Chilantaisaurus maortuensis, (Hu, 1964)

Brusatte, Stephen L., Chure, Daniel J., Benson, Roger B. J. & Xu, Xing, 2010, The osteology of Shaochilong maortuensis, a carcharodontosaurid (Dinosauria: Theropoda) from the Late Cretaceous of Asia, Zootaxa 2334, pp. 1-46 : 3-36

Shaochilong maortuensis ( Hu, 1964)

Figs 1–15

Lectotype. IVPP V.2885.1, well preserved and nearly complete braincase, including parts of the parietals, supraoccipital, exoccipital-opisthotics, basioccipital, basisphenoids, parasphenoid, prootics, and orbitosphenoids; IVPP V.2885.2, paired frontals, paired parietals, and posterior end of right nasal.

Taxonomic note and paralectotype series. Hu (1964) erected Chilantaisaurus maortuensis on the basis of cranial bones, an axis and six caudal vertebrae. Although the material was collected from a single locality, Hu (1964) did not explain the degree of association of the bones or provide quarry maps. It is considered likely that the bones represent a single individual, but to provide for the possibility that they will be shown to belong to multiple taxa in future studies we designate the braincase (IVPP V.2885.1) and skull roof fragment (IVPP V.2885.2) as the lectotype (name-bearing type specimen) and consider the remaining material to belong to the paralectotype series: left and right quadrates (IVPP V.2885.3), a right maxilla (IVPP V.2885.4), an axis vertebra (IVPP V.2885.5) and six caudal vertebrae (IVPP V.2885.6–7). The braincase and skull roof piece are both assigned to the lectotype because they clearly fit together as a single specimen (broken along the parietals, which are shared between both pieces). The remaining skull bones and axis probably belong to the same individual as the braincase and skull roof, due to similar size, proximity in the skeleton, non-duplication of elements, and similar phylogenetic affinities indicated by all elements. The caudal vertebrae are referred with less certainty, as they do not show unambiguous evidence for carcharodontosaurid affinities and are from a more distant part of the skeleton.

In an unpublished thesis, Chure (2000) briefly described the lectotype series of Shaochilong and provided a new generic name, “Alshansaurus”. Chure’s (2000) thesis was circulated to many dinosaur researchers and is often cited. Although the name “Alshansaurus” was never formally published, it has been used by many dinosaur workers as an informal name for the specimen.

Type locality and horizon. Ulansuhai Formation, Maortu, Inner Mongolia Autonomous Region, People’s Republic of China (60 km north of Chilantai). The Ulansuhai Formation is often regarded as Aptian-Albian (late Early Cretaceous) based on perceived faunal similarities to other deposits of this age (e.g., Weishampel et al. 2004b). However, radiometric dating of underlying strata indicates a maximum age of approximately 92 Ma (Turonian, early Late Cretaceous [“mid Cretaceous”]; Kobayashi & Lu 2003, Benson & Xu 2008). We prefer the Turonian date, as it is tied to explicit radiometric data.

Original diagnosis. “Skull small, occipital condyle comparatively large, maxilla with 12 teeth, quadrate relatively small” ( Hu 1964: 59).

Emended diagnosis. Allosauroid theropod possessing the following autapomorphies: maxillary antorbital fossa reduced in extent and nearly absent; paradental groove absent on the medial surface of the maxilla; deep, dorsoventrally oriented grooves located dorsally on maxillary interdental plates; pneumatic recess penetrates to posterior end of nasal; dorsoventrally deep sagittal crest on the frontal; large pneumatic foramen (pneumatopore) in the anterodorsal corner of the dorsal tympanic recess of the prootic ( Brusatte et al. 2009a).

Nomenclatural note. Clade names and phylogenetic definitions follow a recent revision of basal tetanuran taxonomy proposed by Benson et al. (2009). Allosauroidea is comprised of four “family”-level clades: Sinraptoridae , Allosauridae , Carcharodontosauridae , and Neovenatoridae . The latter two clades are sister taxa and comprise the rank-free Carcharodontosauria, which is equivalent in definition to Carcharodontosauridae as commonly used by previous authors (e.g., Holtz et al. 2004; Sereno et al. 2005). These taxonomic changes do not severely affect the terminology used in this monograph. However, Neovenator is no longer a “carcharodontosaurid,” as commonly considered under previous taxonomies, but rather a “neovenatorid carcharodontosaurian”.

Description and comparisons. Skull. A skull reconstruction of Shaochilong, drawn by Brett Booth, is presented in Figure 1 . The snout of Shaochilong is shortened relative to other carcharodontosaurids, which generally possess long snouts despite their large body size (e.g., Sereno et al. 1996; Currie & Carpenter 2000; Eddy 2008). In concert with small body size (see below), the short-snouted skull suggests that Shaochilong possessed a unique morphotype, and perhaps ecotype, among carcharodontosaurids.

Maxilla. Hu (1964: 59) listed both a right maxilla and a “fragmental left maxilla” among the “material” (taken here as the syntype series) of Shaochilong maortuensis . However, we could only locate the right maxilla (IVPP V2885.4; Figs 2–3 ), which is nearly complete and well preserved. This element was illustrated by Hu (1964: fig. 10), but it is difficult to distinguish original bone and broken margins in this figure. A revised version of this figure was published by Zhao et al. (2008: fig. 325) and a photograph was provided by Dong (1992). However, other than a paragraph in the original description ( Hu 1964: 60), this bone has not been thoroughly described in the literature.

The right maxilla is nearly complete but is missing the dorsal part of the ascending process as well as the posterior portion of the jugal process bearing the articular facet for the jugal. As preserved the maxilla is 290 mm long anteroposteriorly and 77 mm deep dorsoventrally at the anterior margin of the antorbital fenestra. The tooth row in Shaochilong is complete and the jugal process extends 35 mm posterior to it as preserved. However, a substantial portion is missing in this region, as in other allosauroids there is an extensive margin of non-dentigerous bone posterior to the posteriormost alveolus (e.g., Madsen 1976; Currie & Zhao 1993a; Sereno et al. 1996; Eddy 2008; Sereno & Brusatte 2008). The main body tapers in depth as it continues posteriorly and becomes confluent with the jugal process, thinning to a depth of 38 mm at the posterior broken margin. A tapering main body and jugal process is common among theropods but contrasts with the condition in abelisaurids (e.g., Lamanna et al. 2002; Sampson & Witmer 2007), some coelurosaurs (e.g., Dromaeosaurus: Currie 1995 ), Monolophosaurus (Brusatte et al. in press), and Zupaysaurus ( Ezcurra 2007), which possess maxillae that maintain a relatively constant depth across their length. Posteriorly the jugal process is deflected posteroventrally, beginning at the anterior end of the jugal articulation. Only the base of this deflection is preserved but this region is oriented at an angle of approximately 20 degrees from the anteroposterior trend of the main body. A similar deflection is present in the carcharodontosaurids Acrocanthosaurus ( Eddy 2008: fig. 14) and Eocarcharia (Brusatte & Sereno 2008: figs. 11-13), as well as those megalosaurids in which this region is preserved ( Afrovenator , UC OBA 1; Megalosaurus , OUNHM J.13506), and was employed as a phylogenetic character by Sereno & Brusatte (2008: ch. 8). In contrast, other allosauroids and basal tetanurans only exhibit ventral deflection at the far posterior tip of the jugal process (e.g., Madsen 1976; review in Sereno & Brusatte 2008).

As in many other basal tetanurans there is a distinct anterior ramus of the maxilla that projects from the main body anterior to the ascending ramus (e.g., Madsen 1976; Sereno et al. 1994; Holtz et al. 2004). The separation between the ascending ramus and anterior ramus is slight in Shaochilong and the anterior ramus is tall relative to its length (78 mm deep by 35 mm long). It is proportionally taller than in most other theropods that possess an anterior ramus that is taller than long, such as Ceratosaurus ( Madsen & Welles 2000) , and its shape and size are similar to those of some individuals of Mapusaurus (MCF-PVPH-108.115; Coria & Currie 2006: fig. 2B). In other Mapusaurus specimens the anterior ramus is essentially absent, as it is confluent with the anterior rim of the maxillary body and ascending process (MCF-PVPH-108.169; Coria & Currie 2006: fig 2A). Similarly, the anterior ramus is deep and either confluent with the ascending ramus or weakly demarcated in most other carcharodontosaurids, including Acrocanthosaurus ( Currie & Carpenter 2000) , Carcharodontosaurus ( Sereno et al. 1996; Brusatte & Sereno 2007), and Eocarcharia (Sereno & Brusatte 2008) . However, Neovenator exhibits a prominent anterior ramus (Brusatte et al. 2008). The shape of the ramus is variable in non-carcharodontosaurian allosauroids, as it is prominent in Allosaurus ( Madsen 1976) but confluent with the ascending ramus in Sinraptor ( Currie & Zhao 1993a).

The ascending ramus extends posterodorsally at approximately 45 degrees from the anteroposterior trend of the main body. This is the case in most basal tetanurans, but differs from the nearly vertical orientation of the ramus in most abelisaurids (e.g., Chatterjee 1978; Bonaparte 1985; Bonaparte et al. 1990; Lamanna et al. 2002; Canale et al. 2009). In Shaochilong the ascending ramus is broken dorsally, but by this point it has already strongly tapered (minimum axis “width” measurement in lateral view of 12 mm, compared with 50 mm at its base at the anteroventral corner of the antorbital fenestra). The entire ramus is very thin anteroposteriorly across its length. The overall proportions of the ramus are narrower than those of Carcharodontosaurus (Brusatte & Sereno 2007) , Giganotosaurus (MUCPv-CH-1), and Mapusaurus ( Coria & Currie 2006) , which have relatively narrow ascending rami and narrow antorbital fossae (see below). In contrast, Acrocanthosaurus ( Currie & Carpenter 2000; Eddy 2008), Allosaurus ( Madsen 1976) , Eocarcharia (Sereno & Brusatte 2008) , Neovenator (Brusatte et al. 2008) , and Sinraptor ( Currie & Zhao 1993a) have proportionally wider ascending rami that accommodate a more extensive antorbital fossa.

Only some regions of the surfaces for contact with the premaxilla, nasal, and jugal are observable. The premaxilla is contacted via the anterior surface of the anterior ramus of the maxilla, which is broadly convex in lateral view. Furthermore, when seen in lateral view, the premaxilla-maxilla suture trends strongly posterodorsally. This is also the case in most other allosauroids (e.g., Currie & Zhao 1993a; Brusatte & Sereno 2007), but differs from the more vertical contact in Allosaurus ( Madsen 1976) and Neovenator (Brusatte et al. 2008) .

The nasal articulates with the anterior surface of the ascending ramus and may have continued onto the dorsal surface of the ramus more posteriorly, although this region is broken in the paralectotype maxilla (IVPP V.2885.4). Few details of the nasal suture are evident and it is unclear whether the maxilla contributed to the floor of the external naris. However, it is evident that the nasal articulation is located solely on the anterior surface of the anterior ramus and does not face laterally, unlike in abelisaurids ( Wilson et al. 2003; Sereno et al. 2004; Sereno & Brusatte 2008). Furthermore, the nasal articulation does not terminate ventrally in the blunt pit that is characteristic of abelisaurids ( Wilson et al. 2003; Sereno et al. 2004).

Although most of the jugal articulation is broken, the jugal clearly sat within a deep trough on the posterior part of the jugal process of the maxilla. Whether this trough was partially exposed laterally as in some allosauroids (e.g., Acrocanthosaurus: Eddy 2008 ; Eocarcharia: Sereno & Brusatte 2008 ), due to a lower lateral wall, is unclear. However, it is evident that the most anterior region of the trough is a deep embayment hidden in lateral view, and thus the complete articular surface on the maxilla is not entirely laterally facing as is often considered a synapomorphy of abelisaurids (e.g., Wilson et al. 2003; Sereno et al. 2004).

The lateral surface of the maxilla is generally smooth, although it is slightly rugose anteriorly and above the tooth row. This form of sculpturing is similar to that of most theropods, and is not as extensive as in the derived carcharodontosaurids Carcharodontosaurus , Giganotosaurus , and Mapusaurus (Brusatte & Sereno 2008) and abelisaurids ( Lamanna et al. 2002; Sampson & Witmer 2007; Sereno & Brusatte 2008). In Carcharodontosaurus , elongate grooves and ridges ornament most of the lateral surface, a texturing that has been described as autapomorphic for the genus (Brusatte & Sereno 2007). The surface texture is mottled, with random rugosities that do not form distinct ridges or grooves, in Giganotosaurus ( Coria & Salgado 1995) , Mapusaurus ( Coria & Currie 2006) , and the neovenatorid carcharodontosaurian Neovenator (Brusatte et al. 2008) . In contrast, the lateral surface of the maxilla in Acrocanthosaurus ( Currie & Carpenter 2000) , Allosaurus ( Madsen 1976) , Eocarcharia (Sereno & Brusatte 2008) , and Sinraptor ( Currie & Zhao 1993a) is smooth and little different in texture from that in Shaochilong.

The lateral surface of the maxilla of Shaochilong is pierced by numerous foramina, which are especially abundant immediately dorsal to the tooth row. These foramina form two distinct series: a primary series that is approximately 10 mm dorsal to the tooth row and a secondary series that is positioned 35 mm above the tooth row. Foramina in both rows are large, measuring up to 5 mm in diameter, and form a linear series that approximately parallels the tooth row. The two rows merge posterior to the eighth alveolus, and the final foramen in the conjoined rows (located above the ninth alveolus) opens posteriorly into a deep and elongate groove. A discrete secondary row is also present in Acrocanthosaurus ( Eddy 2008) , Carcharodontosaurus saharicus (Brusatte & Sereno 2007) , Eocarcharia (Sereno & Brusatte 2008) , and possibly Mapusaurus ( Coria & Currie 2006: fig. 2). Foramina are located in this region in other taxa (e.g., Allosaurus: Madsen 1976 ), but are not always set into a discrete row. However, whether this represents random variation or a phylogenetically informative signal is difficult to determine in the small samples for most theropod taxa. Furthermore, the final foramen of the conjoined row also opens into a deep groove in Eocarcharia (Sereno & Brusatte 2008) . Unfortunately, this region of the maxilla is missing in many closely related taxa, precluding comparison. Finally, the primary row of Shaochilong, like those of other allosauroids, is positioned several millimetres above the tooth row, not immediately above the alveolar margin as in abelisaurids (Sereno & Brusatte 2008).

The antorbital fossa is not extensive on the lateral surface of the maxilla, although this appearance is partially exaggerated by breakage. As preserved, the fossa only extends for approximately 7 mm ventral to the antorbital fenestra across most of the main body. However, the dorsal edge of the fossa is a broken surface, which is quite thick in mediolateral width. It is possible to link this broken surface with original bone on the dorsal margin of a small flange that projects dorsally at the anteroventral corner of the antorbital fenestra. This was not a flange in life, but rather is a preserved flake of bone, completely covered by the smooth fossa, that remains in isolation after the rest of the bone in this area has been broken away. Furthermore, the original dorsal surface of this flange can be linked to original bone surface on the posterior margin of the ascending ramus, giving a complete and fairly accurate reconstruction of the true dimensions of the antorbital fossa ( Fig 2 ). In life, the fossa extended only 10-15 mm ventrally from the antorbital fenestra along the main body of the maxilla. Similar ventral reduction is present in other carcharodontosaurids such as Carcharodontosaurus (Brusatte & Sereno 2007) , Giganotosaurus (MUCPv-CH-1), and Mapusaurus ( Coria & Currie 2006) , as well as abelisaurids (e.g., Bonaparte et al. 1990; Sereno & Brusatte 2008) and the megalosaurid Torvosaurus ( Britt 1991) . In contrast, Allosaurus ( Madsen 1976) , Sinraptor ( Currie & Zhao 1993a), and the basal carcharodontosaurians Acrocanthosaurus ( Currie & Carpenter 2000; Eddy 2008), Eocarcharia (Sereno & Brusatte 2008) , and Neovenator (Brusatte et al. 2008) have a ventrally extensive antorbital fossa.

The antorbital fossa extends anteriorly onto the ascending ramus of the maxilla, but only excavates approximately 15% of the width of the base of the ramus ( Table 1 ). In most allosauroids, including basal carcharodontosaurians such as Acrocanthosaurus , Eocarcharia , and Neovenator , this proportion is 50-65%. A more extreme condition, an extensive fossa along the entire ascending ramus, is a synapomorphy of Coelurosauria ( Sereno et al. 1996; Rauhut 2003a; Holtz et al. 2004). In Carcharodontosaurus and other carcharodontosaurines the fossa is reduced on the ascending ramus ( Table 1 ), but not to the extent seen in Shaochilong. Thus, the extremely limited antorbital fossa on the ascending ramus is an autapomorphy of Shaochilong among allosauroids.

The antorbital fossa and the subcutaneous surface of the main body of the maxilla are not separated by a sharp rim or a swollen ridge (as in Carcharodontosaurus saharicus: Sereno et al. 1996 ; Brusatte & Sereno 2007), but rather by an abrupt change in bone texture. Anteriorly, the rim surrounding the antorbital fossa is rounded, not squared-off as in some carcharodontosaurians ( Eocarcharia , Neovenator: Sereno & Brusatte 2008 ), as well as megalosaurids ( Afrovenator : UC OBA 1; Dubreuillosaurus: Allain 2002 ), coelophysids ( Rauhut 2003a) and Eoraptor ( Sereno et al. 1993) . There is a distinct foramen within the fossa, which faces laterally and posteriorly, level with the region between the eighth and ninth alveoli.

No accessory antorbital openings are readily visible within the antorbital fossa. However, as the anteroventral region of the fossa—the location of these openings in other theropods—is broken, this absence is potentially artifactual. Indeed, the broken medial surface of the maxilla shows that the base of the ascending ramus and the promaxillary process were inflated. These two regions are usually inflated by the maxillary and promaxillary fenestrae, respectively ( Witmer 1997). Whether both openings were actually present is difficult to assess, since Carcharodontosaurus has both inflated internal sinuses but only a single external opening ( Sereno et al. 1996; Brusatte & Sereno 2007). This single opening has been interpreted as a maxillary fenestra ( Sereno et al. 1996), but homology with either the promaxillary or maxillary fenestrae of other theropods is difficult to assess (Brusatte & Sereno 2008). Regardless of which fenestra this lone opening is homologous to, both the phylogenetic position of Carcharodontosaurus (nested within Tetanurae, most of which possess two openings) and its internal morphology (two sinuses) indicate that one of the fenestrae was lost, as in other carcharodontosaurines ( Giganotosaurus : MUCPv-CH-1; Mapusaurus: Coria & Currie 2006 ; RBJB pers. obs.). If reduction of the the antorbital fossa correlates with the loss of an accessory pneumatic opening then it is possible that the condition in Shaochilong was the same as that in the carcharodontosaurines. Any fenestrae that were present, however, probably penetrated anteriorly into the base of the ascending ramus and were concealed in lateral view, due to the very narrow lateral exposure of the antorbital fossa.

Accessory excavations within the antorbital fossa of the ascending ramus (‘excavatio pneumatica’ of Witmer 1997) are clearly absent. These structures are present in Acrocanthosaurus ( Eddy 2008) and Eocarcharia (Sereno & Brusatte 2008) , as well as Sinraptor ( Currie & Zhao 1993a), and possibly Allosaurus ( Witmer 1997) , although homology is difficult to assess since some of these structures differ in form, position, and number.

In medial view two separate antorbital sinus chambers are visible at the base of the ascending ramus. It is unclear whether these chambers were closed medially by a wall of bone in life; if so, this wall has broken away to expose the chambers. The more posterior chamber, which corresponds to the maxillary antrum of Witmer (1997), has several concave depressions on its floor that correspond to the tooth crypts. These depressions are referred to as the interalveolar pneumatic recesses by Witmer (1997), and indicate that the tooth replacement crypts extend far dorsally.

The interdental plates are fused into a single lamina, as is the case in all allosauroids more derived than sinraptorids ( Currie & Zhao 1993a), ceratosaurs (sensu Carrano & Sampson 2008), and the megalosaurid Torvosaurus ( Britt 1991) . This lamina is highly ossified: individual plates are only distinguished by shallow depressions between them and no replacement tooth foramina are present.

Chure (1998) described the interdental plates as “very small” and stated that they could “only be differentiated from the maxilla by their texture.” He considered this an “unusual feature” of Shaochilong, and in his 2000 thesis described the size and form of the interdental plates as autapomorphic. In particular, Chure (1998, 2000) considered the interdental plates to be restricted to the ventral margin of the tooth row, since there is an approximately 15 mm deep strip of bone above the tooth row that is especially rugose and punctured by a preponderance of small foramina. Small interdental plates such as these, which are difficult to distinguish from the remainder of the maxilla, have also been described in Dromaeosaurus ( Currie 1995) and were an important featuring linking Shaochilong with derived maniraptorans in Chure’s (2000) discussion of characters.

However, the discovery and description of new comparative material has helped clarify the anatomy of this region. Importantly, the interdental plates are not small but in fact relatively large. As Shaochilong lacks a maxillary paradental groove (groove for the dental lamina) that cleanly demarcates the interdental plates dorsally in other theropods, their size is not immediately apparent. However, although it is true that the plates are heavily textured ventrally, a similar form of surface texturing that differs only in strength extends approximately 40 mm above the tooth row. This form of texturing, composed of random pits and fine lineations, is characteristic of the interdental plates in other carcharodontosaurids (e.g., Brusatte & Sereno 2007) and not the smooth lingual surface of the maxilla dorsal to the paradental groove. Thus, it is reasonable to consider this entire 40-mm-deep region to represent the heavily fused interdental plates.

Two autapomorphies of the interdental plates are present in Shaochilong. First, the paradental groove (groove for the dental lamina) is absent, and the interdental plates and lingual surface of the maxilla are not cleanly separated but contact directly so that their junction is only discernable by a subtle textural change. Second, the medial surfaces of the interdental plates are excavated dorsally by several deep, elongate, dorsoventrally oriented grooves. These are broader and deeper than the numerous fine, cut-like lineations that are present in abelisaurids ( Rauhut 2004b; Sereno & Brusatte 2008), as well as the less sharp, less dense, and more random array of lineations in carcharodontosaurids. Indeed, these “carcharodontosaurid-type” lineations are present ventrally on the interdental plates of Shaochilong, and are present but more widely scattered and less dense dorsally, where they are located alongside and even within the autapomorphic grooves.

Systematically important characters of the interdental plates are also present. The anterior plates are more than twice as deep as wide, a phylogenetically informative character seen in most carcharodontosaurids ( Carcharodontosaurus , Giganotosaurus , Mapusaurus: Brusatte & Sereno 2008 ; Acrocanthosaurus : contra Brusatte & Sereno 2008, Sereno & Brusatte 2008). In contrast, the basal carcharodontosaurian Neovenator (Brusatte et al. 2008) and other allosauroids ( Allosaurus: Madsen 1976 ; Sinraptor: Currie & Zhao 1993 a) have anterior plates that are smaller and shallower, as in other basal tetanurans (e.g., Bonaparte 1986; Sadlier et al. 2008). The basal carcharodontosaurid Eocarcharia has plates that are less than twice as deep as long, but are intermediate in size between those of more derived carcharodontosaurids and other allosauroids (Sereno & Brusatte 2008). Additionally, in Shaochilong, the line of contact between the plates and the lingual surface of the maxilla is approximately straight across most of the bone, but curves anteroventrally at the second alveolus. This is seen in carcharodontosaurians ( Acrocanthosaurus , Carcharodontosaurus , Eocarcharia , Mapusaurus , Neovenator ) and some megalosaurids ( Britt 1991; Benson 2008a), but not in Allosaurus ( Madsen 1976) and Sinraptor ( Currie & Zhao 1993a), in which the contact (formed in these taxa by the paradental groove) is straight across its entire length.

The tooth row is complete, consisting of 12 alveoli, a low number among basal tetanurans. Tooth number for other allosauroids is as follows: Acrocanthosaurus (15), Allosaurus (15), Carcharodontosaurus (~14), Giganotosaurus (12+), Eocarcharia (15), Mapusaurus (12), Neovenator (15+), Sinraptor (15). In Shaochilong the labial wall of the alveoli, formed by the lateral wall of the maxilla, extends further ventrally than the lingual wall, formed by the interdental plates, as in Neovenator (Brusatte et al. 2008) and the megalosaurid Torvosaurus ( Britt 1991) . In ventral view the alveoli are ovoid to subrectangular in shape. The seventh alveolus is the largest and more posterior alveoli become progressively smaller ( Table 2).

2 27 17

3 27 16

4 28 18

5 30 18

6 25 17

7 32 20

8 27 15

9 26 15

10 25 13

11 21 12

12 15 13

Only a single partially erupted tooth, situated in the eighth alveolus, is observable. This tooth is very similar in overall morphology (shape, thickness, and surface texture) to an unerupted tooth described for Eocarcharia (Sereno & Brusatte 2008) , but in the absence of quantitative metrics it is difficult to assess whether such similarity is phylogenetically informative. This tooth is thick labiolingually as in Eocarcharia , not thin and blade-like as in derived carcharodontosaurids ( Sereno et al. 1996). Enamel wrinkles are not visible and if present must have been subtle; the distinct, high-relief, marginal wrinkles of Carcharodontosaurus and other derived carcharodontosaurids are absent (Brusatte et al. 2007). Both mesial and distal carinae are placed slightly labially, and they are continuous across the tip of the tooth as is usual for theropods (contra Harris 1998).

Nasal. Only a small portion of the nasal is present in Shaochilong: a fragment of the posterior end of the right nasal that remains articulated with the nasal prong of the frontal ( Figs 4 , 5 ; IVPP V2885.2). This fragment was not discussed by Hu (1964) although it is clearly visible in his figures ( Hu 1964: pl. 2). It was described by Chure (2000:252), who noted that it has a “weak ornamentation not found on other skull bones of [Shaochilong].” This suggested to Chure (2000) that the fragment might be a displaced element that was later glued onto the holotype frontal. However, Chure (2000) also listed several features consistent with its identification as a nasal, and proceeded to describe the fragment as such. Our observations agree with this assessment: the fragment clearly articulates with the nasal prong of the frontal, and although its dorsal surface does appear to have a weak array of pits not seen on the frontal, this is likely an artefact of erosion.

This nasal piece is fragmentary (38 mm long anteroposteriorly, 32 mm wide mediolaterally). It is clear that the nasal was not fused to its left counterpart, as its medial surface is well-defined. The opposing nasals would have met each other at a straight, smooth, parasagittal suture along their medial surfaces. Posteriorly, the nasal-frontal articulation is expressed as a nearly transverse contact in dorsal view. Separate medial and lateral projections of the nasal are not apparent here, and if present must have been small. A similar condition is seen in Allosaurus ( Madsen 1976) , Carcharodontosaurus ( Sereno et al. 1996; Sereno & Brusatte 2008), and Neovenator (Brusatte et al. 2008) , whereas in Acrocanthosaurus ( Currie & Carpenter 2000) and Sinraptor ( Currie & Zhao 1993a) the nasal-frontal suture is “W” shaped due to an extensive lateral projection of the posterior end of the nasal.

The nasal fragment is extensively pneumatic. The broken anterior and dorsal surfaces of the fragment expose a large pneumatic internal recess that almost completely hollows out the posterior region of the nasal. This recess is divided into two cavities: a posterior pocket and a more anterior region that also extends ventral to the posterior pocket. These pockets are separated by a thick and stout web of bone. The posterior fossa extends much further medially than laterally: here the medial wall of the nasal is only 5 mm thick whereas the lateral wall is 17 mm thick. However, the anterior pocket is wider mediolaterally and the lateral wall of the nasal is only 2 mm thick (the medial wall is not preserved here).

Nasal pneumaticity is a rare feature in theropods and is often considered a synapomorphy of Allosauroidea (e.g., Holtz 2000; Rauhut 2003a; Holtz et al. 2004). Indeed, nasal pneumatic foramina are present in all known allosauroids (other than Acrocanthosaurus : NCSM 14345) and are generally unknown in other basal theropod dinosaurs (e.g., Ceratosaurus , Cryolophosaurus, Zupaysaurus : see review in Brusatte et al. in press). Nasal pneumaticity is present in the basal tetanuran Monolophosaurus ( Currie & Zhao 1993a; Brusatte et al. in press) and the abelisaurid Majungasaurus ( Sampson & Witmer 2007) . However, the nasal morphology of these taxa differs from that of Shaochilong in detail: both have fused nasals and Monolophosaurus possesses an elaborate cranial crest that is mostly formed by the nasals. Furthermore, the posterior region of the nasal recess of Majungasaurus is a single conjoined cavity, shared between the fused nasals, which lacks even a rudimentary midline septum ( Sampson & Witmer 2007: fig. 6). In Shaochilong, in contrast, there were clearly separate pneumatic recesses in each nasal that were separated medially by the medial surfaces of each unfused nasal.

The morphology of the pneumatic recess of Shaochilong is also unique among allosauroids. Other allosauroid taxa have nasal pneumatic recesses anteriorly, but the posterior regions of the nasal are thin, platelike, and apneumatic where they contact the frontal. This condition is suggested by external morphology (e.g., Currie & Zhao 1993a) and verified by high resolution CT scans of Allosaurus ( Snively et al. 2006: fig. 5). The nasals of derived carcharodontosaurids bear pneumatic excavations anteriorly within the nasal portion of the antorbital fossa, and as these are shallow relative to their width in Giganotosaurus (MUCPv-CH 1) and Mapusaurus (MCF-PVPH-108.1; Coria & Currie 2006) they do not result in extensive hollowing of the bone. In Carcharodontosaurus (SGM-Din 1) the nasals are broken posteriorly and do not show internal pneumatic chambers. Therefore, the posteriorly extending pneumatic internal chambers of the nasal in Shaochilong are autapomorphic among allosauroids.

Frontal. The left and right frontals are preserved in articulation (IVPP V2885.2), with fragments of the nasal (see above) and parietal appressed to them ( Figs 4–6 ). The opposing frontals appear to be fused in dorsal view, as each rises up at the midline to contribute to a tall and thin sagittal crest (see below). Chure (2000) considered the frontals to be unfused, noting apparent gaps between the left and right halves of the sagittal crest, but the only well preserved and complete section of the dorsal edge of the crest is sharp and firmly fused. However, the line of fusion between the frontals is visible and partially open in ventral view. This form of coossification is similar to that in a referred specimen of Eocarcharia (Sereno & Brusatte 2008: fig. 16). Frontal fusion in derived carcharodontosaurids (e.g., Carcharodontosaurus: Brusatte & Sereno 2007 ) is more extensive and the suture is almost entirely obliterated in ventral view. However, the degree of fusion clearly changes throughout ontogeny, as shown by the smaller, unfused holotype frontals and larger, fused referred frontals of Eocarcharia (Sereno & Brusatte 2008) . Thus, we hesitate to subdivide characters relating to frontal fusion into separate states, and instead regard all of these carcharodontosaurids (but not the allosauroids Allosaurus and Sinraptor ) as possessing fused frontals.

The frontals of Shaochilong are mediolaterally broad and anteroposteriorly short, and a single frontal is approximately 67% as broad as long. Similar ratios are seen in carcharodontosaurids (Sereno & Brusatte 2008), which possess frontals that are 60–70% as long as broad, as well as in Allosaurus ( Madsen 1976) . Sinraptor has proportionally longer frontals (52%) and abelisaurids often have frontals that are broader than long (ratios over 100%: Sampson & Witmer 2007). Dromaeosaurids are characterized by ratios of approximately 75% (e.g., Barsbold & Osmólska 1999), due to their enlarged postorbital articular processes which projects far laterally, an autapomorphy of the group (Norell & Makovicky 2004). In Shaochilong an individual frontal is 62.5 mm wide mediolaterally at its greatest extent, where it contacts the postorbitals immediately posterior to the nasal prongs. The prongs are discrete processes, one on each frontal, that keep a relatively constant width as they extend anteriorly.

The most remarkable feature of the frontals is a sharp and tall sagittal crest that trends across the entire dorsal surface of the frontal posterior to the nasal prongs. The crest is formed by contributions from both frontals, which appear to be fused along this contact. It is broken in places and some regions have been reconstructed with plaster, but the reconstructed shape appears to be approximately accurate. One exception, however, is that the crest is reconstructed as slightly bifurcating posteriorly, but there is no evidence for this morphology on the specimen itself. The crest is extremely thin: it is only 2 mm in mediolateral width in the one well preserved and non-reconstructed region at its midpoint. Here, it rises approximately 10 mm above the dorsal surface of the frontal. The dorsal edge of the crest appears to trend posterodorsally when seen in lateral view, and thus the crest expands in depth posteriorly. At its posterior end, where it meets a dorsal knob on the parietal, the crest is 18 mm in mediolateral width when viewed posteriorly.

The presence and morphology of the sagittal crest is an autapomorphy of Shaochilong. The dorsal surface of the frontal is flat in all other basal tetanurans (e.g., Madsen 1976; Currie & Zhao 1993a; Allain 2002; Sadleir et al. 2008), including other carcharodontosaurids such as Acrocanthosaurus ( Stovall & Langston 1950; Currie & Carpenter; 2000; Eddy 2008), Carcharodontosaurus ( Sereno et al. 1996; Brusatte & Sereno 2007), Eocarcharia (Sereno & Brusatte 2008) , and Giganotosaurus ( Coria & Currie 2002) . A variety of frontal ornamentation is seen in abelisaurids, including discrete horns and pronounced mound-like eminences ( Bonaparte 1985; see review in Sampson & Witmer 2007), and the frontals and parietals narrow posteriorly to form a sharp crest that gives the frontoparietal bridge a triangular outline in dorsal view ( Sampson & Witmer 2007: fig. 2; reviewed by Carrano & Sampson 2008: ch. 22). A sharp, narrow crest is also present in coelurosaurs (e.g., Weishampel et al. 2004a), formed by constriction of the frontoparietal bridge by the supratemporal fenestrae along its entire length. However, the sagittal crest of Shaochilong is unlike the crests of coelurosaurs and abelisaurids because it is located upon the otherwise flat dorsal surface of the frontoparietal bridge, rather than resulting from narrowing of the bridge itself. Sues et al. (2002) noted that the frontals of Irritator formed a distinct ridge along their median sutural contact but did not figure the structure. It is possible that this represents dorsal swelling of the bones adjacent to the midline suture, but it may also denote a sagittal crest similar to that of Shaochilong. Pending direct examination of the holotype of Irritator (SMNS 58022) we consider the unique morphology of the sagittal crest to be an autapomorphy of Shaochilong.

Lateral to the sagittal crest the dorsal surface of the frontal is smooth. The supratemporal fossa only extends slightly onto the frontal, and at its longest extent is 34% of the anteroposterior length of the frontal itself. Reduced supratemporal fossae have been described as a synapomorphy of Carcharodontosauridae or a subset of derived members of the group (e.g., Coria & Currie 2002: fossa ‘roofed over by a shelf of the frontoparietal’; Brusatte & Sereno 2008). However, a comparative table of measurements has yet to be presented. Among allosauroids, carcharodontosaurids are unique in having a frontal fossa that is less than 35- 40% of the length of the frontal ( Table 3 ). The fossa of Shaochilong covers a larger portion of the frontal than in any other carcharodontosaurid, but is still much smaller than those of Allosaurus and Sinraptor . Sereno & Brusatte (2008: character 32) considered the fossae of Eocarcharia to be “broadly exposed,” as opposed to the “negligible exposure” of Carcharodontosaurus and Giganotosaurus . However, when measured, all of these carcharodontosaurids have similar ratios ( Table 3 ). With that being said, in Carcharodontosaurus , Giganotosaurus , and Shaochilong the parietal is barely exposed on the dorsal surface of the skull roof posterior to the frontal supratemporal fossa. Therefore, the total length of the supratemporal fossa is longer in Acrocanthosaurus and Eocarcharia , as it extends posteriorly onto the parietal. In summary, although the proportional length of the frontal portion of the supratemporal fossa is roughly uniform among carcharodontosaurids, the total length of the supratemporal fossa is reduced in Shaochilong and carcharodontosaurines, in which the parietal no longer participates.

Although proportionally small, the supratemporal fossae of Shaochilong are widely exposed in dorsal view. The opposing fossae are widely separated on the midline by a thick margin of the frontals, as in Acrocanthosaurus ( Stovall & Langston 1950; OMNH 10146), Carcharodontosaurus (Brusatte & Sereno 2007; SGM-Din-1), and Giganotosaurus ( Coria & Currie 2002; MUCPv-Ch 1). In contrast, the fossae of Eocarcharia are more extensive and nearly contact medially, and are thus separated by a narrower midline bridge of the frontals (Sereno & Brusatte 2008). The condition in Eocarcharia is also present in Allosaurus ( Madsen 1976) and Sinraptor ( Currie & Zhao 1993a) and probably represents the plesiomorphic state.

Within the supratemporal fossa of Shaochilong, and essentially bisecting it, is a sinuous crest that trends mediolaterally. This has been noted in Carcharodontosaurus and described as a possible scar for the attachment of jaw adductor musculature, which filled the fossa in theropods (Brusatte & Sereno 2007). This crest, which differs in shape in different species of Carcharodontosaurus (Brusatte & Sereno 2007) , has yet to be described in any other theropod to our knowledge. However, it is also present in Acrocanthosaurus (NCSM 14345) and Giganotosaurus (MUCPv-CH 1), and its absence in Eocarcharia may be due to erosion (Sereno & Brusatte 2008: fig. 14). It is clearly absent in Allosaurus ( Madsen 1976: fig. 11) and Sinraptor (Currie & Zhao 1993: fig. 7), and thus may be a synapomorphy of carcharodontosaurids or a less inclusive subgroup.

The nasal prongs are tongue-like and underlie the nasals ventrally. The articular surface of the prong slopes anteroventrally and is covered with a series of robust grooves that would have strongly interlocked with the nasal, resulting in a firm and immobile contact.

The lateral surface of the frontal is almost completely covered by the extensive articulations for the lacrimal/prefrontal and postorbital. The former articulation is deep and funnel-like, and faces laterally and anteriorly. This contact is 30 mm long anteroposteriorly and 23 mm deep dorsoventrally at its midpoint. It occupies the entire lateral surface of the nasal prong and is deepest at the corner where the prong meets the body of the frontal. Here, the deep, smooth, and rounded internal socket of the funnel faces mostly anteriorly. This socket is obscured in lateral view by a thick lip of bone that trends anteriorly. However, the lip terminates far posterior to the medial edge of the funnel, thus exposing the funnel in lateral view for most of its length. It is unclear if a separate prefrontal articulated here, as the prefrontal and lacrimal are firmly fused into a single element in carcharodontosaurids ( Sereno et al. 1996; Sereno & Brusatte 2008). However, if present, the prefrontal did not articulate with the frontal across a rugose and interdigitating suture like that seen in Eocarcharia (Sereno & Brusatte 2008) , Allosaurus ( Madsen 1976) , and Sinraptor ( Currie & Zhao 1993a).

The postorbital articulation is notably large in lateral view. It trends posteroventrally-anterodorsally, and has a 47 mm long axis and 22 mm perpendicular minor axis at its greatest extent. Posterior to this articulation there is a small notch on the frontal for the laterosphenoid. The anterior part of the postorbital articulation forms a small but discrete process that faces anteriorly, not laterally. This process is present in other carcharodontosaurids, and is well figured in Eocarcharia (Sereno & Brusatte 2008: figs. 14, 15), but is absent in Allosaurus ( Madsen 1976) and Sinraptor ( Currie & Zhao 1993a). In carcharodontosaurids this region smoothly lines up with an anteriorly facing articular surface on the postorbital, and together they contact the lacrimal to exclude the frontal from the orbital rim. Exclusion of the frontal from the orbital rim is a synapomorphy of carcharodontosaurids (Brusatte & Sereno 2008), and is clearly present in Shaochilong. Not only is the small anterior process of the postorbital articulation present, but the region between the lacrimal and postorbital contacts, which corresponds to the orbital rim in other theropods, is essentially absent. It is reduced to a tiny, 4 mm long notch that faces mostly anteriorly, not laterally as does the orbital rim of most theropods. Furthermore, this notch is not smooth, as is characteristic of the orbital rim, but houses a discrete rugose tuberosity. This narrow margin would not have contributed to the rim of the orbit in Shaochilong.

The ventral surface of the conjoined frontals is marked by two large, crescentric scars for the orbitosphenoid ( Figs 4 , 6 ). In between them, and extending parasagitally along the midline of the frontals, is a groove for the olfactory tract. This groove, which forms the endocranial surface of the frontal, terminates anteriorly in two small (approximately 25 mm long by 10 mm wide), teadrop-shaped depressions for the olfactory bulbs. These begin at approximately midlength of the frontal and extend anteriorly nearly to the point where the nasal prongs diverge from the body of the frontal. The orbitosphenoid scars terminate near the midpoint of the olfactory bulb depressions and clearly do not enclose the olfactory bulbs anteriorly. Thus, the sphenethmoid was not ossified. This condition is also seen in Eocarcharia , Allosaurus , Sinraptor , and most theropods (Sereno & Brusatte 2008), but an ossified sphenethmoid is present in Acrocanthosaurus , Carcharodontosaurus , and Giganotosaurus (e.g., Stovall & Langston 1950; Coria & Currie 2002); this character is further discussed and reviewed below.

The proportions and shape of the endocranial surface are similar to those of other allosauroids. Importantly, the surface is not extremely narrow as is autapomorphic for Eocarcharia (Sereno & Brusatte 2008) . However, the endocranial surface is much broader in many coelurosaurs (e.g., Currie 1985: fig. 3), especially posteriorly, and only narrows as it terminates at the olfactory bulbs. In Shaochilong and other allosauroids the endocranial surface is narrow across its entire length and actually expands at the olfactory bulbs. Additionally, the endocranial surface and olfactory bulb depressions are shallow in Shaochilong, which is characteristic for basal theropods but contrasts with the deeper and more heavily vascularized depressions in most coelurosaurs (e.g., Currie 1985; Osmólska 2004; Kirkland et al. 2005).

Parietal. The parietals are nearly complete but are broken in half, with the anterior regions fused to the frontals (IVPP V2885.2) and the posterior regions conjoined with the braincase (IVPP V2885.1) ( Figs 4 , 5 ). These two regions match up, providing irrefutable evidence that the frontal/nasal piece and the braincase belong to the same individual. The opposing parietals are fused on the midline, where they are 22 mm long anteroposteriorly.

The conjoined parietals are hourglass shaped in dorsal view, due to the medially extensive supratemporal fenestrae. The fenestrae are only separated by a 20 mm width of parietal at their greatest expansion. In comparison, the posterior edge of the parietal is 123 mm wide, meaning that this bone is constricted to only 16% of its maximum width between the supratemporal fenestrae. This constriction is proportionally greater than in other carcharodontosaurids ( Acrocanthosaurus: Stovall & Langston 1950 ; Carcharodontosaurus: Sereno et al. 1996 ; Sereno & Brusatte 2008; Giganotosaurus: Coria & Currie 2002 ). Interestingly, in these carcharodontosaurids the narrow extent of the supratemporal fossae on the frontal (in both anteroposterior and mediolateral dimensions) corresponds with a small degree of parietal constriction. However, in Shaochilong the frontal fossae are not extensive but the parietal is still strongly hourglass shaped.

Unfortunately, the parietal is eroded dorsally, and thus it is unclear whether Shaochilong possessed the tall dorsal parietal eminence that is seen in some carcharodontosaurids ( Coria & Currie 2002). Similarly, it is unclear whether the supraoccipitals or parietals overlapped each other dorsally. However, based on the thickened anterior margin of the parietals where they meet the frontals, it appears as if the frontal sagittal crest did continue onto at least the anterior region of the parietals. The frontals and parietals are heavily fused where they contact, a condition seen in all carcharodontosaurids except for Eocarcharia (Brusatte & Sereno 2008; Sereno & Brusatte 2008). In posterior view the parietal is exposed broadly on the occiput and rises to the same level as the supraoccipital. A foramen for the dorsal head vein pierces the occipital plate of each parietal where it meets the exoccipital-opisthotic and presumably the supraoccipital, although sutures in this region are not entirely clear.

Quadrate. Both left and right quadrates are known (IVPP V2885.3) ( Fig. 7 ). The right element is complete and well preserved, whereas the left is broken into several pieces. The right quadrate is 143 mm tall dorsoventrally along its posterior margin (the “shaft” region). Anteriorly the shaft gives rise to a plate-like flange that articulates with the pterygoid. This flange is 47 mm long anteroposteriorly at its midpoint and 104 mm deep at its tallest extent posteriorly.

The lateral surface of the quadrate is marked by an elongate, laterally-facing, rugose articular scar for the quadratojugal. This scar extends along the dorsal half of the shaft and expands in anteroposterior length dorsally before eventually reaching the quadrate cotylus. The cotylus, or head, is a smoothly rounded ovoid structure, which is 24 mm long anteroposteriorly by 18 mm wide mediolaterally in proximal view. Further ventrally, the lateral surface of the lateral condyle is entirely excavated by a rugose articulation for the quadratojugal. This sutural surface is roughly triangular, 24 mm tall by 20 mm long anteroposteriorly, and faces laterally and dorsally. Thus, the quadratojugal articulates with both the lateral surface of the shaft dorsally and the lateral condyle ventrally, as is usual for basal theropods.

Between the two articular surfaces for the quadratojugal is a smooth, 34 mm deep nonarticular margin. Approximately 18 mm of this margin is indented as a slight concavity, which is the medial edge of the quadrate foramen. Chure (2000) could not locate the quadrates during the course of his study, but suggested that the quadrate foramen was absent based on an interpretation of Hu’s (1964) published figures. However, the smooth, concave margin for the foramen is visible between the two articular surfaces for the quadratojugal in Hu’s (1964: fig. 9) illustration. This margin is subtle and suggests that the quadrate foramen was a small structure in life. It must have been extensively enclosed by the quadratojugal, which would have formed its lateral, dorsal, and ventral margins. Small foramina are also present in other allosauroids (e.g., Madsen 1976; Currie & Zhao 1993a; Coria & Currie 2006; Eddy 2008), and the carcharodontosaurids Acrocanthosaurus ( Eddy 2008) , Giganotosaurus (MUCPv-CH-1), and Mapusaurus ( Coria & Currie 2006) also possess foramina that are broadly enclosed by the quadratojugal. In other allosauroids the foramen is primarily enclosed by the quadrate (e.g., Sinraptor: Currie & Zhao 1993 a). Allosaurus is often considered to possess a foramen fully or almost entirely enclosed by the quadrate ( Madsen 1976), but this condition is variable among specimens (UMNH VP specimens, RBJB pers. obs.). In posterior view a shallow groove leads into the quadrate foramen, as is characteristic for theropods (e.g., Brusatte et al. in press).

The quadrate flange is thin and plate-like. It projects anteriorly and medially relative to the transversely straight condyles and its lateral surface is smooth and flat. In contrast, the medial surface of the flange is flat dorsally but deeply concave ventrally, where there is a smooth pocket that excavates the corner where the flange meets the medial condyle. However, this pocket does not enclose any pneumatopores or other external signs of pneumaticity. Indeed, the quadrate appears to be apneumatic, similar to the condition in Allosaurus ( Madsen 1976) and Sinraptor ( Currie & Zhao 1993a). In contrast, all known carcharodontosaurid quadrates are pneumatized (e.g., Acrocanthosaurus , Giganotosaurus , Mapusaurus ), as are those of tyrannosaurids (e.g., Brochu 2003). Such pneumaticity is usually expressed in two regions of the quadrate in carcharodontosaurids. First, some specimens possess a discrete pneumatopore, which sometimes is fenestra-like, on the posterior surface of the quadrate (e.g., Coria & Currie 2006: fig. 7; Eddy 2008). This structure is also seen in the neovenatorid carcharodontosaurian Aerosteon ( Alcober et al. 1998; Sereno et al. 2008). Second, many specimens possess a deep pneumatopore, which leads into an internal chamber, at the corner of the medial surface where the flange meets the medial condyle (e.g., MUCPv-CH-1; Coria & Currie 2006; Eddy 2008). The presence of a smooth pocket in this region in Shaochilong suggests that a precursor of pneumaticity may be present. However, many other theropods also possess a smooth fossa on the medial surface of the quadrate flange, which often extends ventrally into the region of the pocket in Shaochilong. This is usually described as a shallow pneumatic feature, associated with the paratympanic system, which does not penetrate the quadrate internally (e.g., Currie 2003a).

Separate lateral and medial condyles are present ventrally. Chure (2000) interpreted Hu’s (1964) figures as illustrating a single, undivided distal condyle, which was an important feature linking Shaochilong with the enigmatic Labocania ( Molnar 1974) in Chure’s (2000) discussion of characters. However, although the groove separating them is subtle, separate condyles are clearly present. The lateral condyle is 37 mm wide mediolaterally by 15 mm long anteroposteriorly. Its ventral articular surface is highly convex anteriorly and concave posteriorly, and in distal view it is seen to continue laterally and posteriorly as a thin flange. This flange develops into the laterally-facing articulation for the quadratojugal, and defines its ventral margin. The medial condyle has a long axis (40 mm) oriented slightly anterolaterally-posteromedially, with a 24 mm perpendicular minor axis, and its ventral articular surface is less convex than the lateral condyle. In fact, the convex region of the lateral condyle continues onto the anterior margin of the articular surface of the medial condyle. This upraised margin, which thins and sharpens as it continues medially, defines the anterior edge of the trochlear surface for the jaw articulation. The posterior edge of the trochlea is demarcated by a slighter upraised bulge along the posterior margin of the medial condyle.

Braincase. The braincase (IVPP V2885.1) is well preserved and substantially complete, making it one of the best known basal tetanuran braincases ( Figs 5 , 8–12 ). However, only the bases of both paroccipital processes are currently represented; Hu (1964) figured much of the left paroccipital process, but this piece could not be located during the course of our study. Other missing regions include the right basal tuber, most of the funnel-shaped basisphenoid recess ventrally, and the anterior regions of the laterosphenoids and orbitosphenoids. Most of the sutures between individual bones have been obliterated, and thus the shape and extent of some bones are reconstructed based on landmarks and raised ridges that we consider the fused remnants of original sutures. This degree of fusion suggests that the individual was an adult at its time of death, an assessment supported by the heavily fused interdental plate apron on the maxilla and the fused neurocentral sutures of the axis and most caudal vertebrae.

In general, the braincase is very similar to those of Acrocanthosaurus ( Stovall & Langston 1950; Franzosa & Rowe 2005; Eddy 2008), Carcharodontosaurus ( Sereno et al. 1996; Brusatte & Sereno 2007), and Giganotosaurus ( Coria & Currie 2002) . As in these carchardontosaurids, the braincase of Shaochilong is extremely pneumatic: it is penetrated by numerous pneumatopores and excavated by deep pneumatic fossae, and broken regions show the presence of several internal chambers. Additionally, as in carcharodontosaurids, the braincase of Shaochilong as preserved is short anteroposteriorly and extremely deep dorsoventrally. However, the latter dimension is underestimated since much of the basisphenoid funnel is missing, and thus it would have been even deeper in life. In contrast, proportionally longer braincases are seen in other allosauroids (e.g., Allosaurus: Madsen 1976 ; Sinraptor: Currie & Zhao 1993 a), as well as basal tetanurans (e.g., Piatnitzkysaurus: Rauhut 2004 a) and basal theropods (e.g., Cryolophosaurus: Smith et al. 2007 ; Dilophosaurus: Welles 1984 ; Majungasaurus: Sampson & Witmer 2007 ; Zupaysaurus: Ezcurra 2007) in general. The braincases of derived tyrannosaurids (e.g., Tyrannosaurus: Brochu 2003 ) and spinosaurids (e.g., Baryonyx: Charig & Milner, 1997 ; Irritator: Sues et al. 2002 ) are also short and deep, but basal members of each clade (spinosauroids: Dubreuillosaurus, Allain 2002 ; tyrannosauroids: Dilong , IVPP V14243 ; Guanlong , IVPP V14531 ) have proportionally longer braincases similar to those of most other theropods.

Supraoccipital. The supraoccipital is complete and well preserved. It is widely exposed on the occiput and is inclined posteroventrally. Visible sutures clearly show that this bone contributes to the dorsal rim of the foramen magnum. However, it does not extend ventrally to form the lateral margins of the foramen magnum and contribute to the dorsal surface of the occipital condyle, as has been described in Giganotosaurus ( Coria & Currie 2002) . The posterior surface of the supraoccipital is ornamented by a robust midline crest, which thickens in mediolateral dimension as it expands dorsally. The crest becomes confluent with a large dorsal expansion of the supraoccipital, the “pronounced nuchal process” described by Coria & Currie (2002). This process, which is often referred to as the supraoccipital “knob” or “tuberosity” ( Sampson & Witmer 2007), is extremely rugose, thickened mediolaterally and anteroposteriorly when viewed dorsally, and is greater than twice the width of the foramen magnum in the derived carchardontosaurids Acrocanthosaurus ( Stovall & Langston 1950) , Carcharodontosaurus ( Sereno et al. 1996) , and Giganotosaurus ( Coria & Currie 2002) . A less pronounced and mediolaterally narrower structure is seen in Allosaurus ( Madsen 1976) and Sinraptor ( Currie & Zhao 1993a) and more generally among theropods (e.g., Taquet & Welles 1977; Sampson & Witmer 2007). In Shaochilong the knob is especially prominent: it is more than three times the width of the foramen magnum, 1.3 times the width of the ventral region of the supraoccipital where it roofs the endocranial cavity, and is slightly wider than the occipital condyle. The knob joins with the occipital plate of the parietal to form a tall nuchal crest, which is striking in posterior view. This crest comprises the posterior edge of the supratemporal fenestrae and thus delimits the chamber for the temporal musculature posteriorly.

Basioccipital. The basioccipital forms the majority of the occipital condyle and basal tubera. The occipital condyle is subspherical and projects posteroventrally when the frontals are held horizontally. The basioccipital clearly forms the floor of the foramen magnum above the condyle, as the pedicels of the exoccipital-opisthotic only form the dorsolateral corner of the condyle and do not join across the midline. This latter condition, in which the basioccipital is completely separated from the foramen magnum, has been suggested for Carcharodontosaurus , but based upon unclear sutures (MNN IGU3; Brusatte & Sereno 2007).

A stout ‘neck’ of bone supports the occipital condyle and connects the condyle with the exoccipitalopisthotic and remainder of the basioccipital anteriorly and laterally. Ventrolateral to the occipital condyle, the posterior surface of the basioccipital dorsal to the basal tubera is excavated on both sides by deep pneumatic fossae. It is unclear whether the right fossa penetrates the surface due to breakage in this region, but on the left side there is a large ovoid pneumatopore that leads anteriorly and medially into an extensive recess below the endocranial cavity. Similar ‘paracondylar pneumatopores’ have been described in carcharodontosaurids (e.g., Carcharodontosaurus: Brusatte & Sereno 2007 ; Giganotosaurus: Coria & Currie 2002 ), and are also present in Acrocanthosaurus (OMNH 10146). The presence of these structures has been used as a phylogenetic character (e.g., Coria & Currie 2002; Brusatte & Sereno 2008). Pneumaticity is common in this region in coelurosaurs (e.g., Makovicky & Norell 1998; Currie 2003a; Kirkland et al. 2005) and is also present in other basal theropod taxa. Rauhut (2004a) identified small pneumatopores in Piatnitzkysaurus that he suggested were associated with the subcondylar recess, whereas Sampson & Witmer (2007) described tiny pneumatic foramina in Majungasaurus that lead into an extensive medial cavity underneath the brain, which is confluent with the anterior tympanic recess. However, in neither of these taxa, nor in any other non-avian theropods, are there large pneumatopores entering the posterior surface of the basioccipital ventromedial to the occipital condyle as in carcharodontosaurids. Additionally, the form of the surrounding fossa is different: a distinct fossa is not present in Majungasaurus , whereas a shallow fossa that faces directly posteriorly (instead of a deep fossa that faces posterolaterally as in Shaochilong) is seen in Piatnitzkysaurus . Unfortunately, it is unclear if the median recess underneath the endocranial cavity is associated with the subcondylar or anterior tympanic recesses in Shaochilong because of extensive internal breakage.

The basal tubera project posteroventrally relative to the horizontal dorsal surface of the frontals. Combined with the posteroventral sloping of other occipital structures, such as the posterior surface of the supraoccipital, this gives the entire posterior surface of the braincase a posteroventral inclination, as in Carcharodontosaurus , Giganotosaurus , and Sinraptor ( Coria & Currie 2002; Brusatte & Sereno 2008). In contrast, the tubera of Allosaurus and Acrocanthosaurus descend nearly vertically, and thus are perpendicular to a horizontal plane drawn through the occipital condyle ( Coria & Currie 2002). In Shaochilong the tubera are wider transversely than the occipital condyle as in Carcharodontosaurus , Giganotosaurus , and indeed most theropods, not narrower as in Acrocanthosaurus , Allosaurus , and Sinraptor (Brusatte & Sereno 2008) . The posterior surface of the basioccipitals between the tubera is excavated by a single, deep midline fossa, which continues dorsally underneath the occipital condyle. This depression is an expression of the subcondylar recess, a common structure in theropods ( Rauhut 2004a; Sampson & Witmer 2007). A single midline fossa is present in most theropods, but varies in width. In Allosaurus ( Madsen 1976) , Carcharodontosaurus (Brusatte & Sereno 2007) , Shaochilong ( Figs 8 , 9 ), and Sinraptor ( Currie & Zhao 1993a) the fossa is approximately half the width of the occipital condyle and widely separates the pneumatic fossae on the posterior surfaces of the basal tubera. However, in Acrocanthosaurus the midline fossa is narrow, forming a deep dorsoventrally oriented groove between the fossae on the basal tubera (NCSM 4345, OMNH 10146). This condition may be related to narrowing of the basal tubera in Acrocanthosaurus and seems to be autapomorphic. The condition in Giganotosaurus cannot be determined due to incomplete preservation (MUCPv-CH 1). Much of the fossa appears to be open posteriorly in Shaochilong and is seen to lead into the large recess underneath the endocranial cavity. However, this opening is not a pneumatopore but simply is broken bone, and thus it is unclear if the external subcondylar fossa communicated with the internal median recess (see above). Only the left basal tuber is complete distally, where it is slightly thickened and rugose for muscle attachment. There clearly was an arched, concave ventral margin between the left and right tubera, which in posterior view appears as a deep notch.

The basal tubera are formed mostly by the basioccipital. The basisphenoid forms the anterior portion of the tuber, as shown by a visible suture at the anterolateral corner of the left tuber. However, this suture is only visible in lateral and anterior views, and the basisphenoid contribution is only seen as a slight ventral projection in posterior view. Only a small region of the basisphenoid contribution is clearly preserved, at the anterolateral margin of the left tuber. Anterior and dorsal to this region the suture between the basisphenoid and basioccipital has been obliterated by fusion. The clear basisphenoid contribution corresponds to the “basisphenoid scar” of Bakker et al. (1988), a muscle attachment site. This scar is the only part of the crista ventrolateralis—the web of bone that spans the tuber and basipterygoid process to form the lateral wall of the basisphenoid—that is preserved in Shaochilong.

It is unclear if the exoccipital-opisthotic contributes to the tuber, as the suture between this element and the basioccipital has been obliterated by fusion. In many theropods, including many basal tetanurans, the lateral surfaces of the tubera are formed by descending processes of the exoccipital-opisthotic ( Rauhut 2004a). If this is the case in Shaochilong, then the two bones are smoothly confluent and not separated by a notch ventrally as in Allosaurus ( Madsen 1976) and Acrocanthosaurus ( Stovall & Langston 1950; Eddy 2008). However, other basal theropods (e.g., Majungasaurus: Sampson & Witmer 2007 ) do not exhibit a descending process of the exoccipital-opisthotic, and the entire posterior surface of the tubera is formed by the basioccipital.

Basisphenoid. A large part of the basisphenoid is present, but it is difficult to trace its sutures with other braincase bones (with the exception of the small region of visible suture at the anterolateral corner of the basal tuber, described above). A large part of the basisphenoid contribution to the lateral wall of the braincase is present, but much of the bone is eroded ventrally in the region of the funnel-like basisphenoid recess. In ventral view the basisphenoid is sheared across a planar surface, exposing a cross sectional view of the pneumatic basisphenoid recess, as well as portions of the anterior tympanic recess and the median recess underneath the endocranial cavity, which appears to be partially enclosed by the basisphenoid. Anteriorly, the basipterygoid processes and most of the crista ventrolateralis linking these processes to the basal tubera are absent. Rauhut (2004a) stated that a basipterygoid recess—a pneumatic depression on the lateral wall of the crista above the basipterygoid process—is present in Shaochilong. However, this region of the braincase is not preserved.

The sheared ventral surface of the basisphenoid exposes a number of pneumatic cavities in cross section. A deep, triangular, funnel-like cavity is clearly part of the basisphenoid recess, an enigmatic midline excavation that is hypothesized to be part of the median pharyngeal system ( Witmer 1997; Sampson & Witmer 2007). Because the crista ventrolateralis is mostly missing it is not clear if the basisphenoid recess of Shaochilong was a deep funnel that occupied approximately 30% of the depth of the braincase as in Acrocanthosaurus and Carcharodontosaurus ( Sereno et al. 1996; Brusatte & Sereno 2007; Sereno & Brusatte 2008), a derived condition relative to the smaller and shallower recesses of most theropods. However, the broken walls of the crista ventrolateralis are thick and diverge from each other in a wide, funnel-like shape, suggesting that the recess was extensive, and larger than in most theropods. However, the basal tubera are not nearly as ventrally extensive as they are in Acrocanthosaurus ( Stovall & Langston 1950; Eddy 2008), where they are extremely deep to form the posterior wall of the deep basisphenoid recess. Thus, we suggest that that the basisphenoid recess was intermediate in size between the shallower recesses in theropods such as Allosaurus and Sinraptor and the deep funnels of some carcharodontosaurids.

Other cavities visible in ventral view correspond to other recesses. Immediately anterior to the basisphenoid recess is a large, triangular opening that is either an anterior chamber of the basisphenoid recess or the subsellar recess ( Rauhut 2004a; Sampson & Witmer 2007). Posterior and lateral to the basisphenoid recess, and well exposed on the better preserved left side, are elongate cavities associated with the median recess underneath the endocranial cavity. Finally, lateral to the basisphenoid recess, and well preserved on the right side, is a smaller funnel-shaped depression that leads into the foramen for the internal carotid. This is the anterior tympanic recess ( Rauhut 2004a; Sampson & Witmer 2007), and it is better seen in lateral view where it deeply excavates the lateral wall of the braincase in the region where the prootic and basisphenoid meet. The recess is partially hidden in lateral view by the preotic pendant, which extends posteroventrally as a winglike structure. Sutures between the prootic and basisphenoid in this area are unclear, but a raised ridge trending anterodorsally across the lateral surface of the pendant may represent this contact. If so, the pendant, as well as the anterior tympanic recess medial to it, is nearly evenly divided between these bones. Unfortunately, after entering the anterior tympanic recess the course of the carotid is not clear. It is not possible to determine whether the paired carotid canals united internally, an unusual feature among theropods ( Sampson & Witmer 2007) that has been suggested for Giganotosaurus ( Coria & Currie 2002) .

Whether the various recesses (basisphenoid recess, anterior tympanic recess, medial recess underneath the endocranial cavity) communicated with each other internally is unclear. As preserved, the basisphenoid recess does communicate with the remaining two recesses, but only because of clearly broken surfaces. Therefore, since much of the ventral part of the braincase is missing and the preserved walls are heavily eroded, it is unknown whether the walls of bone between these cavities would have completely separated them in life. This information is probably intractable even with CT data. However, all three recesses are positioned close to one another, and are densely packed within the braincase and only separated by narrow walls of bone. Thus, it is possible that they did communicate in life, since it would only require small foramina between the dividing walls and not complex internal passageways.

Exoccipital-Opisthotic. The exoccipital and opisthotic are indistinguishably fused into a single element as in archosaurs generally (e.g., Currie 1997; Sampson & Witmer 2007). The left and right exoccipitals are separated by the supraoccipital and basioccipital and never come into contact. The exoccipital-opisthotics form the lateral margins of the foramen magnum and are flat in this region, lacking the depressions seen in many coelurosaurs (e.g., Currie & Zhao 1993b; Currie 1995; Makovicky & Norell 1998; Norell et al. 2004). Further ventrally, stout pedicels of the exoccipital-opisthotic contribute to the dorsolateral corners of the occipital condyle, and sutures between the exoccipital-opisthotic and basioccipital are clearly visible on both sides of the condyle.

Ventral to the pedicels, and lateral to the occipital condyle, two large foramina open posterolaterally into a depression, which appears to be mostly present on the exoccipital-opisthotic. This depression is sometimes called the paracondylar pocket ( Welles 1984) or the paracondylar recess ( Chure 2000); we prefer the former term, since “recess” implies that this region is pneumatic, which is clearly not the case in Shaochilong and other large theropod dinosaurs (e.g., Currie & Zhao 1993a; Rauhut 2004a; Brusatte & Sereno 2007; Sampson & Witmer 2007). The two foramina are of equal size, with one placed posterodorsal to the other. A much smaller, ovoid depression between them may represent a third opening that has been filled with matrix. Indeed, three openings in this region—two for the hypoglossal (XII) nerve and one jugular foramen that transmits the jugal vein, vagus (X), and accessory (XI) nerves—is the usual condition in derived theropods. In these taxa, the jugular foramen (= “metotic foramen”) is divided from the remainder of the middle ear by an ossified metotic strut, which serves to reposition the jugular foramen on the posterior occipital surface of the braincase, in contrast to its lateral position in more basal archosaurs (e.g., Gower & Weber 1998; Sampson & Witmer 2007). Even if the small depression of Shaochilong is not a true third foramen, the presence of only two clear foramina in the paracondylar pocket is not an argument for the primitive condition. First, there is no clear jugular foramen on the lateral wall of the braincase. Second, Majungasaurus clearly has a posteriorlyreoriented jugular foramen (as shown by CT data) but only two posterior openings ( Sampson & Witmer 2007). Third, several other basal theropod specimens that lack obvious lateral jugular foramina, but have not been subject to rigorous CT study, also only have two openings in the paracondylar pocket (e.g., Allosaurus : UMNH VP 16605; Baryonyx: Charig & Milner 1997 ; Ceratosaurus : BYU 128930; Giganotosaurus : MUCPv-CH 1; Irritator: Sues et al. 2002 ). Thus, like Majungasaurus and presumably these other basal theropods, Shaochilong may have only a single, larger opening for both branches of the hypoglossal nerve within the paracondylar pocket. Whether the number of hypoglossal foramina is phylogenetically informative or randomly variable awaits further study, although it was employed as a phylogenetic character by Benson (in press).

Anteriorly and laterally to the paracondylar pocket, a bony web separates the jugular foramen from the fenestra ovalis. Although this web is usually referred to as the metotic strut (see review in Sampson & Witmer 2007), we prefer the term crista tuberalis, since the web in extinct reptiles cannot be positively associated with the embryonic metotic cartilage that forms the strut in extant taxa ( Gower & Weber 1998; Sampson & Witmer 2007). The crista tuberalis is a thick and extensive web that connects the paroccipital process dorsally with the basal tubera ventrally, and in doing so separates the posterior and lateral walls of the braincase. The crista is formed completely by the exoccipital-opisthotic with no contribution from the prootic; the latter condition, which is abnormal for basal theropods, has been described in Carcharodontosaurus , but based on equivocal broken bone surfaces (Brusatte & Sereno 2007).

Only the bases of each paroccipital process are preserved on the specimen, although Hu (1964: pls 1, 2) presented photographs showing a nearly complete left paroccipital process. We will use these photographs to augment our description of the paroccipital processes, even though the extensive left paroccipital process (which appears to be preserved as a separate piece based on the photographs) could not be located during the course of our studies, neither by DJC in the 1990s nor SLB in 2009.

The paroccipital processes extend strongly laterally, ventrally, and posteriorly. Their posteriormost extent is unknown and Hu’s (1964) images cannot be of help here, since he did not figure the braincase in dorsal view. However, based on the angle and orientation of the broken bases, the processes would have extended far posteriorly as in carcharodontosaurids ( Coria & Currie 2002), Allosaurus ( Madsen 1976) , and Sinraptor ( Currie & Zhao 1993a). The preserved trend of the processes and Hu’s (1964) photographs indicate that these structures were approximately planar and did not exhibit the distal twisting of some coelurosaurs ( Currie 1995; Norell et al. 2004). Hu’s (1964) photo of the braincase in posterior view shows that the paroccipital processes were oriented strongly ventrally, such that their tips terminated below the occipital condyle. This is seen in all allosauroids, as well as a few non-allosauroid basal theropods (e.g., Ceratosaurus: Madsen & Welles 2000 ; Cryolophosaurus: Smith et al. 2007 ; see review in Brusatte et al. in press). However, Shaochilong lacks one feature that has been described as an allosauroid synapomorphy: ventral margins of the bases of the paraoccipital processes positioned ventral to the occipital condyle (see review in Brusatte et al. in press). This condition is seen in Acrocanthosaurus , Allosaurus , and Sinraptor , whereas other basal theropods have more dorsally positioned paroccipital processes in which the ventral margins of the bases are level with the midpoint of the condyle (Brusatte et al. in press). Shaochilong clearly possesses the latter condition. Brusatte et al. (in press) regarded Carcharodontosaurus as possessing the allosauroid condition, but this was based on a reconstruction of the paroccipital processes on two skulls that are broken in this region (SGM-Din- 1; MNN IGU3). Similarly, Giganotosaurus is also broken in this region ( Coria & Currie 2002). Thus, it may be that some carcharodontosaurids have more dorsally positioned paraoccipital processes, unlike the condition in more basal allosauroids in which the paroccipital processes are web-like and extensive in posterior view due to their ventrally-placed bases.

Although the braincase is extensively pneumatized in general, the broken bases of the paroccipital processes exhibit spongy bone texture in cross section, not the large pneumatic cavities of some coelurosaurs (e.g., Kurzanov 1976; Clark et al. 1994; Sues 1997; Makovicky & Norell 1998; Brochu 2003; Norell et al. 2004). Pneumaticity is also absent in this region in Carcharodontosaurus , despite the otherwise extremely pneumatic nature of the braincase.

Anterior to the crista tuberalis, on that part of the exoccipital-opisthotic that contributes to the lateral wall of the braincase, the fenestra ovalis is visible. This opening faces laterally, as is usual for theropods. Coria & Currie (2002) described the fenestra ovalis of Carcharodontosaurus and Giganotosaurus as being reoriented relative to the normal theropod condition, such that they are exposed on the posterior surface of the braincase due to an enlargement of the jugular foramen and a repositioning of the crista tuberalis. Brusatte & Sereno (2007) reassessed the braincase of Carcharodontosaurus and identified the broken base of the crista tuberalis, which is in the usual position for theropods. They noted that heavy erosion in the fenestra ovalis region makes it appear as if this fenestra faced posteriorly, but that in life the crista tuberalis would have separated it from the occiput as is normal for theropods. Brusatte & Sereno (2007) did not discuss the braincase of Giganotosaurus , but the fenestra ovalis region, the base of the paroccipital process, and the crista tuberalis are heavily eroded on both sides, thus making interpretation difficult (MUCPv-CH 1).

However, although Coria & Currie (2002) were incorrect in placing the fenestra ovalis on the occiput in Carcharodontosaurus and possibly Giganotosaurus , they did correctly and astutely observe that the fenestra ovalis region of carcharodontosaurids is heavily modified relative to other theropods. Shaochilong helps clarify the anatomy of this region. In most theropods the fenestra ovalis opens almost entirely laterally, between the exoccipital-opisthotic and prootic, and immediately posterior to the opening for the facial (VII) nerve. This is seen in Allosaurus ( Madsen 1976) and Sinraptor ( Currie & Zhao 1993a), along with a range of other theropods (e.g., Welles 1984; Currie 1985, 1995; Charig & Milner 1997; Allain 2002; Sues et al. 2002; Xu et al. 2002; Brochu 2003; Rauhut 2004a; Sampson & Witmer 2007; Smith et al. 2007). In Shaochilong, on the other hand, the fenestra ovalis is located entirely within the exoccipital-opisthotic and is placed on the base of the crista tuberalis itself, at the region where the crista and paroccipital process meet. Thus, the fenestra ovalis in Shaochilong faces strongly anteriorly as well as laterally, although it is still technically located on the lateral wall of the braincase because it is anterior to the crista. Remarkably, the fenestra ovalis is placed far lateral (approximately 22 mm) to the endocranial cavity, and the two are linked via an elongate, anteromedially-trending bony canal that is completely enclosed by the exoccipital-opisthotic and perhaps medially by the prootic.

It is not clear that this condition is present in Carcharodontosaurus and Giganotosaurus , but the pattern of breakage in both taxa and the clear lack of a laterally facing fenestra ovalis posterior to the facial nerve opening is strong evidence that this is the case. Indeed, the fenestra ovalis of Giganotosaurus , as figured by Coria & Currie (2002: figs 3, 8) is clearly broken posteriorly, and this broken margin is seen in posterior view as a rounded surface that does appear to open onto the occiput ( Coria & Currie 2002: fig. 5). Therefore, it is apparent that the fenestra ovalis of Giganotosaurus trends strongly anterior-posterior, instead of medial-lateral as in most theropods. However, the broken surfaces delimit a missing section of the braincase, which appears as a notch in posterior view, between the crista tuberalis and the paroccipital process. Coria & Currie (2002:fig. 5) reconstruct this notch as open, and as a posterior continuation of the fenestra ovalis. However, this notch corresponds exactly to the position of the anteriorly-facing fenestra ovalis in Shaochilong. The only difference is that the posterior wall of the fenestra—that section of the braincase linking the crista and the paroccipital process—is completely preserved in Shaochilong whereas it is broken in Giganotosaurus . We suggest that Shaochilong and Giganotosaurus have the same condition, and that the fragile posterior wall of the fenestra ovalis has simply been broken in Giganotosaurus . This broken wall has been interpreted as a real, posteriorly exposed opening in Giganotosaurus , but in life the fenestra would have only opened anteriorly. Thus, Coria & Currie (2002) were correct in noting that the fenestra ovalis is reoriented in carcharodontosaurids, such that it now trends primarily anterior–posterior instead of medial–lateral. However, it is not actually exposed posteriorly. Carcharodontosaurus may have a similar condition, but both known braincases are too eroded in this region to be certain. Similarly, the condition in Acrocanthosaurus deserves further assessment.

Prootic. The prootic is nearly complete on both sides of the braincase but is somewhat crushed and eroded posteriorly on the left side. Sutures with the laterosphenoid, basisiphenoid, supraoccipital, and, if present, the basioccipital, are obscured by partial fusion. However, the prootic clearly overlaps the exoccipital-opisthotic posteriorly and extends slightly onto the paroccipital process, and the shape of this suture is apparent on both sides. Similarly, the prootic-parietal contact within the dorsal tympanic recess is also clear. The prootic does not participate in the margin of the fenestra ovalis, but rather extends posteriorly to terminate immediately anterodorsal to this opening. This relationship is better seen on the right side, which has been less affected by crushing.

Foramina for the trigeminal (V) and facial (VII) nerves are some of the most conspicuous features of the prootic. The openings are not set into the same fossa, but rather are divided by a stout bar of bone. The trigeminal foramen is positioned anterodorsal to the facial nerve fossa, and both openings are posterior to the level of the apex of nuchal wedge of the supraoccipital and parietal when the braincase is oriented with the frontals held horizontal. This is also seen in Carcharodontosaurus and Giganotosaurus ( Coria & Currie 2002; Brusatte & Sereno 2007, 2008) and may be a result of the posteroventrally sloping occiput that these taxa share with Shaochilong and Sinraptor . In Sinraptor the trigeminal (V) foramen is located approximately ventral to the apex of the supraoccipital tuberosity ( Currie & Zhao 1993a: fig. 7B), but in allosauroids that have ventrally sloping occiputs (the usual condition among theropods), such as Acrocanthosaurus ( Franzosa & Rowe 2005) and Allosaurus ( Madsen 1976) , both openings (V and VII) are positioned anterior to the supraoccipital tuberosity ( Coria & Currie 2002).

The prootic forms the anterior, dorsal, and ventral borders of the facial nerve fossa; the posterior border is formed by the exoccipital-opisthotic. There are two separate foramina for the facial nerve, both set into the same triangular fossa and one positioned anterodorsal to the other. The more dorsal opening is for the hyomandibular branch and the ventral foramen is for the palatine branch ( Franzosa & Rowe 2005). A series of distinct and deep grooves continues posteriorly and dorsally from the facial fossa. The most dorsal groove of the series is the deepest and most elongate; it trends dorsally, posteriorly, and laterally towards the fenestra ovalis and would have transmitted the hyomandibular branch of the facial nerve after it emerged from the braincase. However, the groove does not enter the fenestra ovalis, but rather is separated from it by the raised anterior rim of the fenestra. The series of grooves is demarcated dorsally by the otosphenoidal crest, a thin ridge of bone that continues posteriorly onto the exoccipital-opisthotic to form the dorsal rim of the fenestra ovalis and, ventral to the facial fossa, curves ventrally and posteriorly to become confluent with the posterior edge of the preotic pendant. Thus, the facial fossa is located within the confines of the otosphenoidal crest, as in other theropods ( Sampson & Witmer 2007) and would have been part of the middle ear space. Multiple openings for the facial nerve are rarely seen in theropods, but have been described in Acrocanthosaurus ( Franzosa & Rowe 2005) . Additionally, they appear to be present in Giganotosaurus but were not figured by Coria & Currie (2002: fig. 8), as the region of the second opening, anteroventral to the first, was obscured in anteroventrolateral view (MUCPv-CH 1, RBJB pers. obs.). The two foramina of Giganotosaurus and Shaochilong are close together and set within the same fossa, whereas they are more widely spaced and apparently located on different bones (the prootic and the basisphenoid) in Acrocanthosaurus ( Franzosa & Rowe 2005) .

Only a single opening for the trigeminal nerve is present on each side of the braincase. Some allosauroids possess multiple openings, or partially divided foramina, but this may be variable within taxa (Brusatte & Sereno 2007, 2008). In the genus Carcharodontosaurus , for instance, C. saharicus has a single opening whereas C. iguidensis shows a ‘binocular’-shaped opening that may indicate incipient division (Brusatte & Sereno 2007). Additionally, Brusatte & Sereno (2008) described a single foramen in one specimen of Acrocanthosaurus (OMNH 10146) and two foramina in another specimen (NCSM 14345). However, direct observation of NCSM 14345 reveals that only a single foramen is present, and thus there is no variability within Acrocanthosaurus . The trigeminal foramen is usually shared between the prootic and laterosphenoid in most theropods, and this appears to be the case in Shaochilong. Although a clear suture is not present, a raised and rugose margin that may represent a heavily fused suture extends dorsally and posteriorly from the posterodorsal corner of the trigeminal foramen. Ventral to this suture, and extending across what is presumably the laterosphenoid, is a deep depression that trends anterodorsally. This is most likely a groove for the ophthalmic branch of the trigeminal nerve, as is common in theropods generally ( Sampson & Witmer 2007). Above the groove, and along the presumed prootic-laterosphenoid suture, is a rugose surface that corresponds to the epipterygoid articular facet in Majungasaurus and other well described theropod braincases ( Sampson & Witmer 2007).

Dorsally, above the facial and trigeminal foramina and separated from them by a thick bar of bone, is a deep dorsal tympanic recess. This structure in Shaochilong is remarkably deep and more extensive than in any basal theropod we have ever seen, as well as most coelurosaurs we have examined. It extends onto the parietal and is overhung dorsally by a web of bone that projects ventrally from the parietal. A depression in this region is present in many theropods and is often referred to as a dorsal tympanic recess ( Rauhut 2004a), a structure that is present in living birds. However, it is possible that this depression may be apnuematic in some taxa, and instead may house jaw musculature ( Sampson & Witmer 2007). Indeed, it is located within the temporal region of the braincase and is separated from the lateral wall of the braincase by the stout bar of the parietal above the trigeminal and facial openings. However, as discussed by Rauhut (2004a), this depression is clearly pneumatic in Shaochilong, as its anterodorsal corner is penetrated by an enormous pneumatopore on each side of the braincase. The better preserved right pneumatopore is circular, with a diameter of nine millimetres. Pneumatopores such as these are unknown in other basal theropods, and indeed may only otherwise be present in birds, where they are smaller and less distinct ( Rauhut 2004a). Thus, they are considered an autapomorphy of Shaochilong among basal, non-coelurosaurian theropods.

Laterosphenoid. Much of the anterior region of the laterosphenoid is missing on both sides, including the capitate process that contacts the postorbital and the far anterior margin that contacts the frontal. However, a good portion of the posterior part of the laterosphenoid is present. This bone likely forms the anterior margin of the trigeminal foramen and contributes to some or all of the more anterior cranial nerve openings (II, III, IV, VI). However, as some or all of these are also formed by the orbitosphenoid, they are discussed in a single section below.

The posterior portion of the antotic crest—a thick ridge that separates the orbital space anteriorly from the temporal musculature space dorsally ( Sampson & Witmer 2007)—is preserved. The crest arises anterodorsal to the trigeminal nerve opening, and its presence is persuasive evidence that this part of the braincase pertains to the laterosphenoid, as only this bone forms the antotic crest in other well described theropods (e.g., Majungasaurus: Sampson & Witmer 2007 ). However, unlike Majungasaurus , the antotic crest is not essentially continuous with the more posterior otosphenoidal crest. Instead, the two crests are separated by a smooth and broad fossa that houses the trigeminal foramen. Posterior to this fossa the otosphenoidal crest curves ventrally and posteriorly to become confluent with the posterior margin of the preotic pendant. The base of the antotic crest is thick, suggesting that it was a stout and prominent structure. Prominent crests are also seen in Acrocanthosaurus (OMNH 10146), Carcharodontosaurus (SGM-Din-1), and Giganotosaurus ( Coria & Currie 2002) , whereas they are thinner and less offset laterally in Allosaurus and Sinraptor ( Sampson & Witmer 2007) . This may reflect increased attachment area for jaw adductor musculature more ventrally within the temporal space in carcharodontosaurids, as these taxa have reduced attachment sites on the dorsal surface of the frontal. However, the autapomorphic sagittal crest of Shaochilong suggests that the adductors did anchor firmly to the dorsal surface of the frontal, despite the fact that the supratemporal fossa is reduced in size.

Orbitosphenoid. Parts of the orbitosphenoid are clearly present in the vicinity of the pituitary fossa and interorbital region, but sutures with the surrounding bones (laterosphenoid, prootic, basisphenoid) are entirely obliterated. The orbitosphenoids are broken anterior to the openings for the optic (II) nerve but would have extended further anteriorly and dorsally to cup the olfactory bulbs, as shown by rugose attachment scars on the ventral surface of the frontal (see above). The suture with the laterosphenoid probably would have been in the region of the openings for the oculomotor (III) and trochlear (IV) nerves, based on the condition in other theropods (Brusatte & Sereno 2007; Sampson & Witmer 2007). However, the various cranial nerve openings and other foramina and fossae of this region are described together here.

The assorted endocranial structures of this area are divided into two general regions: the pituitary fossa posteriorly (including foramina for nerve VI) and the interorbital region anteriorly (including foramina for nerves II, III, IV). The hypophyseal fenestra itself is not visible since the interorbital septum is unossified (see below), but a depression for the pituitary is present posteriorly. Openings for the abducens (VI) nerve are located within this depression, not lateral to it as in many coelurosaurs ( Currie 1997). There is no prominent midline ridge between the left and right abducens foramina; a ridge is present in most theropods, including Allosaurus and Sinraptor , but is absent in Carcharodontosaurus and Giganotosaurus ( Coria & Currie 2002) . Foramina for the optic (II) and oculomotor (III) nerves appear to be similar in size. These are placed next to each other, with the optic foramen anterior to the oculomotor foramen, on the midline. The foramina for the trochlear (IV) nerves are much smaller than those for the optic and oculomotor nerves, and are more widely separated on the midline. Anterior to the trochlear foramen is a small opening whose function is unknown; it has also been identified in Giganotosaurus ( Coria & Currie 2002: fig. 8). It is not clear if there is a single midline opening for both left and right optic nerves or if there were separate foramina, since the interorbital septum that forms the midline of the braincase here is unossified. This condition is variable in allosauroids and is likely correlated with the ossification of the septum, as carcharodontosaurids with a bony septum have separate openings and those allosauroids with a cartilaginous or membraneous septum have a single foramen ( Franzosa & Rowe 2005).

Sphenethmoid, Mesethmoid, and Interorbital Septum. None of these various structures are present as ossified elements in Shaochilong. However, as they are frequently discussed in the literature and are an important character in allosauroid phylogeny, they deserve to be discussed further. Furthermore, these structures are often confused in the literature, as different structures are often referred to under the same umbrella term or referred to using misleading or incorrect terms ( Ali et al. 2008).

The interorbital septum is a parasagittal sheet oriented along the midline of the braincase that connects the cultriform process of parabasisphenoid ventrally to the sphenethmoid dorsally. It is part of a larger sagittal membrane, which stretches to the tip of the snout, and is usually cartilaginous or membraneous in most archosaurs ( Sampson & Witmer 2007). This is the case in most theropods, including Allosaurus and Sinraptor ( Currie & Zhao 1993a; Coria & Currie 2002; Brusatte & Sereno 2008). However, Carcharodontosaurus and Giganotosaurus , as well as some other large theropods (e.g., Majungasaurus: Sampson & Witmer 2007 ), have ossified or otherwise mineralized this sheet. Acrocanthosaurus is usually regarded as having an unossified septum (e.g., Coria & Currie 2002; Brusatte & Sereno 2008), but this region of the braincase is more extensively ossified than in other theropods and a bony septum may have been present and subsequently eroded in the two known braincases, as small parts remain in both specimens, particularly NCSM 14345. The ossified septum is undoubtedly absent in Shaochilong, as the region between the optic, oculomotor, and abducens foramina is clearly open on the midline. Although a narrow strut of bone could have divided these foramina on the midline, there is no thick, broken base of the septum, which would be present if the septum was ossified in life but eroded away (e.g., Coria & Currie 2002; Sampson & Witmer 2007).

Shaochilong also appears to lack an ossified sphenethmoid ( Fig. 6 ). This bone, located at the junction of the orbital and nasal cavities and usually ventral to the frontal, is a trough-like element that encloses the olfactory bulbs ventrally and anteriorly ( Sampson & Witmer 2007; see Ali et al. [2008] for review of homologies). It is often associated with a second ossification, termed the mesethmoid, which extends dorsally from the sphenethmoid trough to divide the olfactory tracts and bulbs on the midline ( Ali et al. 2008). The mesethmoid is sometimes considered to be an extension of the ossified interorbital septum (e.g., Sampson & Witmer 2007), but Ali et al. (2008) argue that it is a separate ossification that should be given its own name.

The presence of both the sphenethmoid and mesethmoid can be inferred from the shape of the braincase attachment scars on the ventral surface of the frontal. In particular, a midline scar between the olfactory tracts is strong evidence for an ossified mesethmoid, whereas curved scars that extend lateral and anterior to the olfactory bulb depressions indicate an ossified sphenethmoid. In Shaochilong only scars for the orbitosphenoid are present on the frontal; these are crescentic surfaces that extend to only midlength of the olfactory bulb depression. Midline scars or sutural surfaces anterior to the olfactory bulbs are absent. This is also the case in Allosaurus and Sinraptor , and Eocarcharia (Sereno & Brusatte 2008) , whereas other carcharodontosaurids (including Acrocanthosaurus: Stovall & Langston 1950 ; Sereno & Brusatte 2008) have ossified sphenethmoids and mesethmoids. This is well shown in Carcharodontosaurus ( Fig. 6 ; MNN IGU3), in which a raised midline rim between the olfactory bulb depressions is the attachment site for the mesethmoid, and a large, rugose, C-shaped scar anterior to the bulb depressions is the articular surface for the sphenethmoid.

Axis. The axis (IVPP V2885.5) is generally well preserved but is missing the anterior portion of the centrum, parts of the anterior and dorsal regions of the neural spine, and the lateral edges of the epipophyses ( Fig. 13 ). The entire axis is 145 mm tall dorsoventrally. The centrum is 57 mm long anteroposteriorly along its complete and uneroded dorsal margin, immediately ventral to where the centrum and neural arch are firmly fused, obliterating the neurocentral suture. The anterior articular surface of the centrum is eroded but it appears to have been approximately circular, with a reconstructed diameter of 51 mm. The posterior surface is also eroded but was clearly a dorsoventrally elongate oval, with a reconstructed depth of 53 mm and width of 34 mm. Details of the anterior surface are unclear, but preserved regions of the posterior surface indicate that it was shallowly concave. Some broken surfaces reveal what appears to be camellate internal bone structure, as has been described in other carcharodontosaurians ( Harris 1998; Brusatte & Sereno 2008).

Two proportional characters of the axial centrum are unusual in Shaochilong. First, an ovoid posterior articular surface is rare among allosauroids. Allosaurus ( Madsen 1976) , Giganotosaurus (MUCPv-CH-1), and Sinraptor ( Currie & Zhao 1993a) all possess nearly circular posterior surfaces. Brusatte et al. (2008: 22) described the posterior articular surface of Acrocanthosaurus as “substantially higher…than wide,” citing Harris (1998) as justification. However, the table of vertebral measurements provided by Harris (1998) unequivocally shows the posterior surface to be circular. The only other allosauroid with an ovoid posterior surface is the basal neovenatorid carcharodontosaurian Neovenator (Brusatte et al. 2008) . Second, shortened axes, with centra that are approximately as long as tall, are present in the derived carcharodontosaurids Acrocanthosaurus ( Harris 1998) and Giganotosaurus (MUCPv-CH-1), but not Neovenator (Brusatte et al. 2008) or the basal allosauroids Allosaurus ( Madsen 1976) and Sinraptor ( Currie & Zhao 1993a). Molnar et al. (1990) suggested, based on this latter character, that Shaochilong may belong to Tyrannosauridae , as a shortened axis is seen in Daspletosaurus , Tarbosaurus , and Tyrannosaurus . However, the elongate axes of basal allosauroids and tyrannosauroids (e.g., Dilong : IVPP V14243 ; Alioramus: Brusatte et al. 2009 b) suggests that this character evolved independently in the two groups. It is interesting that a shortened axis is mostly seen in derived, fairly large-bodied members of each clade, and may be related to the biomechanical constraints of large body size.

The ventral surface of the axial centrum is smooth and lacks a ventral keel or ridge. A low ventral axial ridge (often referred to as a “keel”) is present in carcharodontosaurids such as Acrocanthosaurus ( Harris 1998) and Giganotosaurus (MUCPv-CH-1), and more prominent keels are present among non-tetanuran theropods (e.g., Ceratosaurus , Dilophosaurus : see review in Brusatte et al. 2008). However, the ventral surface of the axis is rounded in Allosaurus and Sinraptor (Brusatte & Sereno 2008) and only a subtle ridge is present in the the neovenatorid carcharodontosaurian Neovenator (Brusatte et al. 2008) . The lateral surface of the axial centrum of Shaochilong is deeply depressed by a smooth fossa, which is excavated by a single large pneumatic foramen (“pleurocoel”) at its midpoint.

The neural arch is well preserved. The neural spine is extensive: it is 85 mm tall dorsoventrally as preserved and 31 mm wide mediolaterally at its base. It is inclined posterodorsally (contra Hu 1964) and appears to maintain a relatively constant width dorsally until it terminates at a broken margin. The dorsal tip of the spine is too eroded to determine the presence or absence of “crown-like” projections that are seen in some theropods, especially tyrannosaurids (e.g., Brochu 2003). The anterior surface of the neural spine is ornamented with a rugose midline ridge, a common feature of theropods that is likely an attachment site for the splenius capitis musculature ( Brochu 2003). In Shaochilong the ridge is eroded anteriorly but was clearly robust. On either side of this ridge the anterior surface of the neural arch is apneumatic, and lacks the deep pneumatic pockets that are present in some large tyrannosaurids (e.g., Brochu 2003). Similarly, the small pneumatic foramina described in Acrocanthosaurus ( Harris 1998) and Neovenator (Brusatte et al. 2008) , and also present in Giganotosaurus (MUCPv-CH-1), appear to be absent. However, this apparent absence could result from breakage along the anterior portion of the neural arch, as there is a shallow fossa, located anteriorly on the dorsolateral surface of the arch and most clearly visible on the left side, that may be pneumatic. The posterior surface of the neural spine is deeply concave. This concavity is deepest ventrally, where it forms an invaginated pocket, but shallows as it continues dorsally. Within the fossa is a dorsoventrally elongate, thick (5 mm mediolaterally), and rugose ligament attachment scar that trends across the entire height of the neural spine.

Only the left prezygapophysis is preserved. It has a flat, circular (18 mm diameter) articular facet that is barely offset from the remainder of the neural arch. The facet is located at the anteroventral corner of the neural arch and faces laterally but also slightly dorsally and anteriorly. The parapophysis is located at the anterodorsal corner of the lateral surface of the centrum, but is only visible as a heavily eroded region on the right side. The diapophysis is placed at the end of a short and indistinct transverse process, which projects straight ventrally as a small bulge. Ventral and medial to the transverse process, and partially covered by it in lateral view, is a shallow ovoid fossa that trends anterodorsally-posteroventrally. Posterior to this depression is a much deeper, triangular fossa that faces laterally and slightly posteriorly. Posterior to this second fossa, and separated from it by a 10 mm long upraised margin, is a smaller and shallower depression. This third fossa is ovoid, with a transversely oriented long axis, and faces strongly posteriorly and slightly laterally. This fossa is immediately anteroventral to the postzygapophysis. It is unclear if these fossae are homologous to the infraprezygapophyseal, infradiapophyseal, and infrapostzygapophyseal fossae of other theropods ( Wilson 1999), as individual bounding laminae are not clear due to poor preservation.

Each postzygapophysis has a large, flat facet that approximates the shape of a triangle with rounded margins. Each dimension of the rounded triangle is 30 mm long. The facet faces strongly ventrally but also slightly laterally. The base of the epipophysis is preserved on each side. Although both epipophyses are mostly broken it is clear that these structures were robust, pronounced processes that perpendicularly diverged from the neural spine in anterior and posterior views and protruded posteriorly past the postzygapophyses.

Caudal Vertebrae. Hu (1964) stated that six caudal vertebrae were present, three anterior caudals (IVPP V2885.6) and three middle caudals (IVPP V2885.7). Two of these, one anterior caudal and one middle caudal, were figured ( Hu 1964: fig. 12). Chure (2000) could only locate five of these during the course of his study, and SLB could only locate four when accessing the specimen again in January 2009 ( Figs 14 , 15 ). Strangely, each one of these was labeled as IVPP V2885.7, although one of them (which appears to be an anterior caudal) did not have a label written on it and was simply included in a box with the “middle caudal” (IVPP V2885.7) label. Thus, this is almost certainly one of the anterior caudals (IVPP V2885.6). Therefore, it is clear that the anterior caudal figured by Hu (1964: fig.12a) is missing, as is a second anterior caudal.

The four remaining vertebrae do not form a continuous series but can be placed in a relative sequence based on their size and morphology. The anteriormost caudal (referred to as “caudal A” here), which is the only remaining anterior caudal (IVPP V2885.6), has a centrum that is 72 mm long anteroposteriorly ( Fig. 14 a–f). The anterior surface is deeper (59 mm) than wide (47 mm), as is the posterior surface (55 mm high, 50 mm wide). Both surfaces are shallowly concave and the centrum is rounded ventrally, lacking a ridge or groove. However, the posterolateral corners of the centrum project somewhat ventrally to articulate with the chevrons. There is a small but discrete depression on each lateral surface of the centrum. On the left side the depression is an ovoid, shallow fossa (15 mm long anteroposteriorly by 6 mm deep dorsoventrally), but whether it contains any foramina is unclear due to weathering. On the right side there is a single, circular (5 mm diameter) opening located within an ovoid fossa. As the depression and foramen are located immediately ventral to the transverse processes, and penetrate the neural arch, they are unlikely to be homologous with the “pleurocoels” (pneumatic foramina) of the cervical and dorsal vertebral centra of most theropods (Sereno et al. 2008; O’Connor 2009; Wedel 2009), which are also present in the caudal centra in some allosauroids and other basal theropods ( Stromer 1931; Britt 1991; Calvo et al. 2004; Sereno et al. 2008). However, left-right asymmetry suggests the possibility of a pneumatic origin, perhaps homologous with foramina that are present within the infradiapophyseal fossa of some theropod dorsal vertebrae, which correspond in position. However, this caudal vertebra of Shaochilong lacks neural arch laminae ventral to the transverse process that commonly delimit the infradiapophyseal fossa. Furthermore, infraprezygapophyseal and infrapostzygapophyseal fossae, which are usually located anterior and posterior to the infradiapophyseal fossa, are absent. It is possible that the foramen on the right side of IVPP V.2885.6 represents a nutrient foramen, but such non-pneumatic foramina in theropod vertebrae are usually only on the order of 1 mm in diameter (RBJB pers. obs.). Therefore, it is difficult to say with certainty whether the lateral depressions and foramen of IVPP V.2885.6 were formed by pneumatic diverticulae.

The centrum and neural arch are fused but the interdigitating neurocentral suture between them is still partially visible. Only the bases of the transverse processes are preserved, but their thick cross sections (14 mm deep dorsoventrally) indicate that the processes were quite large in life. The trend of the broken base indicates that the processes extended laterally and posteriorly. The dorsal surface of the transverse process, at the point where it diverges from the arch, is indented with a smooth, deep, and broad fossa. None of the zygapophyses are preserved. The neural spine is present and is displaced posteriorly, such that its posterior margin is level with the posterior margin of the centrum but its anterior margin is located 12 mm behind the anterior face of the centrum. The spine is broken dorsally but is 46 mm long anteroposteriorly by 16 mm wide mediolaterally at its base.

The second remaining caudal vertebra (“caudal B”) is that figured by Hu (1964: fig. 12b), and belongs to IVPP V2885.7 ( Fig. 14 g–h). The centrum is 85 mm long, the strongly concave posterior face is deeper (50 mm) than wide (45 mm), and the more shallowly concave anterior face is also deeper (53 mm) than wide (45 mm). The ventral surface is smooth, without any keel or groove, and the lateral surfaces do not contain any fossae or foramina. Only the bases of the transverse processes are preserved, and these are thin (7 mm deep) and project straight laterally. There are no laminae linking the transverse process and centrum ventrally and there is only a shallow fossa on the dorsal surface of the base of the process. The neural spine is centrally located on the centrum and is reduced to a small bulge between the zygapophyses. The postzygapophyses extend 25 mm past the centrum posteriorly and project posterodorsally. The articular facets are flat, ovoid (20 mm long by 15 mm deep), and face strongly laterally and slightly ventrally. A flange continues past the articular facets posteriorly, and in this region there is a midline ridge between the two facets, which is robust in dorsal view. The prezygapophyses do not extend past the centrum anteriorly, but rather terminate 5 mm posterior to the anterior face. The two prezygapophyses diverge laterally, and together they clasp the postzygapophyses of the preceding vertebrae, which together form a single wedge. The prezygapophyseal facet is only preserved on the right side; is is flat, faces strongly medially and slightly dorsally, and is somewhat smaller than the postzygapophyseal facets.

Finally, two distal caudals also belong to IVPP V2885.7 ( Fig. 15 ), and are referred to as caudals C and D, respectively. The first is 85 mm long, with shallowly concave anterior (40 mm deep by 47 mm wide) and posterior (40 by 50 mm) faces. The second is 90 mm long, also with shallowly concave anterior (44 by 43 mm) and posterior (44 by 38 mm) faces. The ventral surface of the first centrum is smooth, whereas that of the second has a very slight, anteroposteriorly elongate, rectangular groove. The neural arch is not preserved on either caudal but articular scars for the arch are present on each centrum. Thus, it is unclear whether these vertebrae are anterior or posterior to the “transition point,” where theropod caudals lose their transverse processes and neural spines. The posterior face of each centrum extends ventrally relative to the anterior face to brace the centrum.

Figures

FIGURE 1. Skull reconstruction of Shaochilong maortuensis, based upon the paralectotype series described here (IVPP V. 2885.1 – 4). Compared to other carcharodontosaurids, Shaochilong has a shortened snout (shorter and deeper skull) and a smaller body size. Reconstruction by Brett Booth. Scale bar equals 5 cm.

FIGURE 2. Photographs of the right maxilla of Shaochilong maortuensis (IVPP V 2885.4) in lateral (a), medial (b), and ventral (c) views. Abbreviations: aof, antorbital fossa; ar, anterior ramus; gr, groove; idp, interdental plates; jpr, jugal process; ma, maxillary antrum; pmr, promaxillary recess; pnr, primary neurovascular foramina row; snr, secondary neurovascular foramina row. Designation “ m ” referrs to maxillary tooth position. Scale bar equals 5 cm.

FIGURE 3. Photograph of the right maxilla of Shaochilong maortuensis (IVPP V 2885.4) in medial view. Abbreviations: gr, groove; idp, interdental plates; ma, maxillary antrum; pmr, promaxillary recess. Scale bar equals 5 cm. l plates; ma, maxillary antrum; pmr, promaxillary recess. Scale bar equals 5 cm.

FIGURE 4. Photograph of the skull roof (right nasal, frontals, parietals) of Shaochilong maortuensis (IVPP V 2885.2) in dorsal (a), ventral (b), and left lateral (c) views. Abbreviations: cr, crest within supratemporal fossa; lc, lacrimal contact; nas, nasal; np, nasal process; npr, nasal pneumatic recess; obd, olfactory bulb depressions; oc, orbitosphenoid contact; of, orbital fossa; on, orbital notch; par, parietal; poc, postorbital contact; sc, sagittal crest; stf, supratemporal fossa. Scale bar equals 5 cm.

FIGURE 5. Photograph of the skull roof piece (right nasal, frontals, parietals; IVPP V 2885.2) articulated with the braincase (IVPP V 2885.1) of Shaochilong maortuensis in dorsal view. Abbreviations: cr, crest within supratemporal fossa; lc, lacrimal contact; nas, nasal; np, nasal process; npr, nasal pneumatic recess; oc, occipital condyle; poc, postorbital contact; pop, paroccipital process; sc, sagittal crest; sok, supraoccipital knob; stf, supratemporal fenestra; stfos, supratemporal fossa. Scale bar equals 5 cm.

FIGURE 6. Photograph of the frontals of Carcharodontosaurus iguidensis (a: MNN IGU 3) and Shaochilong maortuensis (b: IVPP V 2885.2) in ventral views. Abbreviations: mc, mesethmoid contact scar; obd, olfactory bulb depressions; oc, orbitosphenoid contact; of, orbital fossa; sc, sphenethmoid contact scar. Scale bars equal 5 cm.

FIGURE 7. Photograph of the right quadrate of Shaochilong maortuensis (IVPP V 2885.3) in anterior (a), posterior (b), lateral (c), medial (d), dorsal (e), and ventral (f) views. Abbreviations: qf, quadrate foramen; qja, quadratojugal articulation. Scale bar equals 5 cm.

FIGURE 8. Photographs and line drawings of the braincase of Shaochilong maortuensis (IVPP V 2885.1) in posterior (a, b) and right lateral (c, d) views. Abbreviations: aoc, antotic crest; atr, anterior tympanic recess; bo, basioccipital; bs, basisphenoid; bt, basal tuber; dtr, dorsal tympanic recess; ex-op, exoccipital-opisthotic; fm, foramen magnum; fo, fenestra ovalis; for, paracondylar openings representing jugal foramen and foramen for nerve XII; ls, laterosphenoid; oc, occipital condyle; p, parietal; pn, pneumatic foramen (pneumatopore); pop, paroccipital process; pp, preotic pendant; pro, prootic; scr, subcondylar recess; so, supraoccipital; sok, supraoccipital knob; sor, supraoccipital ridge. Roman numerals refer to cranial nerves. Scale bar equals 5 cm.

FIGURE 9. Photographs of the braincase of Shaochilong maortuensis (IVPP V 2885.1) in oblique left posterior (a) and oblique right posterior (b) views. Abbreviations: bs, basisphenoid; for, paracondylar openings representing jugal foramen and foramen for nerve XII; pf, pneumatic fossa; pn, pneumatic foramen (pneumatopore); scr, subcondylar recess. Scale bar equals 5 cm.

FIGURE 10. Photograph of the braincase of Shaochilong maortuensis (IVPP V 2885.1) in ventral view. Abbreviations: atr, anterior tympanic recess; bsr, basisphenoid recess; bsrw, basisphenoid recess web; bt, basal tubera; ex-op, exoccipital-opisthotic; fo, fenestra ovalis; for, foramen; ic, internal carotid entrance; p, parietal; pit, pituitary fossa; pro, prootic; ssr, subsellar recess. Roman numerals refer to cranial nerves. Scale bar equals 5 cm.

FIGURE 11. Photograph of the braincase of Shaochilong maortuensis (IVPP V 2885.1) in right lateral view. Abbreviations: aoc, antotic crest; atr, anterior tympanic recess; bsr, basisphenoid recess; bt, basal tubera; dtr, dorsal tympanic recess; fo, fenestra ovalis; for, paracondylar openings representing jugal foramen and foramen for nerve XII; oc, occipital condyle; pn, pneumatic foramen (pneumatopore). Roman numerals refer to cranial nerves. Arrowhead indicates position of foramen (which is hidden in lateral view). Scale bar equals 5 cm.

FIGURE 12. Photographs of the braincase of Shaochilong maortuensis (IVPP V 2885.1) in right lateral oblique views, including a complete photograph (a) and a closeup of the anterior pituitary region (b). Abbreviations: aoc, antotic crest; atr, anterior tympanic recess; bsr, basisphenoid recess; bt, basal tubera; ct, crista tuberalis (= metotic strut); dtr, dorsal tympanic recess; ecc, endocranial canal; f, fossa; fo, fenestra ovalis; for, foramen; ic, internal carotid entrance; pit, pituitary fossa; pn, pneumatic foramen (pneumatopore); orb, orbitosphenoid articulation scar; ssr, subsellar recess. Roman numerals refer to cranial nerves. Scale bar equals 5 cm and refers to image (a) only.

FIGURE 13. Photographs of the axis of Shaochilong maortuensis (IVPP V 2885.5) in anterior (a), posterior (b), left lateral (c), right lateral (d), and ventral (e) views. Abbreviations: f, fossa; las, ligament attachment site; lf, lateral fossae; mr, medial ridge; nc, neural canal; pa, parapophysis; paf, posterior articular surface; pf, pneumatic fossa; poz, postzygapophysis; prz, prezygapophysis; tvp, transverse process. Scale bar equals 5 cm.

FIGURE 14. Photographs of an anterior caudal vertebra (a – f: caudal A, IVPP V 2885.6,) and a posterior caudal vertebra (g – l: caudal B, IVPP V 2885.7) of Shaochilong maortuensis in left lateral (a, g), right lateral (b, h), ventral (c, i), dorsal (d, j), anterior (e, k), and posterior (f, l) views. Abbreviations: for, foramen; poz, postzygapophysis; prz, prezygapophysis; tvp, transverse process. Scale bar equals 5 cm.

FIGURE 15. Photographs of two posterior caudal vertebrae, caudal C (a – f) and caudal D (g – k), of Shaochilong maortuensis (IVPP V 2885.7) in left lateral (a, g), right lateral (b, h), anterior (c, i), posterior (d), ventral (e, j), and dorsal (f, k) views. Scale bar equals 5 cm.

Tables

TABLE 1. Proportion of the base of the ascending ramus of the maxilla excavated by the antorbital fossa. Measurements are taken along an anteroposterior line, parallel with the tooth row, beginning from the anteroventral corner of the antorbital fenestra and continuing until the anterior margin of the maxilla.

Taxon	Ratio	Source
Shaochilong Acrocanthosaurus	0.15 0.50	IVPP V2885.4 Eddy 2008
Allosaurus	0.60	Madsen 1976
Carcharodontosaurus Eocarcharia	0.29 0.64	SGM-Din-1 MNN GAD2
Giganotosaurus	0.40	MUCPv-CH-1
Mapusaurus Neovenator	0.40 0.65	Coria & Currie 2006 MIWG 6348
Sinraptor	0.62	Currie & Zhao 1993

TABLE 3. Ratio of the longest anteroposterior length of the supratemporal fossa on the frontal to the longest anteroposterior length of the frontal itself. Measurements are taken along an anteroposterior line, parallel to the sagittal axis of the skull.

Taxon Shaochilong	Ratio 0.34	Source IVPP V2885.4
Acrocanthosaurus	0.28	Eddy 2008
Allosaurus Carcharodontosaurus	0.47 0.24	Madsen 1976 SGM-Din-1
Eocarcharia	0.26	MNN GAD2
Giganotosaurus Sinraptor	0.29 0.40	Coria & Currie 2002 Currie & Zhao 1993