The Phylogenetic Position of the Ornithischian Dinosaur Stenopelix valdensis from the Lower Cretaceous of Germany and the Early Fossil Record of Pachycephalosauria

The holotype of Stenopelix valdensis is the most completely known dinosaur specimen from the “Wealden” (Lower Cretaceous) of northwestern Germany, but its phylogenetic position has remained highly controversial. Most recent authors have suggested affinities with the ornithischian clade Marginocephalia, and most commonly to the marginocephalian subclade Pachycephalosauria. A pachycephalosaurian identity would make Stenopelix the only confirmed pre-Late Cretaceous member of this clade, breaking up an extensive ghost lineage which extends to the inferred origin of Pachycephalosauria in the Middle—Late Jurassic. Based upon re-examination of the holotype we here review the characters that have previously been used to assign Stenopelix to either Pachycephalosauria or Ceratopsia. All of these characters are problematic, being based upon inaccurate anatomical interpretations, or having more widespread distributions within Ornithischia than previously realised. We conclude that although the overall anatomy of Stenopelix is consistent with marginocephalian affinities, there is insufficient evidence to support referral to either Pachycephalosauria or Ceratopsia; we consider Stenopelix ?Marginocephalia. A brief review indicates that there is no compelling fossil evidence for pachycephalosaurs prior to the Late Cretaceous.


Introduction
The enigmatic ornithischian dinosaur Stenopelix valdensis Meyer, 1857, is the most completely known dinosaur speci− men from the "Wealden" (Lower Cretaceous) of Germany, and the phylogenetic position of this taxon has been the sub− ject of considerable debate. The holotype and only known specimen (GZG 741/2) was collected from the Obernkirchen Sandstone (Berriasian) near Bückeburg, Lower Saxony (NW Germany), in 1855. Since its discovery Stenopelix has been assigned to a variety of clades within Ornithischia; most re− cent speculation has focused on the suggestion that Steno− pelix might be referable to the clade Marginocephalia (e.g., Sereno 1987Sereno , 1999Sereno , 2000Dodson 1990;Maryańska et al. 2004;Sullivan 2006). Stenopelix has been assigned to both of the constituent clades of Marginocephalia, Ceratopsia (e.g., Ostrom 1970;Sues and Galton 1982) and Pachycepha− losauria (e.g., Maryańska and Osmólska 1974;Coombs 1982;Sereno 1987Sereno , 2000Maryańska et al. 2004), and has additionally been considered as the sister taxon to Margino− cephalia (Dodson 1990). The few published numerical phylogenetic analyses to include Stenopelix as a terminal taxon have identified it as the most basal known pachy− cephalosaur (Sereno 1999(Sereno , 2000Butler et al. 2008), al− though this phylogenetic position is generally weakly sup− ported and the taxon acts as a "wildcard" in suboptimal trees (Butler et al. 2008). An identification as a basal pachy− cephalosaur would make Stenopelix an extremely important taxon for understanding the evolution of the clade; such an identification would break up the extensive ghost lineage be− tween the inferred origin of Pachycephalosauria in the Mid− dle-Late Jurassic (based upon the Oxfordian age of the ceratopsian Yinlong downsi, Xu et al. 2006) and the abundant Late Cretaceous pachycephalosaur material known from North America and Asia (e.g., Maryańska and Osmólska 1974;Sereno 2000;Sullivan 2003Sullivan , 2006Maryańska et al. 2004; all other proposed pre−Late Cretaceous pachycephalo− saur records are questionable, see below). However, in a re− cent review of pachycephalosaurian taxonomy, Sullivan (2006) dismissed the evidence in favour of a pachycephalo− saurian identity for Stenopelix. Furthermore, Sullivan (2006) and Bakker et al. (2006) have argued that the evidence in support of marginocephalian monophyly is weak, although work published contemporaneously by Xu et al. (2006), and later by Butler et al. (2008), provided additional evidence for the monophyly of this clade. Given its putative basal posi− tion, Stenopelix is a potentially important taxon in elucidat− ing pachycephalosaurian evolution and testing margino− cephalian monophyly. For those reasons we here provide a review of the validity of the characters that have been sug− gested in support of pachycephalosaurian and ceratopsian af− finities for Stenopelix; we additionally note several poten− tially phylogenetically informative characters not previously discussed. We do not attempt to provide a comprehensive redescription of Stenopelix (other than to document relevant inaccuracies in previous descriptions): descriptions and fig− ures of Stenopelix are provided by Meyer (1859), Koken (1887), Schmidt (1969), Sues and Galton (1982), and Sereno (1987).

Material and methods
The holotype of Stenopelix valdensis was examined first− hand in the collections of the GZG on two separate occasions in the spring of 2005 and 2007 by RJB and independently by RMS on a separate occasion, also in the spring of 2007. Comparative material of basal ceratopsians, pachycephalo− saurs and other ornithischians has been examined in numer− ous other institutions; see Sullivan (2003) for a summary of pachycephalosaur material examined by RMS, and Butler (2005) and Butler et al. (2008) for details of ornithischian material examined by RJB. Furthermore, the holotypes of the recently described basal ceratopsian Yinlong downsi (IVPP 14530;Xu et al. 2006) and the basal ornithopod Chang− chunsaurus parvus (JLUM L0203−j−Zn2; Zan et al. 2005) were examined by RJB as part of this study.
Interpretation of the holotype of S. valdensis is difficult due to its mode of preservation; the partial skeleton is pre− served as impressions on two sandstone slabs (Fig. 1A, C) that only partly overlap, making detailed examination of the material difficult. Furthermore, the skeleton is dorsoven− trally compressed and many of the bones have been sub− jected to crushing. Latex casts of the holotype, prepared by Sues and Galton (1982), are currently held at Göttingen, and form the primary basis for this discussion (Figs. 1B, D, 2-6), although details were checked against the original sandstone slabs. The larger of these latex casts (Fig. 1B) shows the left partial forelimb and scapula, dorsal vertebrae, sacral and pel− vic region, hindlimbs and proximal caudal vertebrae in dor− sal view; the smaller latex cast (Fig. 1D) exposes the sacral and caudal vertebrae and parts of the pelvis and hindlimb in ventral view. The difficulty of interpreting the preserved ma− terial should be kept in mind and provides one major problem in assessing the phylogenetic position of this taxon; the other major problem is the absence of cranial material (Sues and Galton 1982). This is particularly unfortunate as most pachy− cephalosaur synapomorphies (and most ceratopsian and ornithopod synapomorphies) are cranial character states (e.g., Sereno 1999Sereno , 2000Butler et al. 2008).
For the purposes of this study we assume that Margino− cephalia is a monophyletic clade, as supported by recent nu− merical phylogenetic analyses (Sereno 1999(Sereno , 2000Butler 2005;Xu et al. 2006;Butler et al. 2007Butler et al. , 2008, although one of us (RMS) retains doubts as to the accuracy of this result (e.g., Bakker et al. 2006;Sullivan 2006). Marginocephalia and Ornithopoda are sister−taxa within the node−based clade Cera− poda, which is itself a subclade of the stem−based clade Neo− rnithischia (see Barrett et al. 2005 andButler et al. 2008 for discussion of the definitions of Neornithischia and Cerapoda) which also includes some probable non−cerapodan taxa such as the Middle Jurassic Agilisaurus and Hexinlusaurus (e.g., Butler 2005; Barrett et al. 2005) and the Early Jurassic ornith− ischian Stormbergia (Butler 2005). Neornithischia is the sis− ter−taxon to Thyreophora, together forming Genasauria. Non− genasaurian taxa include Eocursor parvus (Butler et al. 2007), and possibly Lesothosaurus diagnosticus (Sereno 1986(Sereno , 1999 although the phylogenetic position of this latter taxon is highly unstable (e.g., Butler 2005;Butler et al. 2007Butler et al. , 2008. The phylogenetic position of Heterodontosauridae within Ornith− ischia is poorly resolved: the clade has been identified as basal members of Ornithopoda (Sereno 1986(Sereno , 1999, the sister−clade to Marginocephalia (Xu et al. 2006) or as non−genasaurian basal ornithischians (Butler et al. 2007(Butler et al. , 2008. Based upon the senior author's phylogenetic analyses, we favour a non−gena− saurian phylogenetic position, although this interpretation does not significantly affect our discussion of the position of Stenopelix.

Stenopelix valdensis
Holotype: GZG 741/2 (formerly GPI Gö 741−2), impressions in sand− stone of a partial articulated skeleton in dorsal and ventral views, includ− ing at least 13 dorsal vertebrae, five or six sacral vertebrae, more than 30 caudal vertebrae, ribs, ossified tendons, distal left scapula, distal left hu− merus, left radius, left ulna, left metacarpal 1, left and right ilia, pubes, and ischia, left femur, proximal right femur, left and right tibiae, fibulae and proximal tarsals, complete right pes and partial left pes. The holotype rep− resents a small individual (less than one metre in length); subadult status is suggested by the absence of fusion between neural arches and centra, ad− jacent sacral centra, and sacral/caudal ribs and neural arches. Emended diagnosis.-Small ornithischian dinosaur with the following autapomorphies: postacetabular process of ilium tapers uniformly to a rounded point; ischial blade with dis− tinct bend at mid shaft; ischial blade is broadest at mid shaft, tapers proximally and distally; prominent proximodistally extending ridge present on the medial surface of the proximal 50% of the shaft of the ischium; distal end of ischial blade is arched dorsoventrally, with a convex lateral surface and a concave medial surface (modified from Sereno 1987).
Comments. -Sereno (1987) provided the first formal diag− nosis for Stenopelix, based upon five autapomorphies, al− though these autapomorphies were neither described nor dis− cussed. The first of the autapomorphies proposed by Sereno (1987) refers to the morphology and length of the post− acetabular process of the ilium (Fig. 2). The relatively short postacetabular process probably represents an ornithischian plesiomorphy (see below), but the uniformly tapering form of the process does appear to be unique.
Three of the autapomorphies proposed by Sereno (1987) refer to the morphology of the ischia (Figs. 3, 4), and a fourth ischial autapomorphy is added here. The ischia of Stenopelix are highly unusual and are marked by a distinct and promi− nent bend at mid shaft ( Fig. 4). At this point the ischial blade is at its broadest, and it tapers both proximally and distally from this point. As noted by Sues and Galton (1982), but not by Sereno (1987), there is a prominent ridge present on the medial surface of the proximal half of the ischial shaft (Fig.  3). Distally, the ischial blade is transversely compressed, and is arched dorsoventrally, with the lateral surface convex and the medial surface gently concave.
The last of the possible autapomorphies listed by Sereno (1987) refers to the dorsoventrally compressed prepubic pro− cess. As discussed by Sereno (1987), the prepubic process of Stenopelix (Figs. 1D, 3B) is considerably longer than recon− structed by Galton and Sues (1982), appears to be strongly dorsoventrally compressed, and tapers slightly anteriorly. The strong dorsoventral compression of the process is not considered here to be autapomorphic for Stenopelix: a simi− lar strongly dorsoventrally compressed prepubic process is also known in Yinlong (Xu et al. 2006: fig. S2F), Archaeo− ceratops (IVPP V11114) and Psittacosaurus (Sereno 1987;Averianov et al. 2006).

General features
The ornithischian identity of Stenopelix is confirmed by a large number of features identified as ornithischian synapo− morphies by previous phylogenetic analyses (e.g., Sereno 1986Sereno , 1999Norman et al. 2004a;Butler 2005;Butler et al. 2007Butler et al. , 2008, including: the presence of ossified epaxial ten− dons in the dorsal region; the increased number of sacral ver− tebrae (at least five are present, see below); the elongated, strap−like preacetabular process of the ilium (Fig. 2); the presence of an opisthopubic pelvis and a prepubic process (Figs. 1D, 3B); the well−developed anterior trochanter of the femur; the strong transverse expansion of the distal end of the tibia (Fig. 6). Within Ornithischia, a number of features link Stenopelix to Neornithischia-several of these features were listed by Sereno (1987: 364-365). The features linking Ste− nopelix to Neornithischia include: the absence of a supra− acetabular flange on the ilium ( na, neural arches and spines; s1, s2, s3, s4, s5, sacral centra; ?s6, possible 6 th sacral centrum (may instead represent first caudal centrum); sr, sr1, sr2, sr5, sacral ribs; sr6, possible 6 th sacral rib (may instead represent first caudal rib).
Ischial peduncle of the ilium broadly swollen and projects ventrolaterally.-In basal ornithischians (e.g., Stormbergia dangershoeki, BMNH R11000) and in Stenopelix (Sereno 1987) the ischial peduncle of the ilium (Fig. 2) projects ven− trally and is not broadly swollen. By contrast, in most basal cerapodans including ornithopods (e.g., Hypsilophodon foxii, Galton 1974: figs. 48A, 50A), pachycephalosaurs (e.g., Ho− malocephale calathocercos, Maryańska and Osmólska 1974: fig. 5A1), and ceratopsians (e.g., Archaeoceratops oshimai, IVPP V11114) the ischial peduncle is broadly swollen and projects ventrolaterally. Sereno (1987: 355) noted the absence of this feature in Stenopelix, and suggested that it might be a result of post−mortem compression. However, a strongly later− ally projecting ischial peduncle is also absent in the basal ceratopsian Psittacosaurus sibiricus (Averianov et al. 2006), so it may be that the distribution of this character is more com− plex than previously realised. Butler et al. (2008) included this character in their analysis (their character 177) and found that the derived state represented an unambiguous synapomorphy  Coria and Salgado 1996;Orodro− meus makelai, Scheetz 1999), ceratopsians (You and Dodson 2004), and pachycephalosaurs (Maryańska et al. 2004). The number of sacral vertebrae/ribs in Stenopelix valdensis has been controversial and is difficult to interpret. Koken (1887) and Schmidt (1969) believed that at most three sacrals were present, whereas Sues and Galton (1982) argued that there were at least six (possibly seven) sacrals. Sereno (1987) argued that there was evidence for at least five and probably six sacrals, but that there is no evidence for a seventh sacral.
Adjacent sacral vertebrae are unfused, and sacral ribs are not fused to sacral centra. The first two sacrals are clearly identifiable in ventral view (Fig. 3B), and their centra are transversely expanded and anteroposteriorly elongate. The first sacral rib is borne on the lateral surface of the first sacral centrum; the right first sacral rib is disarticulated and angles posteroventrally, distally overlapping the second right sacral rib, whereas the left first sacral rib projects nearly straight lat− erally. At its distal end the left first sacral rib is expanded anteroposteriorly and forms a small but distinct contact with the distal end of the left second sacral rib (Sues and Galton 1982;contra Sereno 1987 ond sacral rib is positioned at the anterior end of sacral centrum two and extends a short distance onto the posterior end of sacral centrum one; the right second sacral rib is disarticulated and angles posteroventrally whereas the left second sacral rib projects laterally and only slightly posteri− orly. In dorsal view the left second sacral rib contacts the ilium medial to the acetabulum. More posteriorly, additional sacral ribs are visible in dorsal and ventral views, although matching ribs with neu− ral arches is problematic because the centra have been tilted anteroventrally, resulting in their unfused neural arches be− ing displaced anteriorly in dorsal view (Fig. 3A). As dis− cussed by Sereno (1987), the left fifth sacral rib is clearly identifiable in ventral view by counting posteriorly from the first and second sacral centra and ribs; it is also exposed in dorsal view (Figs. 3A, B), is offset laterally from its articu− lation, is expanded anteroposteriorly at its distal end, and would have contacted the medial surface of the postace− tabular process of the ilium. The next (sixth) pair of ribs is visible in both dorsal and ventral views, although the left of the pair is displaced laterally and slightly anteriorly. As noted by Sereno (1987), the distal end of the right sixth rib extends nearly as far laterally as the right fifth sacral rib, but is not expanded anteroposteriorly and tapers to a narrow pointed tip. There is no obvious surface for contact with the postacetabular process of the ilium, and it is appears un− likely that this rib formed a substantial contact with the iliac blade. The neural arch of the vertebra that bore this rib re− sembles the neural arches of the succeeding proximal cau− dal vertebrae in having an anteroposteriorly narrow neural spine, and postzygapophyses positioned at a high angle (about 80°) to the horizontal. It is unclear whether this ver− tebra and ribs represent the sixth sacral or the first caudal. Therefore, there is unambiguous evidence for at least five sacral vertebrae in Stenopelix and no evidence to support the suggestion that there might be a seventh sacral (Sues and Galton 1982); however, the evidence for a sixth sacral is ambiguous.

Ceratopsian characters
Characters that have been suggested to link Stenopelix val− densis with ceratopsians are: Ischium with downward curvature (Ostrom 1970). -Ostrom (1970: 134) commented on the phylogenetic relation− ships of Stenopelix valdensis, suggesting it was of "psittaco− saurian, if not ceratopsian, affinity". Ostrom (1970) did not ex− plicitly outline the reasons for this referral, but he did note that Stenopelix shared with ceratopsians a strong downward curva− ture of the ischium (Figs. 3, 4). However, Sereno (1987) noted that the ischium is curved, to varying degrees, in ankylosaurs, advanced ornithopods, pachycephalosaurs and ceratopsians, and that curvature of the ischium is absent in psittacosaurs (Sereno 1987; although more recent data indicates that a curved ischium is present in Psittacosaurus sibiricus, Averia− nov et al. 2006). Moreover, as noted by Sereno (1987) and dis− cussed above (see diagnosis), the ischia of Stenopelix val− densis have a pronounced bend midway along the bone (Fig.  4), a unique morphology that does not resemble the gentle bow or arc that characterizes most ceratopsians and pachy− cephalosaurs. We agree with Sereno (1987) in concluding that this character provides little evidence for uniting Stenopelix valdensis with ceratopsians.
"The structure of the pelvic girdle, especially the form of the ilium and the reduced pubis" (Galton and Sues 1982: 188).-As discussed by Sereno (1987Sereno ( , 2000, there do not appear to be any characters of the pelvic girdle that are shared uniquely by ceratopsians and Stenopelix, nor to date have pelvic synapomorphies been identified for Ceratopsia. The prepubic process of Stenopelix is well−developed and not ob− viously "reduced". Although the preserved base of the post− pubis is slender and may not have supported a substantial process, it is impossible to accurately assess the postpubis length on the basis of the available material (Sereno 1987). Furthermore, reduction of the postpubis also characterises pachycephalosaurs (e.g., Maryańska and Osmólska 1974).

Pachycephalosaurian characters
The characters that supposedly link Stenopelix valdensis with pachycephalosaurs are: Anterior caudal ribs long (Maryańska and Osmólska 1974). -As discussed by Sues and Galton (1982) and Sereno (1987), the anterior caudal ribs of Stenopelix (see Fig. 3A) are not elongate like those of pachycephalosaurs: Maryańska and Osmólska (1974) misidentified the elongate posterior sacral ribs (sacral rib 5, and the rib belonging to the following verte− bra-see discussion above) present in Stenopelix as anterior caudal ribs. The only possible caudal with elongate ribs is the vertebra that as discussed above may represent either the last (sixth) sacral or the first caudal. More posterior caudals have short caudal ribs. This character does not, therefore, provide support for pachycephalosaurian affinities.
Pubis excluded from the acetabulum (Maryańska and Osmólska 1974;Coombs 1982;Maryańska et al. 2004).-As previously noted by Schmidt (1969), Sues and Galton (1982) and Sereno (1987), a reconstruction of the pelvic girdle of Stenopelix valdensis requires a substantial contribution of the pubis to the acetabulum, and there is no substantial evidence that the pubis was excluded from the acetabulum. This char− acter does not, therefore, provide support for pachycephalo− saurian affinities. (Coombs 1982). -Coombs (1982: 99) suggested that a postpubic process was absent in Stenopelix valdensis; however, as noted by Schmidt (1969: fig. 1), Sues and Galton (1982), and Sereno (1987), the base of a postpubic process is clearly preserved on the right pubis, although, as discussed above, the length of the postpubic pro− cess is known. (Maryańska and Osmólska, 1974).-As discussed by Maryańska and Osmólska (1974), Sues and Galton (1982) and Sereno (1987), the tibia is dis− tinctly shorter than the femur in Stenopelix with a tibia:femur ratio of approximately 0.93. The equivalent ratio in Stego− ceras validum is 0.98 (Sues and Galton 1987) and the tibia is also shorter than the femur in Wannanosaurus yansiensis, al− though the exact ratio cannot be determined due to poor pres− ervation (RJB and Zhao Qi, unpublished data). Tibiofemoral ratios greater than 1.0 occur primitively for Ornithischia (e.g., Thulborn 1972;Galton 1974;Santa Luca 1980), but tibiofemoral ratios less than 1.0 also occur independently among at least four independent clades within Ornithischia (in pachycephalosaurs, in some derived "hypsilophodontids" and Iguanodontia, in Ceratopsidae, and in nearly all thyreo− phorans). Furthermore, as discussed by Sereno (1987), the tibiofemoral ratio of Stegoceras is actually closer to that in Psittacosaurus sinensis (1.01; Sereno 1987) than to that of Stenopelix. We therefore agree with Sereno (1987) that this character provides only weak support for pachycephalo− saurian affinities. (Sereno 1987(Sereno , 2000. -Sereno (1987: 365) noted that, "In Stenopelix and pachy− cephalosaurs the fourth through sixth sacral ribs are notice− ably longer than the second and third [sacral ribs]." The re− sult of this elongation, according to Sereno (1987), is that the posterior end of the sacrum is at least 30% broader than the anterior end. This statement was supported for Stenopelix valdensis by comparing measurements of the right second sacral rib (21 mm according to Sereno 1987) with the left fifth and sixth sacral ribs (32 mm according to Sereno 1987).

Posterior sacral ribs elongate
The second sacral ribs are moderately shorter than the more posterior sacral ribs (see Table 1). However, the maxi− mum length of the right first sacral rib (29.3 mm), measured in ventral view, is nearly 90% of the length of the most elon− gate posterior sacral rib. Furthermore the right second sacral rib has a maximum length (25 mm) that exceeds that given by Sereno (1987) and exceeds the maximum measurable length of the second left sacral rib. The discrepancy in measure− ments between right and left anterior sacral ribs is explained by the fact that the left sacral ribs are incompletely exposed. Sereno (1987Sereno ( , 2000 appears to have underestimated the length of the anterior sacral ribs of S. valdensis, and the first sacral rib certainly had a length nearly comparable to that of the posterior sacral ribs. Even if the posterior sacral ribs were slightly more elongate than the anterior sacral ribs, this does not necessarily imply that the posterior end of the sacrum was broadened as occurs in pachycephalosaurs: the bases of the anterior sacral ribs would have been positioned further from the midline than the bases of the posterior sacral ribs, as a result of the transverse expansion of the anterior sacral centra. Furthermore, there is no evidence that the postace− tabular process of the ilium was deflected laterally to accom− modate a posterior transverse expansion of the sacral compa− rable to that seen in pachycephalosaurs (in fact, as noted by Sereno [1987:354], the postacetabular process appears to be deflected weakly medially rather than laterally; see below). Slight broadening of the posterior part of the sacrum also oc− curred in some basal ceratopsians, as demonstrated by the holotype specimen of Archaeoceratops (IVPP V11114) in which the last sacral rib is notably more elongate than sacral ribs 1-3 (You and Dodson 2003: fig. 2C). In light of these observations, we do not believe that this character provides substantial evidence to support a pachycephalosaurian iden− tification for Stenopelix. (Sereno 1987(Sereno , 2000.-The dorsal margin of the ilium of Stenopelix valdensis is thickened into a narrow horizontal plane along much of its length and slightly overhangs the lat− eral margin of the ilium (Figs. 2, 3A). This thickening is weak and narrow at the distal end of the postacetabular pro− cess, but expands transversely immediately posterior to the ischiadic peduncle to a width of around 7 mm. It remains at a similar (although variable) thickness along most of the length of the blade anterior to the ischiadic peduncle. Anteriorly, the preacetabular process curves gently laterally and undergoes weak torsion along its length so that its lateral surface faces dorsolaterally rather than strictly laterally. This torsion, and possibly post−mortem compression, has reduced the distinc− tion between the dorsal and lateral surfaces of the preace− tabular process of the ilium such that they resemble a single surface; however, these surfaces can still be distinguished from one another by a weak anteroposteriorly extending ridge, referred to by Sereno (1987: 354) as a "subtle crest". Thus, strictly speaking the dorsal margin of the preacetabular process of the ilium does not expand significantly toward its distal end; the apparent expansion is a result of the torsion of the preacetabular blade. A comparable morphology occurs in the basal ceratopsian Yinlong (RJB personal observations of IVPP 14530; Xu et al. 2006: fig. S2a-d): the dorsal margin of the ilium is thickened along most of the length of the blade, while the lateral surface of the preacetabular process under− goes torsion such that it faces dorsolaterally towards its ante− rior end. Combined, this results in the anterior end of the preacetabular process being compressed dorsolaterally/ ventromedially; moreover, in dorsal or ventral view the preacetabular process appears to expand slightly in the trans− verse plane at its distal tip (Xu et al. 2006: fig. S2c, d). A sim− ilar morphology is also present in a previously undescribed fragment of the preacetabular process of the left ilium of the paratype specimen of the primitive pachycephalosaur Wan− nanosaurus yansiensis (Butler and Zhao 2009 By contrast, in other pachycephalosaurs the dorsal mar− gin of the ilium is everted into a much broader shelf along most of its length and this shelf expands anteriorly as a flat horizontal surface (e.g., Maryańska and Osmólska 1974: pl. 29: 1). It is probable that the thickened dorsal margin of the ilium and the torsion of the preacetabular process seen in Stenopelix and Wannanosaurus represent the initial stage in the development of the morphology seen in other pachy− cephalosaurs. However, given that a similar condition is present in the most basal known ceratopsian, Yinlong, it is plausible that this character represents a marginocephalian plesiomorphy. Furthermore, expansion of the dorsal margin of the ilium is seen in a range of other ornithischians, includ− ing all thyreophorans (RJB and Susannah Maidment, unpub− lished data), the ornithopod Zalmoxes (Weishampel et al. 2003), and ceratopsids (Hatcher et al. 1907), demonstrating that this character is highly homoplastic within Ornithischia. As a result, this character provides little support for pachy− cephalosaurian affinities. (Sereno 1987(Sereno , 2000.

Scapula blade narrow and strap−like
-Only the distal−most portion of the left scapula (Fig. 5) is preserved in Stenopelix valdensis, and was first identified by Sereno (1987). It is narrow and only slightly expanded at its distal end, and was compared (Sereno 1987(Sereno , 2000 with the scapula of the pachycephalosaur Stegoceras validum (Sues and Galton 1987: fig. 10). However, as noted by Sereno (1987Sereno ( , 2000, a narrow and strap−like scapula is also known in Heterodontosaurus (Santa Luca 1980: fig. 10), and in many basal ceratopsians including Protoceratops andrewsi (Brown and Schlaikjer 1940: fig. 26), Graciliceratops mon− goliensis (Maryańska and Osmólska 1975: fig. 2) and Psit− tacosaurus sinensis (Young 1958: fig. 53). As a result, this character probably has a widespread distribution at the base of Marginocephalia. Furthermore, the true proportions of the scapula are difficult to confirm for Stenopelix (due to its in− complete exposure in the holotype specimen). For these rea− sons, this character provides little support for pachycephalo− saurian referral. (Sereno 1987).-Sereno suggested that the postacetabular process of Stenopelix valdensis underwent a subtle medial deflection at its base, and that this might be homologous with the stronger medial deflec− tion of the distal portion of the pachycephalosaurian post− acetabular process. The right ilium has a well−preserved post− acetabular process that does appear to undergo a weak medial deflection at its base. However, as noted by Sereno (1987) the medial flexure in Stenopelix is extremely subtle (and could possibly be explained by post−mortem compression), and oc− curs at the base of the postacetabular process, rather than at mid length along the process as occurs in pachycephalosaurs (e.g., Maryańska and Osmólska 1974). We therefore agree that "the subtle flexure of the postacetabular process in Steno− pelix does not constitute convincing evidence for pachyce− phalosaur affinity" (Sereno 1987: 370).

Iliac blade with gentle sigmoid curve
In summary, the preceding eight characters provide only weak and ambiguous support for pachycephalosaur affini− ties. As noted by Sereno (1987), a number of pachycephalo− saurian synapomorphies are definitively absent in Stenopelix . Tibia and fibula are exposed in posterior view; pes is exposed in anterior view. Note first phalanx of digit I. Roman numerals I-IV correspond to respective digits.
valdensis. These include: sixth sacral rib more elongate than fifth sacral rib; anterior caudal vertebrae with elongate cau− dal ribs; caudal "basket" of ossified tendons; reduction in rel− ative length of forelimb; medially directed tab−shaped pro− cess on the dorsal margin of the ilium; pubic peduncle of the ischium strongly dorsoventrally compressed; body of pubis very small and nearly excluded from the acetabulum. Al− though Sereno (1987) suggested that interlocking grooves and ridges were absent from the dorsal pre− and postzygapo− physes, this can only be assessed for two vertebrae located at the anterior end of the preserved dorsal series.
Prepubic process is dorsoventrally flattened along its en− tire length.-Primitively for Ornithischia, the short pre− pubic process is transversely compressed and dorsoventrally deep (Eocursor parvus, Butler et al. 2007;Heterodonto− saurus tucki, Santa Luca 1980). In some basal neornith− ischians and all basal ornithopods the process is either rod− like with a cross−section that is subcircular or slightly wider than deep (e.g., Hexinlusaurus multidens, He and Cai 1984;Hypsilophodon foxii, Galton 1974;Othnielosaurus, Galton and Jensen 1973;Parksosaurus warreni, Parks 1926), or "sword−like", being dorsoventrally shallow but transversely compressed (e.g., Orodromeus makelai, Scheetz 1999). In pachycephalosaurs the pubis is only known for Homalo− cephale calathocercos; it differs from the pubis of ornitho− pods in being dorsoventrally compressed at its base, but transversely compressed at its anterior end (Maryańska and Osmólska 1974). By contrast to all of these conditions, the prepubic process of Stenopelix appears to be dorsoventrally compressed along its entire length; Sereno (1987) considered this an autapomorphic feature, but as discussed above, this feature is also seen in a number of basal ceratopsians. This character might provide some support for a ceratopsian iden− tity for Stenopelix; however, more work on the range of pu− bic variation within Ornithischia is required-the range of variation within Pachycephalosauria, for example, is com− pletely unknown.

Taxonomic position of Stenopelix
The holotype of Stenopelix valdensis is difficult to interpret and, without the discovery of additional material, this taxon may never be phylogenetically positioned with certainty. Most features of the skeleton suggest affinities with Cera− poda, but assignment of Stenopelix to any one of the constitu− http://app.pan.pl/acta54/app54−021.pdf ent clades of Cerapoda (Ornithopoda, Pachycephalosauria, Ceratopsia) is difficult. No characters have been proposed to support referral to Ornithopoda, and we feel that such a refer− ral is unlikely, largely in light of the absence of an obturator process on the ischium and the presence of a number of fea− tures that suggest affinities with Marginocephalia. These fea− tures include: the strap−like scapula blade; the transverse ex− pansion of the dorsal margin of the ilium and the transverse expansion of the distal end of the preacetabular process in dorsal view; dorsoventral compression of the proximal part of the prepubic process; the absence of an obturator process; and the elongate pedal phalanx I−1. Within Marginocephalia evidence in favour of either a ceratopsian or a pachycephalo− saurian identity is weak, and we cannot support referral to ei− ther clade on the basis of the available evidence. Dodson (1990) suggested that Stenopelix might be a sister taxon to Marginocephalia; this is plausible, but we are unaware of any features that are shared by pachycephalosaurs and cerato− psians but absent in Stenopelix. As discussed above, recent phylogenetic analyses place Stenopelix within Marginoce− phalia, but find it to be a highly unstable wildcard taxon (But− ler et al. 2008). In light of these uncertainties we recommend that Stenopelix is best considered as ?Marginocephalia.
The early fossil record of pachycephalosaurs  Parker et al. 2005;Irmis et al. 2007) have dem− onstrated the problems inherent in determining the phylogen− etic position of isolated, ornithischian−like teeth. As also re− cognised by Sullivan (2006), there is no substantial evidence to support referral of F. adenticulatum to Pachycephalosauria; F. adenticulatum is here considered Ornithischia incertae sedis, and may represent a nomen dubium. Galton (1971) described a new taxon of pachycephalo− saur, Yaverlandia bitholus, on the basis of an isolated skull fragment from the Wessex Formation (Early Cretaceous: Barremian) of England. Although highly fragmentary, this taxon has survived many recent pachycephalosaurian taxo− nomic reviews (e.g., Maryańska 1990;Sereno 2000;Mary− ańska et al. 2004). However, a number of authors have noted that the phylogenetic position of this taxon is problematic. Sullivan (2000Sullivan ( , 2003Sullivan ( , 2006 argued that Y. bitholus lacks synapomorphies of Pachycephalosauria, and removed this taxon from the clade. Darren Naish (personal communica− tion 2005Naish (personal communica− tion , 2007; cited in Sullivan 2006) indicates that Y. bitholus possesses synapomorphies of theropods (see list of supporting synapomorphies in Sullivan 2006). The putative pachycephalosaur taxon Majungatholus atopus (Sues and Taquet 1979), from the Late Cretaceous of Madagascar, was revealed on the basis of more complete material to represent an abelisaurid theropod (Sampson et al. 1998). It is not there− fore implausible that Y. bitholus could also represent a non− ornithischian taxon.
This brief review demonstrates that there are no definite pachycephalosaurs known prior to the Late Cretaceous, with all well−verified pachycephalosaurian taxa known from the Santonian-Maastrichtian of North America and Asia (Sulli− van 2006). Assuming that recent work supporting margino− cephalian monophyly (e.g., Xu et al. 2006;Butler et al. 2008) is correct, this implies a ghost lineage of around 75 million years extending from the Oxfordian (based upon the occur− rence of the basal ceratopsian Yinlong downsi) to the base of the Santonian. Even if discovery of additional material even− tually demonstrates that Stenopelix does represent an early pachycephalosaur, this would still only represent a point oc− currence that would break−up rather than reduce the overall ghost lineage duration, which is the longest known for any dinosaurian clade. The absence of definite pre−Cretaceous pachycephalosaurs creates problems for understanding the early evolution of the clade; this is compounded by recent work which hypotheses that "flat−headed" pachycephalosaurs such as Goyocephale and Homalocephale, generally thought to represent plesiomorphic morphological intermediates be− tween domed pachycephalosaurids and other ornithischians (e.g., Sereno 1999Sereno , 2000, may in fact represent early onto− genetic growth stages of domed pachycephalosaurs such as Prenocephale (Horner et al. 2007;Sullivan 2007). If such hy− potheses are proved correct, then it will mean that we have vir− tually no insight into the sequence and tempo of construction of the pachycephalosaur body plan. Discoveries of pre−Late Cretaceous pachycephalosaurs will be critical for a deeper un− derstanding of the group's evolution. for details. In addition to all the people acknowledged in previous pachycephalosaur studies, RMS thanks James Gardner (Royal Tyrrell Museum of Palaeontology, Drumheller, Canada) for loan of the "Triebold specimen" and Amy Henrici (Carnegie Museum of Natural History, Pittsburgh, USA) for loan of the cast of the Stegoceras validum postcrania (UALP−2). Thanks to Philip Currie (University of Alberta, Edmonton, Canada) for additional access to UALP−2.