorsal edge of the premaxillary crest is damaged, but it seems to have been anteroposteriorly longer than dorsoventrally high. The proportions of the nasoantorbital fenestra are not directly clear due to the disarticulation of the posterodorsal margin (nasal and lacrimal), but can be estimated at around 2.5 based on its length and mid-height. The dentary symphysis accounts for roughly 55% of total mandibular length, and sports a dorsal eminence as well as a low ventral crest. The anterior symphyseal region is pierced by slit-like foramina close to the occlusal line. Although the forelimbs are incompletely preserved, a partial humerus and both wing fingers are completely preserved. Metacarpal I preserves a clear proximal end and extends for only about 40% the length of metacarpal IV. The fourth wing phalanx accounts only for 20% of the first wing phalanx length. Metatarsal II is the longest one.

Ontogenetic assessment

This specimen lacks fusion of the posterior skull elements, humeral epiphyses, carpals, extensor tendon process of the first wing phalanx, and tarsal elements. Unfortunately, neither pectoral nor pelvic girdles are preserved. This specimen may be a juvenile or a an early subadult. It is roughly equivalent in body size to the holotypes of H. jii, H. corollatus, and H. benxiensis, with a humerus of 79 mm in length and a wingspan of 1,546 mm.

Morphological summary

This specimen includes most of the skeleton, including a premaxillomaxilla, an almost complete mandible, incomplete cervical and dorsal series, and almost complete fore and hindlimb elements (Fig. 11). The rostrum is relatively robustly built and ventrally deflected at an angle of 20°. The rostrum deflection point is located roughly beneath the anterior margin of the nasoantorbital fenestra, where a bulge also seems to indicate the presence of a slight lateral palatal expansion. The premaxillary crest is unfortunately incompletely preserved, but it extends anterior to the anterior margin of the nasoantorbital fenestra and its broad base suggests it was relatively large. Despite the incompleteness of the skull, the length of the nasoantorbital fenestra can be assessed based on the location of the remains of the base of the lacrimal process of the jugal. The height of the nasoantorbital fenestra was measured at its mid-length, to account for the typical position of its maximum height limit as seen in more complete specimens. In this way, the length/height ratio of the nasoantorbital fenestra of BPMC 104 can be estimated at roughly 2.3. The lacrimal process of the jugal is not preserved. The jagged dorsal skull margin is reminiscent of the conditions seen in Tupandactylus (Campos & Kellner, 1997; Frey, Martill & Buchy, 2003), suggesting it sported a soft tissue crest. The dentary exhibits a dorsal eminence as well as a low ventral crest. Cervical formula cannot be assessed. The sacral vertebrae (number unclear) are partially fused and bear intersacral fenestrae. The coracoid bears a large ventral flange. The extension of metacarpal I can be assessed due to the good preservation of its proximal tip, despite the loss of some of the diaphysis (Fig. 11G). It extends for 41% the length of metacarpal IV, similar to H. benxiensis. The first wing phalanx exhibits a single pneumatic foramen on its ventral surface. The fourth wing phalanx is relatively reduced, corresponding to 20% of first wing phalanx length. In the pelvic girdle, the medial margin of the postacetabular process is excavated by a fossa, similar to Tapejara wellnhoferi and Vectidraco daisymorrisae (Eck, Elgin & Frey, 2011; Naish, Simpson & Dyke, 2013). The neck of the postacetabular process is relatively thick and elongate, similar to Vectidraco daisymorrisae (Naish, Simpson & Dyke, 2013) and unlike the rather constricted condition seen in Tapejara wellnhoferi (Eck, Elgin & Frey, 2011) or short condition seen in Tupandactylus navigans (Beccari et al., 2021). The femoral head exhibits a thick neck, with no visible constriction in posterior view. The greater trochanter is well-developed, and a large pneumatic foramen is present near its base. The distal end of the femur is expanded. In lateral view, the femur bows posteriorly. Two (?femoral) unfused epiphyses are present near the proximal end of the tibia. In the pedes, metatarsal II is the longest one.

Ontogenetic assessment

This specimen exhibits fusion of the scapulacoracoid, pelvic girdle (with closed, but still faintly visible, sutures), and distal tarsals. The extensor tendon process of the wing phalanx is still unfused, as are the posterior skull elements, ?femoral epiphyses, and proximal tarsals. The specimen is thus interpreted as a subadult. Concerning body size, this is the second largest specimen in our sample, with a humerus of 100 mm in length and a wingspan of 2,124 mm. This is one of only three Jiufotang tapejarid specimens at the 2-m wingspan size-class, along with D2525 (see above) and BPMC 107 (see below).

Morphological summary

Despite being relatively complete, this specimen is badly preserved—most bones are jumbled together, and most bone surfaces are badly weathered or cracked beyond the point of bearing relevant anatomical details (Fig. 12). Notwithstanding, the outlines of some bones and structures still reveal some interesting data. The skull, exposed in left lateral view, exhibits a trapezoidal premaxillary crest that is conspicuously protrusive, higher than anteroposteriorly long. The shape of the rostrum and the configuration of its ventral deflection are unclear. The nasoantorbital fenestra is about 2.2 times as long as high. The orbit seems to have been piriform. The dentary symphysis bears a dorsal eminence and a low ventral crest. Measurements for visible limb bones are given in File S1 (Sheet 1), but not much further comparative information can be retrieved. Wing proportions closely match those of H. corollatus and H. benxiensis, with the fourth wing phalanx accounting for roughly 20% the length of the first wing phalanx. The second metatarsal is the longest. Not much further information can be assessed.

Ontogenetic assessment

Unfused elements include posterior skull bones and humeral epiphyses, carpals, the extensor tendon process of the first wing phalanx, and the tibiotarsus. It is, in this way, compatible with the holotype of S. dongi in both ontogenetic correlates and also body size, with a humerus of 69 mm in length and a wingspan of 1,288 mm (File S1, Sheet 6). This specimen may thus be regarded as a juvenile.

Morphological summary

This small specimen preserves mainly a partial skull (missing the rostrum) and partial forelimbs other than partial cervical and dorsal series, although not much can be observed (Fig. 13). A triangular, dorsally oriented premaxillary crest is present, located anterior to the inferred anterior limit of the nasoantorbital fenestra, similarly to the holotype of H. atavismus and specimens IVPP V 23388 and D4019. The dorsal edge of the premaxillary crest, and of the posterior process of the premaxilla as well, is jagged (as in Tupandactylus spp.; see Frey, Martill & Buchy, 2003), indicating the potential presence of soft tissue extension. The proportions of the nasoantorbital fenestra are unclear due to the incompleteness of the rostrum and disarticulation between the premaxillomaxilla and the posterior skull region. The shape of the jugal indicates the base of the orbit was broad, implying the orbit was probably subquadrangular/subcircular in shape. The first metacarpal is quite elongate, reaching at least 95% the length of the wing metacarpal.

Ontogenetic assessment

Unfused elements include the posterior skull elements, humeral epiphyses and the carpals. In terms of size, this specimen is relatively small in the Jiufotang tapejarid sample, with a humerus length of 69 (close to specimen D4019, and intermediate between the holotypes of S. dongi and H. jii; File S1, Sheets 1, 6). It is too incomplete for a confident wingspan estimate. The specimen is interpreted as a juvenile.

Morphological summary

This specimen comprises an almost complete skeleton, despite exhibiting badly preserved bone surfaces (Fig. 14). The rostrum is relatively slender and exhibits only a faint, incipient premaxillary crest, very similar to the holotype of S. dongi. The nasoantorbital fenestra is very elongate, with an estimated length/height ratio of about 3. The orbital region is not preserved. The posterodorsal region of the skull exhibits a short frontoparietal crest. The dentary symphysis is exposed in ventral view. It exhibits a dentary crest, which is dorsoventrally crushed and thus appears as a crushed keel. The dentary symphysis accounts for about half of mandibular length. The posterior region of the symphysis is damaged. The left mandibular ramus is complete, including the articular region and the retroarticular process, allowing for estimation of the location of the quadratomandibular articulation in the skull despite the absence of a preserved quadrate (and hence allowing for a rough estimation of the proportions of the nasoantorbital fenestra). The cervical series is incompletely preserved, and the longest cervical vertebra cannot be assessed. The preserved wings exhibit morphology and proportions comparable to the holotype of Sinopterus dongi, although metacarpals I–III cannot be assessed (File S1, Sheet 3). The sternum exhibits a rounded posterior margin. Metatarsal I is slightly longer than metatarsal II.

Ontogenetic assessment

Specimen BPMC 107 is the largest Jiufotang tapejarid specimen within the sample analyzed here, with a humerus length of 111 mm and a totalized wingspan of ~2.18 m (only slightly larger than D2525 and BPMC 104). It bears several signs of an advanced ontogenetic stage, including complete fusion of the scapulacoracoid, humeral epiphyses, syncarpals, extensor tendon process, tibiotarsus, and almost complete fusion of the pelvic elements. Only the posterior skull elements were probably not completely fused, given their absence (presumably derived from disarticulation, implying lack of complete fusion). This specimen thus may be an advanced subadult close to skeletal maturity.

Rostrum, proportions (RI and RV)

Some variation in rostrum proportions in the Sinopterus complex had already been noted by Zhang et al. (2019). RI values vary from 2.85 (Sinopterus lingyuanensis) to 1.33 (BPMC 104), and RV values range from 6.53 (Huaxiapterus atavismus) to 3.5 (BPMC 104). It is clear that smaller, younger specimens tend to exhibit slenderer rostra, while larger, more mature specimens exhibit stouter rostra. In the present work, our SMA analysis indicates that both RI and RV are negatively allometric relative to body size. Since the measurement of RI is directly influenced by the presence and development of premaxillary crests, this result indicates that premaxillary crests grow with ontogeny in the Sinopterus complex. This result was to be expected, as cranial crest development is already well-known as an ontogenetic feature in pterosaurs, as demonstrated by taxa such as Caiuajara dobruskii (Manzig et al., 2014), Anhanguera spp. (Pinheiro & Rodrigues, 2017), and Pteranodon spp. (Bennett, 1993).

It is clear, as already noted by Witton (2013) and Naish, Witton & Martin-Silverstone (2021), that cranial crest development (and by extension, RI) should not be utilized as a tool for diagnosing potential Sinopterus complex species. Furthermore, we add here that the same is true for rostrum measurements that disregard crest development (i.e., RV), which show that the rostrum itself (exclusive of cranial crests) also develop to be stouter with ontogeny in the Sinopterus complex.

Rostrum, deflection angle

The ventral deflection of the rostrum is one of the most conspicuous cranial features of tapejarids, as seen in Caupedactylus, Tapejara, Tupandactylus, Caiuajara, Afrotapejara, Wightia, Eopteranodon, and in the Sinopterus complex. In the Sinopterus complex, Morphotype I exhibits a deflection angle range of 12°–15°, whereas the range is 20°–21° in Morphotype II (File S1, Sheet 5). Our SMA analysis indicates that this variation is not correlated to body size, and is thus interpreted as not ontogenetic in nature.

Martill et al. (2020a) had already reported on the intra- and interspecific variation of this feature within tapejarids. According to our own observations, deflection angles (as measured between the long axis of the deflected rostrum and the long axis of the maxilojugal bar) vary between 21°–25° in Tupandactylus imperator, 23°–25° in Tup. navigans, 25°–28° in Tapejara wellnhoferi, and 32°–37° in Caiuajara dobruskii (File S1, Sheet 5). We were unable to reproduce the measurements given by Martill et al. (2020a) for Caiuajara dobruskii, which produced a remarkably larger variation spectrum. This is probably explained by the variation in the shape of the palatal expansion bulge in Caiuajara dobruskii, which hampers the accurate measurement of the deflection angle if the maxillojugal bar posterior to it is not preserved (L. Piazentin, 2022, personal communication), and thus measurements of deflection angles in Caiuajara should be restricted to specimens with relatively complete maxillojugal bars. However, intraspecific variation in rostrum deflection in tapejarids does not seem to surpass a ~5° range.

Given the patterns of variation in other tapejarid species (File S1, Sheet 5), as well as the pattern that, within the Jiufotang sample, angles of 20°–21° are exclusive to Morphotype II while the other morphotypes are restricted to angles of 12°–15° (a difference which is statistically significant as indicated by our Kruskal-Wallis analysis), and that this variation is not related to body size, we regard this as a potential taxonomic signal for the Sinopterus complex.

Nasoantorbital fenestra, length/height ratio

Within the Jehol tapejarid sample, Morphotype I stands out due to its relatively elongate nasoantorbital fenestra, which is typically ~3 times as long as high (File S1, Sheet 1), as seen in the holotypes of S. dongi, S. jii, S. gui and S. lingyuanensis, and as can be roughly inferred from the holotype of S. atavismus as well as specimen IVPP 23388V. This contrasts with Morphotype II, in which the range for the nasoantorbital fenestra length/height ratio is 2.1–2.5. This seems to be roughly the typical condition for tapejarids, as seen in Eopteranodon lii (~2.1), Caupedactylus ybaka (~2.1), Tupandactylus imperator (∼2.5), and Tupandactylus navigans (2.1–2.2). Caiuajara and Tapejara stand out in exhibiting a relatively low ratio (~1.3; see File S1, Sheet 1). In this way, the particularly elongate nasoantorbital fenestra in Morphotype I is highly distinctive. Our SMA analysis indicates that nasoantorbital fenestra length/height ratio is not correlated to body size, and is thus interpreted as not ontogenetic in nature. It is worth highlighting that the elongate nasoantorbital fenestra of Morphotype I can be found in all skulls attributed to this morphotype, irrespective of ontogenetic stages—from the small juvenile holotype of S. gui to the large, subadult holotype of S. jii. Furthermore, our Kruskal-Wallis analysis indicate a significant difference between Morphotypes I and II regarding this feature. This feature could thus suggest a taxonomic distinction between Morphotype I and other tapejarids.

Nasal process, shape

In the holotypes of S. dongi and S. lingyuanensis (Morphotype I), the descending nasal processes are long, subvertically oriented, and extremely thin. This condition is also present in Tupandactylus navigans (Frey, Martill & Buchy, 2003). In Thalassodromeus sethi, the nasal process is also subvertical, although it is extremely reduced instead of elongated (Pêgas, Costa & Kellner, 2018). In the holotype of H. benxiensis (Morphotype II), however, the nasal process is anteriorly oriented, as already pointed out before in the data matrix of Andres, Clark & Xu (2014), although this feature was not explicitly reported in the original description of the specimen (Lü et al., 2007). Despite the substandard preservation of the nasal bones in the holotype of H. benxiensis (BXGM V0011), it can be seen upon close inspection that the right nasal, seen in medial view and partially overlayed by the incomplete left nasal, displays a preserved nasal process that is anteriorly inclined. Thus, we corroborate the coding provided by Andres, Clark & Xu (2014) and regard this very unusual condition as unique and possibly of taxonomic value for Morphotype II.

Orbit shape

Within the Sinopterus complex, variation exists concerning orbit shape, with the orbit being piriform in some specimens and subquadrangular/subcircular in others. This variation can be translated into the measurement of the angle between the lacrimal and postorbital processes of the jugal (ventral orbit angle), whereby subquadrangular orbits exhibit an angle close to 90° while piriform orbits exhibit a lower, acute angle. This difference in orbit shape distinguishes Morphotypes I and II. Morphotype I exhibits orbit angles of ~90°, as seen in the holotypes of S. dongi and S. lingyuanensis, as well as in IVPP 23388 V and BPMC SC04. Morphotype II exhibits lower values, with angles of 68° in the holotype of H. benxiensis and 65° in BPMC SC003. Within tapejarids, the peculiar subquadrangular orbit shape seen in Morphotype I is unique to this morphotype and to Eopteranodon lii (see below), with the piriform condition being the typical one, as seen in Morphotype II, Tapejara wellnhoferi, Tupandactylus navigans, Tupandactylus imperator, Caiuajara dobruskii, and Caupedactylus ybaka, as well as in thalassodromids (e.g., Pêgas, Costa & Kellner, 2018) and chaoyangopterids (e.g., Wu, Zhou & Andres, 2017). Our SMA analysis indicates that this feature is not correlated to body size. In effect, in Morphotype I, a perpendicular orbit angle can be found in both juveniles (holotypes of S. dongi and S. lingyuanensis, and specimen BPMC 106) and advanced subadults (IVPP 23388 V and holotype of H. jii). Our Kruskal-Wallis analysis indicate a significant difference between Morphotypes I and II regarding this feature. In this way, we regard that the distinctive orbit shape of Morphotype I most likely indicates distinctiveness between Morphotypes I and II.

Quadrate reclination

Within tapejarids, quadrate reclination (as measured between the maxillary ramus of the jugal and the quadrate) is usually between 140°–150°, as seen in Tapejara wellnhoferi (145°), Caiuajara dobruskii (147°), Tupandactylus navigans (145°), and Caupedactylus ybaka (150°). The holotype of Huaxiapterus benxiensis falls within this pattern, with a quadrate reclination of 147°. However, Morphotype I specimens exhibit a stronger quadrate reclination of 160°–162°, as seen in the holotypes of S. dongi and S. lingyuanensis, and specimens D4019 and BPMC 104 (File S1, Sheet 1). According to our SMA analysis, this variation is uncorrelated to body size. Quadrate reclination has been regarded as of taxonomic importance for Pteranodon (Bennett, 1994), tapejarids (Kellner, 2013), Nurhachius (Zhou et al., 2019), and wukongopterids (Zhou et al., 2021). Based on pterosaur species known from multiple specimens, intraspecific variation in quadrate reclination seems to surround 3°–6°, as seen in Pterodactylus antiquus, Aerodactylus scolopaciceps, Hamipterus tianshanensis, Pteranodon longiceps, and Pteranodon sternbergii (see Zhou et al., 2019). Although we do expect some minor influence of preservational distortion over this feature, this influence should also be accounted for in the 3°–6° variation pattern seen in the aforementioned species, as, similar to Jiufotang tapejarids, these are also represented by crushed specimens preserved on slabs as in the cases of Pterodactylus, Aerodactylus, and Pteranodon (e.g., Bennett, 1994). We thus regard this variation in the Sinopterus complex, which surpasses 10°, as of potential taxonomic value.

Cervical formula

Concerning the relative lengths of cervical vertebrae, the typical pterodactyloid condition is that the fifth is the longest one. This can be observed in chaoyangopterids (Leal et al., 2018; Wu, Zhou & Andres, 2017), azhdarchids (Naish & Witton, 2017), and Tupuxuara (Shen et al., 2021). However, some variation exists within tapejarids. In Tapejara wellnhoferi, cervicals four and five are roughly the same size (Vila Nova et al., 2015). In contrast, in Tupandactylus navigans the four cervical is longer than the fifth (Beccari et al., 2021). The same is true for Eopteranodon lii (Lü et al., 2006c) and for Morphotype I of the Sinopterus complex, as seen in the holotypes of S. dongi, S. lingyuanensis, specimen IVPP V 23388, and D3072. In contrast, the fourth cervical is shorter than the fifth in Morphotype II, as seen in the holotype of H. benxiensis (Lü et al., 2007). Our SMA analysis indicates that variation in cervical IV length is uncorrelated to body size in the Sinopterus complex. We thus regard this feature as of potential taxonomic value, distinguishing Morphotypes I and II.

Metacarpal I, length relative to metacarpal IV

An elongate metacarpal I that reaches the carpus is the plesiomorphic condition for azhdarchoids, as seen in Tapejara wellnhoferi, Eopteranodon lii, Tupuxuara leonardii, chaoyangopterids and azhdarchids (e.g., Kellner & Campos, 2007; Wu, Zhou & Andres, 2017). An interesting amount of variation regarding this feature has already been reported before for the Jiufotang tapejarids, with some specimens reportedly bearing either elongate metacarpals I that were subequal to (~90–100% the length of) metacarpal IV (e.g., Wang & Zhou, 2003) or reduced metacarpals I (e.g., Lü et al., 2006a).

Recently, Shen et al. (2021) expressed concern and recommended caution regarding this variation in the Sinopterus complex, since broken/obscured metacarpals could be mistaken for reduced metacarpals in some specimens. In the present work, our close inspection corroborates the presence of reduced metacarpals I (about 30–40% of the length of metacarpal IV) in specimen D2525 and in the holotypes of “H.” benxiensis and “H.” corollatus, along with the new specimens BPMC 103 and BPMC 104. Similarly, elongate metacarpals I (about 90–100% the length of metacarpal IV) are confirmed for the holotypes of S. dongi and H. jii, as well as specimens IVPP V 23388 and D3072, and the new specimen BPMC 106. The condition is unclear in the holotypes of S. lingyuanensis and “H.” atavismus.

Our SMA analysis indicates that this feature is uncorrelated to body size, and our Kruskal-Wallis analysis indicates a significant difference between Morphotypes I and II regarding this feature. This large amount of variation is unreported for pterosaur species and is highly suggestive that more than one species is present in this sample. Thus, we conclude that Morphotype II can be characterized by a short metacarpal I (about 30–40% of metacarpal I length), which does not reach the carpus. Such condition is unique for Morphotype II and Tupandactylus within all known tapejarids.

Wing digit, fourth phalanx length

Similar to what has been found for Rhamphorhynchus (Hone et al., 2021) and anurognathids (Yang et al., 2022), our SMA analyses indicate that most wing elements exhibit isometric growth within the Sinopterus complex. The sole exception to this pattern concerns fourth wing phalanx length, as our SMA analysis shows that its variation is not correlated to body size.

In fact, a noticeable variation occurs in this feature between Morphotypes I and II, irrespective of ontogenetic stage. The length ratio between wing phalanges 4 and 1 is about 0.30–0.40 in Morphotype I, as seen in the holotypes of S. dongi, S. jii, and “H.” atavismus, and specimen D3702. This is similar to Tapejara wellnhoferi and Caiuajara dobruskii, while in Eopteranodon lii the ratio is 0.45. In contrast, this ratio is no higher than 0.20 in Morphotype II, as seen in the holotypes of “H.” corollatus and “H.” benxiensis, as well as specimens D2525, BPMC SC001, BPMC SC002, and BPMC SC003. This is similar to Tupandactylus navigans, for which the same ratio is about 0.13 (Beccari et al., 2021). As indicated by our Kruskal-Wallis analysis, the difference between Morphotypes I and II regarding this feature is statistically significant in our sample. We thus regard that the short wing phalanx 4 (under 20% the length of wing phalanx 1, or under 50% the length of the humerus) of Morphotype II suggests taxonomic distinction from Morphotype I (and all other tapejarids, representing a potential diagnostic apomorphy).

Metatarsal I, relative length

The relative proportions of the metatarsals have already been deemed of taxonomic importance before (Zhang et al., 2019). The general tapejaroid condition is that metatarsal I is shorter than metatarsals II–III, and subequal to metatarsal IV; as found in Tapejara (Eck, Elgin & Frey, 2011), Eopteranodon (Lü et al., 2006c), chaoyangopterids (Wu, Zhou & Andres, 2017), and dsungaripterids (Hone, Jiang & Xu, 2018). This general condition can be seen in Morphotype II specimen D2525, although unclear in the holotypes of “H.” corollatus and “H.” benxiensis. On the other hand, the holotype of Sinopterus dongi and specimens D3702 and BPMC 106 are unique within tapejaroids in exhibiting an elongate metatarsal I (longer than metatarsal II). However, metatarsal I is subequal to metatarsal II (90–95% of its length) in other Morphotype I specimens, as seen in IVPP V 23388 (Zhang et al., 2019), and the holotypes of S. lingyuanensis (Fig. 7) and H. atavismus (Fig. 8), what is not significantly different from Morphotype II. Our SMA analysis indicates that metatarsals I and II grow isometrically relative to humeral length.

It is noticeable that an unusually long metatarsal I is exclusive to a subset of Morphotype I within the whole known tapejaroid sample, but, considering the data as a whole, it is difficult to set this subset of Morphotype I from the remainder of the morphotype, and thus this feature may only represent a polymorphism. Our Kruskal-Wallis analysis reveals no significant difference between Morphotypes I and II regarding this feature. We also note that this feature can be found elsewhere within pterosaurs, such as in Anurognathus ammoni (Bennett, 2007) and Anhanguera piscator (R. Pêgas, 2019, personal observation).

Premaxillary crest, development and size

Within the entire sample of Jiufotang tapejarids here analyzed, only two specimens lack premaxillary crests: the holotypes of Sinopterus gui and S. lingyuanensis. These are the two smallest specimens analyzed here, and are both interpreted as juveniles. The holotype of Nemicolopterus crypticus, which is a near-hatchling, may represent a third specimen of crestless Jiufotang tapejarid (see Witton, 2013; Naish, Witton & Martin-Silverstone, 2021). Thus, it can be said that premaxillary-crestless Jiufotang tapejarids are restricted to very young individuals.

All remaining Jiufotang tapejarid specimens exhibit premaxillary crests, but of differing sizes. Within our Morphotype I (=Sinopterus dongi), the holotype of Sinopterus dongi exhibits but a very discrete crest; that is, it only discretely disrupts the skull margin (by protruding anterodorsally). This specimen is interpreted as a juvenile (see above). Specimen BPMC 107 exhibits a similarly incipient premaxillary crest, despite being regarded as close to skeletal maturity. In contrast, the holotype of ‘H.’ atavismus, which is regarded as a juvenile and is smaller in size than the holotype of S. dongi, bears a more conspicuous premaxillary crest than the latter specimen.

Because premaxillary-crestlessness is restricted to the smallest juvenile specimens, it seems clear that premaxillary crest absence/presence is an ontogenetic feature. Ontogenetic variation in the presence and development of premaxillary crests is corroborated by our SMA analysis, which indicates negative allometry between rostrum index and body size. The negatively allometric growth of rostrum index can easily be explained by the ontogenetic development of the premaxillary crest, which is a feature that augments the value of the rostrum index.

In addition, because premaxillary crest size still varies between larger juveniles and subadults of each species, it also seems likely that variation in premaxillary crest size is influenced also by individual and/or sexual variation, and not only to growth. These variations concerning premaxillary crest presence and size are clearly affected by intraspecific (ontogenetic, individual, and sexual) variations and seem to apply to the Sinopterus complex as a whole. Thus, these variations (concerning crest presence and size) should not be regarded as taxonomic informative for the Sinopterus complex. On the other hand, stating that crest presence/size cannot differentiate between species within our analyzed sample does not imply that a single species exists. Rather, it suggests that sexual and ontogenetic variation (expressed in premaxillary crest size/development) is present in all potentially valid species, whether a single one or more (two as we propose here).

Premaxillary crest, shape

Apart from premaxillary crest presence and size, variation in crest shape can also be seen in Jiufotang tapejarids. Crested specimens exhibit crests of roughly two shapes: heaped and trapezoidal.

Morphotype I is characterized by heaped crests, as seen in the holotypes of S. dongi, ‘H.’ jii, ‘H.’ atavismus, and specimens IVPP 23388 V, D4019, BPMC 106, and BPMC 107. In contrast, Morphotype II is characterized by trapezoidal crests, as seen in the holotypes of ‘H.’ corollatus and ‘H.’ benxiensis, as well as specimens BPMC 103 and BPMC 105.

It is important to highlight that we do not mean to imply that these proposed shape categories are homogenous. Some degree of variation is evidently present within each of them and no two crests are the same, as should be expected given the intraspecific variation in cranial ornamentation that is seen in extant vertebrates, such as in the casques of Numida (Angst et al., 2020) and Casuarius (Naish & Perron, 2016; Green, Kay & Gignac, 2022).

It is notorious that these two shape categories match the two recognized morphotypes/species, apparently without overlap or relation to ontogeny. Consequently, the favored explanation under this scenario is that each shape is characteristic of each morphotype/species.