Reconstructions of the Axial Muscle Insertions in the Occipital Region of Dinosaurs: Evaluations of Past Hypotheses on Marginocephalia and Tyrannosauridae Using the Extant Phylogenetic Bracket Approach

Authors

  • Takanobu Tsuihiji

    Corresponding author
    1. Department of Geology, National Museum of Nature and Science, Shinjuku-ku, Tokyo, Japan
    • Department of Geology, National Museum of Nature and Science, 3-23-1 Hyakunin-cho, Shinjuku-ku, Tokyo 169−0073, Japan
    Search for more papers by this author
    • Fax: +81-3-3364-7104


Abstract

The insertions of the cervical axial musculature on the occiput in marginocephalian and tyrannosaurid dinosaurs have been reconstructed in several studies with a view to their functional implications. Most of the past reconstructions on marginocephalians, however, relied on the anatomy of just one clade of reptiles, Lepidosauria, and lack phylogenetic justification. In this study, these past reconstructions were evaluated using the Extant Phylogenetic Bracket approach based on the anatomy of various extant diapsids. Many muscle insertions reconstructed in this study were substantially different from those in the past studies, demonstrating the importance of phylogenetically justified inferences based on the conditions of Aves and Crocodylia for reconstructing the anatomy of non-avian dinosaurs. The present reconstructions show that axial muscle insertions were generally enlarged in derived marginocephalians, apparently correlated with expansion of their parietosquamosal shelf/frill. Several muscle insertions on the occiput in tyrannosaurids reconstructed in this study using the Extant Phylogenetic Bracket approach were also rather different from recent reconstructions based on the same, phylogenetic and parsimony-based method. Such differences are mainly due to differences in initial identifications of muscle insertion areas or different hypotheses on muscle homologies in extant diapsids. This result emphasizes the importance of accurate and detailed observations on the anatomy of extant animals as the basis for paleobiological inferences such as anatomical reconstructions and functional analyses. Anat Rec 293:1360–1386, 2010. © 2010 Wiley-Liss, Inc.

Reconstructions of muscle morphology and attachments on skeletal elements have been common practices in dinosaur paleontology mainly due to their importance in inferring biological functions in dinosaurs. Because such functions as feeding and locomotion have been analyzed extensively, there have been many studies on reconstructions on muscles of the jaws (e.g., Haas,1955,1963,1969; Ostrom,1961,1966) and appendicular systems (e.g., Romer,1923a,b; Tarsitano,1983; Norman,1986; Nicholls and Russell,1985). More recently, reconstructions of these muscular systems have been re-evaluated in explicitly phylogenetic frameworks, either by using methods proposed by Bryant and Russell (1992) and Witmer (1995) and parsimoniously reconstructing the anatomy (e.g., Dilkes,2000; Carpenter and Smith,2001; Jasinoski et al.,2006; Holliday,2009), or by analyzing sequences of character evolution (e.g., Hutchinson2001a,b; Carrano and Hutchinson,2002).

Among the muscular systems that have been reconstructed less frequently in dinosaurs is the axial musculature in the cervical region. Exceptions, however, are Marginocephalia (a clade of ornithischian dinosaurs consisting of Pachycephalosauria and Ceratopsia) and large-bodied theropods, for which insertions of this muscular system on the occipital region have been reconstructed in several studies. In Marginocephalia, such reconstructions have been made both in Pachycephalosauria (Maryańska and Osmólska,1974; Sues and Galton,1987) and in Ceratopsia (Lull,1908; Russell,1935). Among large-bodied theropods, similar reconstructions of muscle insertions have been made in tyrannosaurids, Allosaurus, and Ceratosaurus (Bakker et al.,1988; Bakker,2000; Snively and Russell,2007a,b). Although these are phylogenetically rather divergent groups of dinosaurs, both marginocephalians and large-bodied theropods are characterized by possessing large and heavy skulls. Ceratopsids have especially large skulls to the extent that certain members of this clade possess largest skulls among terrestrial vertebrates (Dodson et al.,2004). From a functional morphological point of view, therefore, a particularly intriguing question is: how did these dinosaurs support and move their heads? A clue for this question lies in the morphology of the occipital region of the skull. That is, these dinosaurs tend to have a large occipital surface of the skull that provides a potentially large area for insertions of cervical axial muscles. Marginocephalians, especially ceratopsians, show an evolutionary trend of enlargement of the parietosquamosal shelf: although this structure is shelf-like in Pachycephalosauria and Psittacosauridae, it becomes a small frill in basal neoceratopsians and further expands to be an enlarged frill in Ceratopsidae (e.g., You and Dodson,2004; Dodson et al.,2004). Although elaboration of the parietosquamosal frill in ceratopsians is primarily attributed to sexual selection (e.g., Farlow and Dodson,1975; Sampson et al.,1997; Dodson et al.,2004), expansion of this structure would also have provided increased areas for attachment of cervical axial muscles. In large theropods, especially tyrannosaurids, on the other hand, development of the nuchal crest of the parietal is pronounced (e.g., Holtz,2004), thus providing a large area for axial muscle insertions. In both groups of dinosaurs, therefore, the morphology of the occiput suggests strong development of cervical axial muscles inserting on this region, presumably correlated with great mass of their heads. Reconstructions of detailed insertion sites of various axial muscles on the occiput, therefore, serve as a first, necessary step for understanding the morphology and function of the anatomical system for carrying their massive heads.

Past reconstructions of axial muscle insertions on the occiput in marginocephalians and large-bodied theropods, however, not without a problem. First, most such reconstructions in Marginocephalia (Fig. 1) relied on the anatomy of just one clade of reptiles, usually that of lepidosaurians. Lull (1908) used the chameleon as a model to reconstruct the cervical and jaw muscles of the ceratopsid Triceratops (T. horridus and T. prorsus). Maryańska and Osmólska (1974) reconstructed insertions of the axial musculature in the occipital region of the pachycephalosaurian Prenocephale prenes based on published accounts on the anatomy of lepidosaurians by Oelrich (1956) and Ostrom (1961). Following Maryańska and Osmólska (1974), Sues and Galton (1987) made a similar reconstruction in another pachycephalosaurian, Stegoceras validum, based on the lepidosaurian condition. Russell (1935) is a possible exception to this trend in that his reconstruction of axial muscle insertions in the ceratopsid Chasmosaurus belli was based on the anatomy of Sphenodon punctatus as well as on a published account on the anatomy of the raven. With the exception of Russell (1935), therefore, these authors did not use the information of the muscular anatomy of the closest extant relatives of Marginocephalia, that is, Aves and Crocodylia, and thus their reconstructions lack phylogenetic support.

Figure 1.

Past reconstructions of muscle attachments on the occiputs in marginocephalians in posterior view. (A) Prenocephale prenes, modified from Maryańska and Osmólska (1974; Fig. 6); (B) Stegoceras validum, modified from Sues and Galton (1987; Fig. 6); (C) Triceratops horridus, modified from Lull (1908; pl. III); (D) Chasmosaurus belli, modified from Russell (1935; Fig. 7). Original illustrations reproduced courtesy of Palaeontologia Polonica (A), E. Schweizerbart'sche Verlagsbuchhandlung (B), American Journal of Science (C), and the Canadian Museum of Nature, Ottawa, Canada (D). Abbreviation: oc, occipital condyle.

In large-bodied theropods, in contrast, axial muscle insertions on the occiput have been inferred based on the anatomy of crocodylians (Bakker,2000) or on that of both a crocodylian (American alligator) and a bird (ostrich; Bakker et al.,1988). Recently, Snively and Russell (2007a,b) reconstructed the anatomy of cervical muscles, including their insertions on the occiput, in these dinosaurs by applying phylogenetic, parsimony-based methods of Bryant and Russell (1992) and Witmer (1995) mainly based on the anatomy of crocodylians and birds. In terms of the methodology, therefore, reconstructions by Snively and Russell (2007a,b) represent phylogenetically justified inferences. Muscle insertions reconstructed by Snively and Russell (2007a,b), however, appear to be inconsistent with those observed in extant diapsids such as the ones illustrated in Tsuihiji (2005,2007), thus requiring further testing.

In this study, past reconstructions of insertions of cervical axial muscles in the occiputs of marginocephalians and tyrannosaurids are evaluated through rigorous identification of such insertions based on the anatomy of birds, crocodylians, and lepidosaurians examined in detail in my previous studies (Tsuihiji,2005,2007). The basically same methodology as the one used by Snively and Russell (2007a,b) was used, that is, phylogenetic, parsimony-based methods of Bryant and Russell (1992) and, more specifically, the Extant Phylogenetic Bracket (EPB) approach of Witmer (1995). The present application of this methodology based on Tsuihiji's (2005,2007) anatomical data resulted in new reconstructions of muscle insertions in the occiputs of these dinosaurs. Interestingly, reconstructions proposed in this study turned out to be rather different even from the results of the most recent studies by Snively and Russell (2007a,b) on tyrannosaurids. Such differences between the present and previous studies are discussed in detail.

MATERIALS AND METHODS

Data on insertions of axial muscles on the occipital region of the skull in extant diapsid taxa were mainly derived from my previous studies (Tsuihiji,2005,2007), in which four species of lepidosaurians (Sphenodon punctatus, Iguana iguana, Varanus exanthematicus, and V. salvadorii), three species of crocodylians (Alligator mississippiensis, Caiman crocodilus, and Osteolaemus tetraspis), and five species of birds (Struthio camelus, Rhea americana, Meleagris gallopavo, Gallus gallus, and Pygoscelis sp.) were examined. For this study, additional observations were made on some of these specimens. In addition, two additional palaeognath birds, the tinamid Nothura maculosa [YPM 114797 (Peabody Museum of Natural History, Yale University, New Haven, CT)] and casuariid Casuarius casuarius [KPM-NF 2001025 (Kanagawa Prefectural Museum of Natural History, Odawara, Kanagwa, Japan)], were dissected and insertions of the axial muscles on the occiput were examined. In addition to axial muscles, attachments of m. depressor mandibulae and the m. cucullaris complex on the occiput were also examined because of their close topological relationships with the insertions of axial muscles.

Homologies and synonymies of cervical axial muscles among diapsid clades follow those proposed by Tsuihiji (2005,2007) and are summarized in Table 1. Different nomenclatural schemes have been used for muscle names among lepidosaurians, crocodylians, and birds. In the following description and discussion, the lepidosaurian terminology is used, mainly based on that of Nishi (1916), to refer to muscles in all diapsids clades unless noted otherwise. That is, crocodylian and avian homologs of a lepidosaurian muscle are referred to by the name of the lepidosaurian muscle (Table 1). For example, the avian m. biventer cervicis and medial part of the crocodylian m. transversospinalis capitis are both called m. spinalis capitis in this study because the former avian and crocodylian muscles were hypothesized to be homologous with the latter lepidosaurian muscle in Tsuihiji (2005). Past studies by Maryańska and Osmólska (1974) and Sues and Galton (1987) used the lepidosaurian terminology in reconstructing muscle insertions in marginocephalians. Using the same terminology, therefore, facilitates comparison between the result of this study and these past reconstructions. One exception, however, is a subvertebral muscle inserting on the paroccipital process. Because this muscle is absent in Lepidosauria, the avian name m. rectus capitis lateralis is used to refer to this muscle (Table 1).

Table 1. Homologies of muscles attaching to the occiput in extant diapsids based on Tsuihiji (2005,2007)
LepidosauriaCrocodyliaAves
  1. Muscle names in bold are those used in this study.

Epaxial musculature
 M. rectus capitis posteriorM. atloïdo-capitisM. splenius capitis, medial part
 M. obliquus capitis magnusM. epistropheo-capitisM. splenius capitis, lateral part
 M. spinalis capitisM. transversospinalis capitis, medial partM. biventer cervicis (+ m. longus  colli dorsalis, pars caudalis  inserting on cervical vertebrae)
 M. longissimus capitis, pars  articuloparietalisM. transversospinalis capitis, lateral partM. complexus
 M. longissimus capitis,pars  transversalis capitisM. longissimus capitis superficialis(Absent)
 M. longissimus capitis, pars  transversalis cervicisM. longissimus capitis profundusPart of m. rectus capitis dorsalis
 M. iliocostalis capitisPart of “m. iliocostalis capitisPart of m. rectus capitis dorsalis
Hypaxial musculature
 M. rectus capitis anteriorM. rectus capitis anticus majorM. rectus capitis ventralis
 Part of “m. iliocostalis capitisM. rectus capitis lateralis
Other muscles attaching to the occiput
 M. depressor mandibulaeM. depressor mandibulaeM. depressor mandibulae
 M. episternocleidomastoideusPart of “m. iliocostalis capitisM. capitisternalis

Data on extant diapsids, especially those on Crocodylia and Aves, served as the basis for applying the EPB approach for reconstructing muscle attachments on the occiputs of non-avian dinosaurs. Attachments of muscles in extant diapsids discussed in this study are summarized in Table 2. Such a muscle attachment was reconstructed in fossil specimens if homologous muscles have similar attachments between Crocodylia and Aves, either with or without the presence of a clear osteological correlate such as a depression, process, or rugosity characterizing muscle attachment on the bone surface in these dinosaurs (Level I or I′ inferences of Witmer,1995, respectively). That is, muscle attachments were safely inferred when their osteological correlates were identified in fossil specimens (Level I inferences). In addition, even when definite osteological correlates were not present, such attachments were still inferred if their topological relationships with the bone elements or with attachments of other muscles were conserved between Crocodylia and Aves (Level I′ inferences). On the other hand, when homologous muscles inserted on different areas between Crocodylia and Aves, or did not insert on the occiput in one of these clades, the condition (i.e., the position or presence/absence of the insertion) in marginocephalians or tyrannosaurids was equivocal, leading to Level II or II′ inferences dependent on the presence or absence of an osteological correlate. In such a case, an additional outgroup, Lepidosauria, was used to parsimoniously determine the condition of the most recent common ancestor of Aves and these dinosaurs. If the assessment at this node was decisive, such a condition was reconstructed in these dinosaurs. If the assessment at this node was equivocal, however, the muscle attachment was either not reconstructed or only tentatively reconstructed using the condition in one of the extant archosaurian clades as a model. Levels of inference of muscle attachments discussed in this study are summarized in Table 2.

Table 2. Muscle attachments on the occiput in extant diapsids based on Tsuihiji (2005,2007) and their reconstructions, osteological correlates, and levels of inference in non-avian dinosaurs
 Muscle attachments in extant diapsidsInferences in non-avian dinosaurs
LepidosauriaCrocodyliaAvesMuscle attachmentsOsteological correlates of muscle attachmentsLevels of inference
  • a

    The insertions of m. longissimus capitis, pars transversalis cervicis and m. iliocostalis capitis are not distinguished in non-avian dinosaurs because these muscles are merged in Aves.

  • b

    M. iliocostalis capitis, m. rectus capitis lateralis, and m. episternocleidomastoideus are merged in one muscle slip in Crocodylia.

Epaxial musculature
 M. rectus capitis  posteriorPosterior surface of the supraoccipitalPosterior surfaces of the supraoccipital and paroccipital processPosterior surface of the supraoccipitalPosterior surface of the supraoccipitalDepression on the supraoccipitalLevel I or I′
 M. obliquus  capitis magnusDorsal and posterior surfaces of the paroccipital processPosterior surface of the lateral part of the paroccipital processPosterior surfaces of the supraoccipital and/or paroccipital processPosterior surface of the paroccipital processDepression on the paroccipital processLevel I or I′
 M. spinalis  capitisDorsal end of the supraoccipital and adjacent surface of the parietalDorsal end of the supraoccipital and dorsomedial corner of the otoccipitalPosterior surface of the supraoccipital or parietalDorsal end of the supraoccipital and adjacent area of the parietalRugosity or process on the dorsal end of the supraoccipital; scar or depression on the parietalLevel I or I′
 M. longissimus capitis,  pars articuloparietalisPosterior surface of the parietalDorsal end of the supraoccipital and dorsomedial corner of the otoccipitalPosterior surface of the parietalPosterior surface of the parietal (also extending onto the squamosal in marginocephalians)Depressions on the parietal and/or squamosalLevel I or I′
 M. longissimus  capitis,pars  transversalis capitisDistal end and ventral edge of the paroccipital processDistal end of the paroccipital process(Absent)(Not reconstructed)(No unambiguous osteological correlate present)Level II′
 M. longissimus capitis,  pars transversalis  cervicisPosterior surface of the basal tuberaPosterior surface of the basal tuberaPosterior surface of the basal tuberaPosterior surface of the basal tuberaaScar along the distal margin of the basal tuberaaLevel I
 M. iliocostalis  capitisPosterior surface of the basal tuberaDistal end and ventral edge of the paroccipital processbPosterior surface of the basal tuberaPosterior surface of the basal tuberaaScar along the distal margin of the basal tuberaaLevel II
Hypaxial musculature
 M. rectus capitis  anteriorPosterior surface of the basal tuberaPosterior surface of the basal tuberaVentral surface of the basitemporal platePosterior surface of the basal tuberaScar along or close to the midline of the basal tuberaLevel I
 M. rectus capitis  lateralis(Absent)Distal end and ventral edge of the paroccipital processbDistal end of the paroccipital processDistal end of the paroccipital processScar on the distal end of the paroccipital processLevel I or I′
Other muscles attaching to the occiput
 M. depressor  mandibulaePosterior surfaces of the squamosal, parietal, and/or paroccipital processDistal end of the paroccipital process and posterior and lateral aspects of the squamosalPosterior surfaces of the squamosal and/or parietal, and distal end of the paroccipital processDistal end of the paroccipital process and posterior surface of the squamosal (also extending to the lateral part of the parietal in tyrannosaurids)Scar or depression on the squamosal (also tab-like lateral expansion of the parietal in tyrannosaurids)Level I or I′
 M. episternocleido-  mastoideusDistal end of the paroccipital process and posterior surfaces of the parietal and/or squamosalDistal end and ventral edge of the paroccipital processbPosterior surfaces of the parietal, squamosal, temporal membrane, or postorbital process of the frontalDistal end of the paroccipital process, posterior surface of the parietal, or posterior surface of the squamosal (indeterminate)(No unambiguous osteological correlate present)Level II′

The following museum specimens of non-avian dinosaurs were examined for reconstructing muscle attachments. In addition to marginocephalians and tyrannosaurids, specimens of some basal ornithopods were also studied as immediate outgroups of Marginocephalia for comparison with the conditions in this clade: Pachycephalosauria—Stegoceras validum [YPM 3742, cast of UA 2 (University of Alberta, Edmonton, Alberta, Canada)], Homalocephale calathocercos [MPC-D 100/1201 (Mongolian Paleontological Center, Ulaanbaatar, Mongolia)], Dracorex hogwartsia [TCMI 2004.17.1 (The Children's Museum of Indianapolis, Indianapolis, IN)], and Stygimoloch spinifer [MPM 8111 (Milwaukee Public Museum, Milwaukee, WI) and YPM 56085, cast of UCMP 119433 (University of California Museum of Paleontology, Berkley, CA)]; Ceratopsia—Psittacosaurus mongoliensis [IGM 100/1032 (Institute of Geology, Ulaanbaatar, Mongolia) and AMNH 6254 (American Museum of Natural History, New York, NY)], Protoceratops andrewsi [AMNH 6429 and 6466, and Hayashibara Museum of Natural Sciences (HMNS, Okayama, Japan) uncataloged cast of MPC-D 100/533], Chasmosaurus belli [YPM 2016 and BMNH R 4448 (Natural History Museum, London, UK)], Triceratops horridus (YPM 1820 and 1823), and Triceratops prorsus [BSP 1964 I 456 (Bayerischen Staatssammlung für Paläontologie und Historische Geologie, Munich, Germany)]; Basal ornithopods—Hypsilophodon foxii (BMNH R 197), Zephyrosaurus schaffi [MCZ 4392 (Museum of Comparative Zoology, Harvard University, Cambridge, MA)], and Tenontosaurus tilletti [OMNH 10132 (Sam Noble Oklahoma Museum of Natural History, University of Oklahoma, Norman, OK)]; Tyrannosauridae—Gorgosaurus sp. (TCMI 2001.89.1), Daspletosaurus torosus [CMN 8506 (Canadian Museum of Nature, Ottawa, Ontario, Canada)], Daspletosaurus sp. [MOR 590 (Museum of the Rockies, Montana State University, Bozeman, MT)], Tarbosaurus bataar (MPC-D 107/14), and Tyrannosaurus rex (MOR 555). Figure 2 depicts phylogenetic relationships of these examined taxa.

Figure 2.

Non-avian dinosaurian taxa examined in this study with a cladogram depicting phylogenetic relationships of these taxa and extant bracketing taxa. The topology of the cladogram is based on Sereno (1986), Dodson et al. (2004), Norman (2004), Norman et al. (2004), You and Dodson (2004), Carr and Williamson (2010), and Longrich et al. (2010).

Recent studies by Horner and Goodwin (2009) proposed that two pachycephalosaurians examined in this study, Dracorex hogwartsia and Stygimoloch spinifer, represent younger growth stages of Pachycephalosaurus wyomingensis and thus synonymized those two taxa with the latter. The morphology of the occipital region, however, is fairly different between the examined specimens of D. hogwartsia and S. spinifer. Accordingly, these two taxa are treated separately in this study.

RESULTS

Attachments of axial muscles, as well as those of m. depressor mandibulae and the m. cucullaris complex, in the occiput in extant diapsids are shown in Fig. 3 and summarized in Table 2. Based on these data, attachments of these muscles in marginocephalians and tyrannosaurids are inferred and discussed. Descriptions of the general morphology of these muscles are based on Tsuihiji (2005,2007) in the following discussion.

Figure 3.

Muscle attachments on the occiputs in extant diapsids in posterior view (partly modified from Tsuihiji,2005,2007). Attachments of homologous muscles are represented by identical colors and are labeled with lepidosaurian, crocodylian, and avian names. In Aves, m. longissimus capitis, pars transversalis cervicis and m. iliocostalis capitis comprise a single muscle, m. rectus capitis dorsalis. The crocodylian “m. iliocostalis capitis” is also a composite muscle consisting of m. rectus capitis lateralis (as shown in the same color in this figure), m. episternocleidomastoideus, and m. iliocostalis capitis. In Casuarius casuarius, the boundary between m. rectus capitis posterior and m. obliquus capitis magnus is not clear, and muscle attachments on the basal tubera and basitemporal plate are not shown. Not to scale. Abbreviations: fm, foramen magnum; lat, lateral part; med, medial part; oc, occipital condyle.

Muscle Attachments in the Occiput in Marginocephalia

M. rectus capitis posterior (Level I inference).

This is a short suboccipital muscle that arises from the neural arch of the atlas as well as from the neural spine of the axis. It inserts on the supraoccipital in extant diapsids, and the insertion extends onto the paroccipital process in Alligator mississippiensis and some birds (Fig. 3). A shallow depression or concavity characterizes the insertion of this muscle in extant taxa.

In basal ornithopods such as Hypsilophodon foxii (BMNH R 197), there is a pair of shallow concavities on the supraoccipital marking the insertion of this muscle (Fig. 4A–C). Similar depressions in the supraoccipital are commonly seen in marginocephalians. In pachycephalosaurians such as Stegoceras validum (YPM 3742, cast of UA 2) and Homalocephale calathocercos (MPC-D 100/1201), a large depression lies on the supraoccipital and parietal, lateral to the median ridge for the supraspinal ligament (Fig. 4D,E,G). The ventral part of this depression on the supraoccipital is interpreted as the insertion of m. rectus capitis posterior, whereas the dorsal part of this depression is considered as the insertion of m. spinalis capitis (Fig. 4D,E,G; see below). In these pachycephalosaurian specimens, there is no separation of the insertion of m. rectus capitis posterior into two parts on each side unlike in Prenocephale prenes discussed by Maryańska and Osmólska (1974; Fig. 1A). In H. calathocercos, the ventral part of this depression is bounded laterally by a low ridge, presumably separating the insertion of this muscle from that of m. obliquus capitis magnus (Fig. 4G). In Stygimoloch spinifer (MPM 8111), on the other hand, the same area on the supraoccipital and parietal lateral to the median ridge is occupied by a deep, dorsoventrally elongated depression, which extends laterally onto the proximal part of the otoccipital above the foramen magnum (Fig. 5D). Within this depression, a more pronounced, deeper concavity extends from the parietal to the dorsal part of the supraoccipital along the median ridge, demarcated from a shallower, ventrolateral part that lies mainly on the supraoccipital but also extends to the parietal and otoccipital. The latter, shallower part is interpreted as the insertion of m. rectus capitis posterior whereas the deeper concavity is considered as the insertion of the medial part of m. spinalis capitis as discussed below (Fig. 5C,D).

Figure 4.

Occipital regions of select basal ornithopods (AC) and pachycephalosaurians (DG). (A–E) Muscle attachments reconstructed in posterior view. The insertion of m. episternocleidomastoideus indicated by “?” is only tentatively reconstructed on the distal end of the paroccipital process using the crocodylian condition as a model. (A) Hypsilophodon foxii (BMNH R 197); (B) Zephyrosaurus schaffi (MCZ 4392), left side of the occiput (reversed) with an enlarged view of the parietal showing the separation between the insertions of m. spinalis capitis and m. longissimus capitis,pars articuloparietalis; (C) Tenontosaurus tilletti (OMNH 10132) with the squamosals reconstructed; (D) Stegoceras validum (YPM 3742, cast of UA 2); (E) Homalocephale calathocercos (MPC-D 100/1201). (F) Lateral view of the squamosal and paroccipital process, and (G) posterior view of the right half of the occiput of H. calathocercos, showing several osteological correlates of muscle attachments (1–4). (1) Scar on the lateral edge of the squamosal representing the origin of m. depressor mandibulae; (2) depression on the paroccipital process representing the insertion of m. obliquus capitis magnus; (3) depression on the supraoccipital and parietal representing the insertions of m. rectus capitis posterior and m. spinalis capitis; and (4) tuberosity on the dorsal end of the supraoccipital representing the ventral-most part of the insertion of m. spinalis capitis. Not to scale. Abbreviations: fm, foramen magnum; oc, occipital condyle; p, parietal; parocc, paroccipital process; so, supraoccipital; sq, squamosal.

Figure 5.

Occipital regions of pachycephalosaurians, Dracorex hogwartsia (TCMI 2004.17.1, reversed, A and B) and Stygimoloch spinifer (MPM 8111, C and D). (A, C) Muscle attachments reconstructed in posteroventral view. The insertion of m. episternocleidomastoideus indicated by “?” is only tentatively reconstructed on the distal end of the paroccipital process using the crocodylian condition as a model. (B, D) Enlarged views of several osteological correlates of muscle attachments (1–4) in posteroventral view. (1) Rugosity in a depression on the dorsal end of the supraoccipital representing the ventral-most part of the insertion of m. spinalis capitis; (2) scar on the parietal representing the dorsal part of the insertion of m. spinalis capitis; (3) and (4) shallow and deep concavities representing the insertions of m. rectus capitis posterior and m. spinalis capitis, respectively, within a dorsoventrally elongated depression on each side of the median ridge on the supraoccipital and parietal. Not to scale. Abbreviations: fm, foramen magnum; oc, occipital condyle; oo, otoccipital; p, parietal; parocc, paroccipital process; so, supraoccipital; sq, squamosal.

In the basal ceratopsian Psittacosaurus mongoliensis (IGM 100/1032), the supraoccipital has a well-developed median ridge for the attachment of the supraspinal ligament dorsal to the foramen magnum (Fig. 6A–D). On each side of this ridge lies a deep depression that extends to the ventral part of the parietal and proximal part of the otoccipital/paroccipital process (Fig. 6C,D). This depression is interpreted here as the insertion of m. rectus capitis posterior (Fig. 6A,B) although it may also have contained a venous sinus (see below). In Protoceratops andrewsi (AMNH 6429 and 6466, and HMNS uncataloged cast of MPC-D 100/533), in contrast, the supraoccipital bears only weak depressions for the insertion of this muscle (Fig. 6E).

Figure 6.

Occipital regions of basal ceratopsians, Psittacosaurus mongoliensis (IGM 100/1032, reversed, AD) and Protoceratops andrewsi (HMNS uncataloged cast of MPC-D 100/533, E). (A, B, E) Muscle attachments reconstructed in posterior (A, E) and posteroventral (B) views. The insertion of m. episternocleidomastoideus indicated by “?” is only tentatively reconstructed on the distal end of the paroccipital process using the crocodylian condition as a model. In (E), a tab-like projection on the ventral margin of the parietal fenestra is regarded as one of the two sites of insertion of m. longissimus capitis, pars articuloparietalis (as reconstructed with “?”). This projection, however, may instead have served as the insertion of m. episternocleidomastoideus if the distal end of the paroccipital process was not the insertion of this muscle. (C, D) Several osteological correlates of muscle attachments (1–4) in posterior (C) and oblique posteroventral (D) views. (1) Depression on each side of the median ridge on the supraoccipital representing the insertion of m. rectus capitis posterior; (2) depression on the dorsal part of the paroccipital process and ventral part of the parietal putatively representing a second insertion of m. rectus capitis posterior. It is also possible, however, that this depression was instead occupied by a venous sinus; (3) tuberosity on the dorsal end of the supraoccipital representing the ventral-most part of the insertion of m. spinalis capitis; and (4) large depression on the parietal representing the insertions of m. longissimus capitis, pars articuloparietalis and the dorsal part of m. spinalis capitis. Not to scale. Abbreviations: fm, foramen magnum; oc, occipital condyle; p, parietal; parocc, paroccipital process; so, supraoccipital; sq, squamosal.

Ceratopsids, derived members of ceratopsians, tend to have well-developed depressions on the supraoccipital. Not surprisingly, these depressions were proposed as muscle attachments in the past reconstructions (Fig. 1C,D). In Triceratops, Lull (1908) described “two deep depressions separated by a thin lamina of bone” in the supraoccipital (p 391) and interpreted them as the insertions of the “complexus major,” which presumably corresponds to m. longissimus capitis, pars articuloparietalis in this study. In Chasmosaurus belli, Russell (1935) found an “ovoid depression” above the foramen magnum and interpreted this entire depression as the insertion of m. biventer cervicis, which is an avian homolog of m. spinalis capitis (Tsuihiji,2005; Table 1). Comparison with the conditions in extant diapsids suggests that these depressions in Triceratops and Chasmosaurus are the insertions of m. rectus capitis posterior instead (Figs. 7 and 8). Details of the morphology of these depressions, however, merit further consideration. In Triceratops horridus (YPM 1820), a pair of dorsoventrally elongated depressions lies dorsal to the foramen magnum and is separated from each other by a median ridge (Fig. 7C) as Lull (1908) described. The ventral part of each depression is a deep fossa lying mainly on the supraoccipital. A shallower depression extends dorsally from this fossa onto the ventral part of the parietal and appears to expand mediolaterally at its dorsal end. In Chasmosaurus belli (YPM 2016), Russell's (1935) “ovoid depression” is a large, median concavity on the supraoccipital and ventral end of the parietal. This concavity contains a pair of deep fossae ventrally and another pair of round, shallow depressions with scarring dorsally (Fig. 8B). Recently, Witmer and Ridgely (2008) showed that the caudal middle cerebral veins exit in a pair of deep fossae on the supraoccipital in the ceratopsid Pachyrhinosaurus lakustai. This suggests that these deep fossae may have been occupied by a venous sinus in this and other ceratopsians. If this is the case, the insertions of m. rectus capitis posterior in T. horridus and C. belli are likely represented by shallow depressions lying dorsal to these deep fossae. It is noteworthy, however, that in extant birds a groove through which the external occipital vein similarly enters the endocranial cavity in the occipital region is covered by a membrane onto which the insertion of the m. rectus capitis posterior or m. obliquus capitis magnus extends (personal observations on Struthio camelus, Ardea cinera, and other birds), showing that a muscle attachment can coexist with a venous structure. Accordingly, it is also possible that the insertions of m. rectus capitis posterior extend ventrally onto these deep fossae in ceratopsids by covering the venous sinus.

Figure 7.

Occipital regions of the ceratopsid Triceratops, T. prorsus [BSP 1964 I 456, (A)] and T. horridus [YPM 1823, (B) and (D); YPM 1820, reversed, (C)]. (A, B) Muscle attachments reconstructed in posterior (A) and posteroventral (B) views. The insertion of m. episternocleidomastoideus on the distal end of the paroccipital process indicated by “?” is only tentatively reconstructed using the crocodylian condition as a model. An alternative position of the insertion of this muscle is the lateral part of the parietal, which is regarded as a part of the insertion of m. longissimus capitis, pars articuloparietalis (as reconstructed with “?”) in the present figure. (C, D) Several osteological correlates of muscle attachments (1–9) in posterior (C) and posteroventral (D) views. (1) and (2) Depressions representing the insertions of m. rectus capitis posterior. The ventral part of (1) is a deep fossa that may have been occupied by a venous sinus in life; (3) depression on the paroccipital process representing the insertion of m. obliquus capitis magnus; (4) rugosity on the distal end of the paroccipital process representing the insertion of m. rectus capitis lateralis; (5) rugosity on the squamosal representing the origin of m. depressor mandibulae; (6) depression on the lateral part of the parietal representing the insertion of m. longissimus capitis, pars articuloparietalis (or alternatively representing the insertion of m. episternocleidomastoideus); (7) large depression on the squamosal representing the insertion of m. longissimus capitis, pars articuloparietalis; (8) depression on the squamosal representing the origin of m. depressor mandibulae; and (9) pair of shallow depressions and a low ridge in between representing the insertions of m. spinalis capitis and the supraspinal ligament, respectively. Not to scale. Abbreviations: fm, foramen magnum; oc, occipital condyle; p, parietal; parocc, paroccipital process; sq, squamosal.

Figure 8.

Occipital region of the ceratopsid Chasmosaurus belli (YPM 2016). (A) Muscle attachments reconstructed in posterior view; and (B) enlarged view of the central part of the occiput showing several osteological correlates of muscle attachments (1–4). (1) Scar and concavity representing the insertion of m. spinalis capitis; and (2) shallow depression with scarring representing the insertion of m. rectus capitis posterior. Ventral to this depression lies a pair of deep fossae, which would have been occupied by a venous sinus; (3) and (4) depressions on the parietal and paroccipital process, respectively, representing a second insertion of m. rectus capitis posterior. Not to scale. Abbreviations: oc, occipital condyle; p, parietal; parocc, paroccipital process; sq, squamosal.

In addition to the depression on the supraoccipital described above, many ceratopsians (except for Protoceratops andrewsi among the taxa examined in this study) have a second depression lying lateral to it on the dorsal part of the paroccipital process and ventral part of the parietal (Figs. 6, 7, 8, 6–8). In several extant diapsid taxa, including Sphenodon punctatus and especially crocodylians, the insertion of m. rectus capitis posterior expands laterally onto the dorsal/posterior surface of the paroccipital process (Fig. 3), and sometimes this muscle further divides into two distinct slips. Therefore, it is likely that this second depression in ceratopsians marks the insertion of such a lateral slip of this muscle, as reconstructed in Figs. 6, 7, 8, 6 through 8. Alternatively, it is also possible that a vascular sinus occupied this depression. Lull (1908) reconstructed this depression in Triceratops as the insertion of m. rectus capitis posterior (his “rectus capitis posticus major”) as in this study (Fig. 1C). Russell (1935) recognized two, dorsal and ventral depressions in this area in Chasmosaurus belli. He regarded the dorsal depression on the parietal as the insertion of the “longissimus cervicocapitis” (presumably corresponding to m. longissimus capitis, pars articuloparietalis in this study) and the ventral one on the paroccipital process as the insertion of the “rectus capitis posticus major” (m. rectus capitis posterior in this study; Fig. 1D). In C. belli (YPM 2016) that was examined in the present study, these two depressions are continuous on the suture line between the parietal and paroccipital process (Fig. 8B) and apparently corresponds to the single depression in Triceratops (Fig. 7C). Accordingly, these two depressions are here regarded as together representing the insertion of the lateral slip of m. rectus capitis posterior (Fig. 8A,B).

M. obliquus capitis magnus (Level I or I′ inference).

This is another, short suboccipital muscle that arises mainly from the neural spine of the axis. This muscle almost invariably inserts on the posterior surface of the paroccipital process in extant diapsid taxa (Fig. 3), enabling a parsimonious inference that the same area would have been the insertion of this muscle in fossil archosaurians, even when definite muscle scars are absent. In past reconstructions, the insertion of this muscle in pachycephalosaurians was inferred to be the depression adjacent to the foramen magnum and below the strong, horizontal ridge on the paroccipital process (Maryańska and Osmólska,1974; Sues and Galton,1987; Fig. 1A,B). In extant crocodylians and birds, however, the area surrounding the foramen magnum is the attachment of the atlanto-occipital capsular membrane/ligament (Fig. 9). Thus, it is suggested here that at least the proximal part of this depression in pachycephalosaurians marks the attachment of such a membrane/ligament (Fig. 9). It is more likely that m. obliquus capitis magnus would have inserted on the posterior surface of the paroccipital process above the ridge and depression (Figs. 4D–G and 5A). In Homalocephale calathocercos (MPC-D 100/1201), for example, the putative insertion of this muscle is a mediolaterally elongated depression demarcated medially by a low ridge, separated from the putative insertion of m. rectus capitis posterior as mentioned above (Fig. 4E–G). Lateral to this depression, the paroccipital process expands dorsoventrally. The insertion of m. obliquus capitis magnus may have been expanded onto this part of the paroccipital process as well, as reconstructed in Figure 4E. In ceratopsians, a mediolaterally expanded paroccipital process suggests strong development of this muscle. In Psittacosaurus mongoliensis (IGM100/1032), a flat, posterior surface of the paroccipital process would presumably have served as the insertion of this muscle (Fig. 6A,B). Protoceratops andrewsi (HMNS uncataloged cast of MPC-D 100/533) has a long, horizontally elongated depression on the paroccipital process, marking the insertion of this muscle (Fig. 6E). A large depression that is expanded dorsoventrally is present on the paroccipital process of Triceratops horridus (YPM 1820 and 1823) and Triceratops prorsus (BSP 1964 I 456), suggesting massive development of this muscle (Fig. 7A–C).

Figure 9.

Attachments of the atlanto-occipital capsular membrane and supraspinal ligament in the occipital region of extant archosaurians and their reconstructions in marginocephalians in posterior view. Not to scale. Abbreviations: fm, foramen magnum; oc, occipital condyle.

Extant crocodylians have a muscle called m. spino-capitis posticus that inserts on the paroccipital process, lateral to the insertion of m. obliquus capitis magnus (Fig. 3). This muscle, hypothesized either as differentiated from the suboccipital muscles (Vallois,1922) or as a part of m. spinalis cervicis homolog (Tsuihiji,2005), is unique and apomorphic to Crocodylia among extant diapsids. Accordingly, it is hypothesized here that this muscle was lacking in non-avian dinosaurs including marginocephalians.

M. spinalis capitis and m. longissimus capitis, pars articuloparietalis (Level I or I′ inference).

These are the epaxial muscles that lie most dorsally and/or superficially in the neck and insert on the skull (Fig. 3). These muscles are dealt together here because they are partially merge together at their insertions on the occiput in Lepidosauria, and form a single muscle mass (m. transversospinalis capitis) in Crocodylia. The more medially lying m. spinalis capitis arises from neural spines of cervical vertebrae and anterior dorsal vertebrae in Lepidosauria and Crocodylia and from neural spines at around the cervicodorsal boundary in Aves. On the other hand, m. longissimus capitis,pars articuloparietalis arises from zygapophyseal capsules and lateral aspects of prezygapophyses of cervical and anterior dorsal vertebrae in Lepidosauria and Crocodylia and from lateral surfaces of prezygapophyses and epipophyses of anterior cervical vertebrae in Aves.

In extant diapsids except for crocodylians, at least one of these two muscles inserts on the parietal in addition to the supraoccipital (Fig. 3). In Lepidosauria, m. spinalis capitis inserts on the parietal only (e.g., Sphenodon punctatus) or on the parietal and supraoccipital (e.g., Iguana iguana). The insertion of m. longissimus capitis,pars articuloparietalis is the parietal and squamosal in S. punctatus, but is restricted to the parietal only in I. iguana. In Aves, m. spinalis capitis inserts on the supraoccipital, particularly on the dorsal end of the bone along the crista nuchalis transversa in many birds, whereas m. longissimus capitis,pars articuloparietalis inserts on the occiput along crista nuchalis transversa, which marks the boundary between the supraoccipital and parietal (Baumel and Witmer,1993). Fusion of cranial bones makes it difficult to identify exactly which bone serves as the insertion site for m. longissimus capitis,pars articuloparietalis in adult birds. However, it was confirmed that the insertion of this muscle is on the parietal in an embryonic Struthio camelus in which suture lines among cranial bones are visible (Tsuihiji,2005; Fig. 3).

In Crocodylia, on the other hand, the insertions of m. spinalis capitis and m. longissimus capitis,pars articuloparietalis are restricted only to the supraoccipital (Fig. 3). The exposure of the parietal on the posterior surface of the skull is minimal in Crocodylia except for a small area to which the skin attaches. This is in contrast to the conditions in lepidosaurians and some birds, in which the posterior exposure of the parietal is fairly large. In fact, in Archosauromorpha on the line toward Crocodylia, the parietal is also more or less exposed on the posterior surface of the skull, making it available for insertion of one or both of these muscles (Fig. 10). Thus, the posterior exposure of the parietal appears to be a plesiomorphy for Archosauria. Crocodylia is an exception within this clade in that very little parietal surface is exposed posteriorly. Combined with the observation that m. spinalis capitis, m. longissimus capitis,pars articuloparietalis, or both of these muscles insert on the parietal in other extant diapsids whenever this bone is available for their insertion, this suggests that the crocodylian condition in which the insertions of these muscles are restricted to the supraoccipital may be a modification specific to this clade within Diapsida due to the lack of the posterior exposure of the parietal, and that the presence of the insertion of at least one of these muscles on the parietal is a plesiomorphy for Diapsida. This justifies a parsimonious inference that the posterior surface of the parietal in fossil dinosaurs including marginocephalians would also have served as the insertions for these muscles.

Figure 10.

The parietal and postparietal in Archosauromorpha in posterior view. Note that at least a part of the parietal in these taxa is exposed posteriorly and would have served as the insertions of m. spinalis capitis and/or m. longissimus capitis, pars articuloparietalis. The topology of the cladogram is based on Gauthier (1994) and Benton (2004). Not to scale. Institutional abbreviations: GPIT, Geologisch-Paläontologisches Institut, Universität Tübingen, Germany; SMNS, Staatliches Museum für Naturkunde, Stuttgart, Germany.

In some basal ornithopods, the parietal has a fairly large surface exposed posteriorly. In the parietal of Zephyrosaurus schaffi (MCZ 4392), for example, the possible insertions of m. spinalis capitis and m. longissimus capitis,pars articuloparietalis are separated from each other by a ridge (Fig. 4B). In marginocephalians, on the other hand, the posterior surfaces of the parietal and squamosal form a continuous surface for muscle attachments, suggesting that the insertions of m. spinalis capitis and m. longissimus capitis,pars articuloparietalis would likely have expanded from the parietal onto the squamosal (Figs. 4, 5, 6, 7, 8, 4–8) unlike in most extant diapsids (other than Sphenodon punctatus), in which the insertions of these muscles are restricted to the supraoccipital and parietal as described above (Fig. 3). Maryańska and Osmólska (1974) and Sues and Galton (1987) indeed reconstructed a large area on the parietal and squamosal as the insertion of m. spinalis capitis while not recognizing m. longissimus capitis,pars articuloparietalis as a separate muscle (Fig. 1A,B).

At least a part of the insertion of m. spinalis capitis can be readily identified in pachycephalosaurians because this muscle has distinct osteological correlates in extant diapsids. In Iguana iguana, for example, the ventral-most part of this muscle inserts on a process on the dorsal end of the supraoccipital whereas the insertion of the dorsal part of this muscle is on a depression on the parietal. In crocodylians, m. spinalis capitis and m. longissimus capitis,pars articuloparietalis merge together and insert on a process on the dorsal end of the supraoccipital. In Aves, m. spinalis capitis has a small area for its insertion on the occiput characterized by either a depression (e.g., Struthio camelus) or process (e.g., Meleagris gallopavo) in the supraoccipital or on the dorsal end of this bone along the crista nuchalis transversa. Pachycephalosaurians similarly have osteological correlates of m. spinalis capitis at the dorsal end of the supraoccipital. For example, the one in Homalocephale calathocercos (MPC-D 100/1201) is a tuberosity on the median ridge at the dorsal end of the supraoccipital (Fig. 4G), and is similar to the process in I. iguana and A. mississippiensis described above. It is rugosity in a depression on each side of the median ridge for the insertion of the supraspinal ligament in Dracorex hogwartsia (TCMI 2004.17.1; Fig. 5B). The insertion of this muscle in pachycephalosaurians would likely have expanded dorsally from these structures as in Iguana iguana and would have occupied a large area on the parietal (Figs. 4D,E and 5A,C). In D. hogwartsia, for example, the dorsal part of such a putative insertion of m. spinalis capitis on the parietal bears scarring, suggesting extension of the insertion of m. spinalis capitis onto this bone (Fig. 5B).

Distinction of the insertions between m. spinalis capitis and m. longissimus capitis,pars articuloparietalis is often ambiguous in extant diapsids, especially in non-avian taxa in which these two muscles merges at least partially at their insertions on the occiput (e.g., Tsuihiji,2005). Accordingly, the boundary between their insertions may not be clearly distinguishable in fossil dinosaur specimens. It is proposed here that a deep depression lateral to the median ridge for the supraspinal ligament represents the insertion of m. spinalis capitis and that the bone surface lateral to this depression represents the insertion of m. longissimus capitis,pars articuloparietalis in Homalocephale calathocercos (MPC-D 100/1201; Fig. 4E,G) and Dracorex hogwartsia (TCMI 2004.17.1; Fig. 5A). In Stygimoloch spinifer (MPM 8111), the putative insertion of m. spinalis capitis is a deep concavity extending from the parietal to the dorsal part of the supraoccipital within a dorsoventrally elongated depression lateral to the median ridge on the supraoccipital and parietal as described above (Fig. 5C,D). This deep concavity is apparently continuous with a shallow depression on the adjacent part of the squamosal in MPM 8111. Lateral to this depression lies another depression on the squamosal. The demarcation between these two depressions is very prominent in another specimen of the squamosal of S. spinifer (YPM 56085, cast of UCMP 119433), which has a prominent ridge between them. In the reconstruction presented here, the insertion of m. spinalis capitis is considered as expanding onto the medial depression on the squamosal whereas the lateral depression on this bone is interpreted as the insertion of m. longissimus capitis, pars articuloparietalis (Fig. 5C).

The insertions of m. spinalis capitis and m. longissimus capitis, pars articuloparietalis can similarly be reconstructed in ceratopsians. In Psittacosaurus mongoliensis, the median ridge on the supraoccipital has a tubercle on its dorsal end that likely represents the ventral-most part of the insertion of m. spinalis capitis (Fig. 6C,D). The parietal expands posteriorly as a shelf bearing a wide depression, onto which the insertion of m. spinalis capitis likely extended (Fig. 6B,D). The insertion of m. longissimus capitis, pars articuloparietalis is also reconstructed as occupying a large area on this depression (Fig. 6B,D). In Protoceratops andrewsi, on the other hand, there is no definite osteological correlate for m. spinalis capitis. In this study, the insertion of this muscle is reconstructed as being on the dorsal end of the supraoccipital and expanding dorsally to the medial part of the parietal as in pachycephalosaurians and P. mongoliensis (Fig. 6E). There is a shallow, mediolaterally elongated depression extending from the lateral part of the parietal to the squamosal, and this depression is interpreted here as the insertion of m. longissimus capitis,pars articuloparietalis (Fig. 6E). In addition, a tab-like projection on the ventral margin of the parietal fenestra is marked by a shallow depression with scarring, representing another possible insertion of this muscle (Fig. 6E). It is also possible, however, that this projection instead represents the insertion of m. episternocleidomastoideus as discussed below.

In Triceratops horridus (YPM 1823), there is a pair of shallow depressions in the parietal, between which lies a low ridge (Fig. 7D). These depressions are continuous with shallow, groove-like depressions dorsally. In Triceratops prorsus (BSP 1964 I 456), the corresponding area appears as a dorsoventrally elongate, general depression although details of the morphology are obscured by a metal frame in this particular specimen (Fig. 7A). Lull (1908) interpreted these depressions in the parietal as the insertions of “m. complex major” continuing from the supraoccipital. Based on the EPB approach, however, these depressions more likely represent the insertions of m. spinalis capitis (Fig. 7A,B). The weak ridge between the depressions in T. horridus would have been the insertion of the supraspinal ligament continuing from the median ridge in the supraoccipital. As Lull (1908) described, lateral or posterolateral to the putative insertion of m. spinalis capitis lies a depression on the parietal just medial to the suture with the squamosal in T. horridus (YPM 1823; Fig. 7D) and T. prorsus (BSP 1964 I 456). This depression likely represents the insertion of m. longissimus capitis,pars articuloparietalis (Fig. 7A,B) while Lull (1908) labeled this depression as a possible insertion of his “m. levator claviculae” (part of the m. cucullaris complex) in his figure (Fig. 1C). The insertion of m. longissimus capitis,pars articuloparietalis would probably have continued laterally and ventrally onto a smooth, depressed surface on the adjacent part of the squamosal (Fig. 7A,B,D). Without prominent scars in this bone surface, however, it is impossible to determine to what extent it would have been covered by the insertion of this muscle. The present reconstruction based on these inferences implies that the insertion areas of m. spinalis capitis and m. longissimus capitis,pars articuloparietalis are fairly extensive in T. horridus and T. prorsus.

In Chasmosaurus belli, in contrast, the potential insertions of these muscles are restricted to much smaller areas relative to the size of the entire frill due to the presence of a pair of parietal fenestrae (Fig. 8A). The putative insertion of m. spinalis capitis is a shallow, median concavity on the parietal (Fig. 8A,B), which Russell (1935) inferred as the insertion of his “capiti-dorsi-clavicluaris” muscle (part of the m. cucullaris complex; Fig. 1D). This concavity has a scar in YPM 2016 (Fig. 8B), but is just a general, smooth depression in BMNH R 4448. The insertion of m. longissimus capitis,pars articuloparietalis cannot be confidently identified in the specimens of C. belli examined in this study. Russell (1935) reconstructed his “longissimus cervicocapitis” as inserting on a depression on the ventral end of the parietal (Fig. 1D). This depression is instead inferred as a part of the insertion of the lateral slip of m. rectus capitis posterior in this study (Fig. 8A,B). In YPM 2016, a shallow depression is present below the ventral margin of the parietal fenestra. This depression is tentatively interpreted as the insertion of m. longissimus capitis,pars articuloparietalis in this study (Fig. 8A).

M. longissimus capitis, pars transversalis capitis (Level II′ inference).

This muscle arises from the lateral process on the neural arch of the atlas and inserts on the lateral and ventral margins of the paroccipital process in Lepidosauria (Fig. 3). In Crocodylia, it arises from lateral aspects of neural arches of cervical vertebrae and its insertion on the skull is restricted to the distal end of the paroccipital process, proximal to that of “m. iliocostalis capitissensu Seidel (1978; Fig. 3). On the basis of lepidosaurian condition, Maryańska and Osmólska (1974) and Sues and Galton (1987) reconstructed the insertion of m. longissimus capitis, pars transversalis capitis on the distal end of the paroccipital process (Fig. 1A,B). Lull (1908) also reconstructed his “complexus minor” muscle, which apparently corresponds to m. longissimus capitis, pars transversalis capitis, on the distal end of the paroccipital process and the adjacent, medial part of the squamosal in Triceratops (Fig. 1C). In Aves, however, m. longissimus capitis, pars transversalis capitis inserting on the occiput is lacking (Tsuihiji,2007; see, however, Snively and Russell,2007b, for a different view). In addition, this muscle tends to leave no definite osteological correlate specific to it in the occiput of lepidosaurians or crocodylians. Its origins on cervical vertebrae similarly do not have specific osteological correlates that are distinguishable from those of other muscles belonging to the m. longissimus group in these diapsids. Therefore, it is not possible to determine the presence or absence of this muscle in non-avian dinosaurs based on observations on their osteological features. Accordingly, the insertion of m. longissimus capitis, pars transversalis is not reconstructed in the marginocephalian occiput in this study.

M. longissimus capitis, pars transversalis cervicis (Level I inference) and m. iliocostalis capitis (Level II inference).

M. longissimus capitis, pars transversalis cervicis arises from lateral aspects of neural arches, as well as from synapophyses or transverse processes, and inserts on the basal tubera in Lepidosauria and Crocodylia (Figs. 3 and 11). In Lepidosauria, m. iliocostalis capitis arises from the fascia between the m. longissimus and m. iliocostalis groups in the anterior-most cervical region and similarly inserts on the basal tubera (Figs. 3 and 11). In Crocodylia, on the other hand, m. iliocostalis capitis appears to be incorporated in one muscle slip, merged with m. rectus capitis lateralis and m. episternocleidomastoideus (Tsuihiji,2007). This muscle slip, called “m. iliocostalis capitis” by Seidel (1978) and m. capitisternalis by Fürbringer (1876), arises from the lateral surface of the atlas rib and inserts on the distal end of the paroccipital process (Fig. 3). In Aves, m. longissimus capitis, pars transversalis cervicis and m. iliocostalis capitis comprise a single muscle, m. rectus capitis dorsalis (Tsuihiji,2007), which inserts on the basal tubera (Figs. 3 and 11). Therefore, the insertion of m. longissimus capitis, parstransversalis cervicis is on the basal tubera in all extant diapsid clades, leading to a rather safe inference that this muscle would have inserted on the same structure in fossil dinosaurs. The insertion of m. iliocostalis capitis, on the other hand, is different between Lepidosauria and Aves on the one hand and Crocodylia on the other. The phylogenetic distribution of these character states in extant diapsids suggests that m. iliocostalis capitis inserting on the paroccipital process is apomorphic to Crocodylia and thus that the insertion of this muscle can be parsimoniously hypothesized to be on the basal tubera in fossil dinosaurs as in Lepidosauria and Aves. The insertions of m. longissimus capitis, parstransversalis cervicis and m. iliocostalis capitis in Aves are not distinguishable from each other because these two muscles comprise a single muscle, m. rectus capitis dorsalis, as described above. Therefore, the insertions of these two muscles are dealt together in reconstructing their positions in marginocephalians in this study.

Figure 11.

Muscle attachments on the basal tubera and/or basitemporal plate in extant diapsids (modified from Tsuihiji,2007). Attachments of homologous muscles are represented by identical colors and are labeled with lepidosaurian, crocodylian, and avian names. Note that m. rectus capitis anterior lies closer to the occipital condyle than m. longissimus capitis, pars transversalis cervicis in Squamata while the opposite is the case with other diapsid clades. Not to scale. Abbreviations: oc, occipital condyle; parocc, paroccipital process; q, quadrate.

A subvertebral muscle, m. rectus capitis anterior, arising from the ventral aspect of the vertebral column in the anterior cervical region also inserts on the basal tubera or area surrounding it in all diapsid clades (Figs. 3 and 11). The insertion of this muscle lies closer to the occipital condyle than, or proximal to, those of m. longissimus capitis, parstransversalis cervicis and m. iliocostalis capitis in Squamata (Figs. 3 and 11). In Aves, the insertion of m. rectus capitis anterior occupies a wide area of the basitemporal plate, and the insertion of m. longissimus capitis, parstransversalis cervicis and m. iliocostalis capitis combined lies closer to the occipital condyle than this insertion (Figs. 3 and 11). Similarly, the insertion of m. longissimus capitis, parstransversalis cervicis lies proximal to that of m. rectus capitis anterior and is surrounded by the latter in Crocodylia (Figs. 3 and 11). It is found that the condition in Sphenodon punctatus is similar to the one in Crocodylia and Aves, in that m. rectus capitis anterior partially surrounds m. longissimus capitis, parstransversalis cervicis (as well as m. iliocostalis capitis) at their insertions, or at least the insertion of the former muscle extends distally to that of the latter (Figs. 3 and 11), as illustrated in Ostrom (1961; fig. 55). It follows that the condition seen in Squamata is unique to this clade within Diapsida, and is not an appropriate model for reconstructing the insertions of these muscles in fossil dinosaurs.

In the past, both Maryańska and Osmólska (1974) and Sues and Galton (1987) used the squamate condition to infer muscle insertions on the basal tubera in pachycephalosaurians and reconstructed the insertion of m. rectus capitis anterior more proximal, or closer to the midline, than that of m. longissimus capitis, parstransversalis cervicis (Figs. 1A,B and 12C). Application of the EPB approach instead suggests that the relative positions of these insertions should be just the opposite in these pachycephalosaurians (Fig. 12C). Lull (1908), on the other hand, reconstructed only subvertebral muscles, and not a longissimus muscle, as inserting on the basal tubera in Triceratops (Fig. 1C).

Figure 12.

Basal tuberas of basal ornithopods (A, B) and marginocephalians (CF) with muscle attachments reconstructed in posterior (A, C–F) and left posterolateral (B) views. (A) Tenontosaurus tilletti (OMNH 10132); (B) Zephyrosaurus schaffi (MCZ 4392); and (C) Prenocephale prenes modified from Maryańska and Osmólska (1974), with their reconstruction of muscle attachments based on the squamate condition (above) compared with the present reconstruction using the EPB approach (below). Note that relative positions of the insertions of m. longissimus capitis, pars transversalis cervicis and m. iliocostalis capitis combined and m. rectus capitis anterior are reversed between these two reconstructions; (D) Psittacosaurus mongoliensis (AMNH 6254); (E) Protoceratops andrewsi (HMNS uncataloged cast of MPC-D 100/533); and (F) Triceratops horridus (YPM 1820). Not to scale. Abbreviation: oc, occipital condyle.

In most marginocephalian and ornithopod specimens that were examined, two distinct scars for muscle attachment were recognized on the basal tubera, one along or close to the midline and another along the distal margin of this structure (Fig. 12). Based on the EPB approach, the former scar is identified as the insertion of m. longissimus capitis, parstransversalis cervicis and m. iliocostalis capitis combined and the latter as that of m. rectus capitis anterior in these dinosaurs (Fig. 12). In some taxa such as Protoceratops andrewsi (HMNS uncataloged cast of MPC-D 100/533) and Triceratops horridus (YPM 1820), a shallow depression between these two scars may represent the insertion of additional, fleshy fibers of m. longissimus capitis, parstransversalis cervicis (Fig. 12E,F).

Subvertebral muscles: M. rectus capitis anterior (Level I inference) and m. rectus capitis lateralis (Level I or I′ inference).

In Lepidosauria, the subvertebral muscle, m. rectus capitis anterior, inserts on the basal tubera as described above (Figs. 3 and 11). In Crocodylia and Aves, the subvertebral muscle inserting on the occiput is divided into two parts. One part inserts on the basal tubera as in Lepidosauria (Figs. 3 and 11), leading to the reconstruction of its insertion in marginocephalians described above (Fig. 12). The other part, m. rectus capitis lateralis, inserts on the paroccipital process. In Aves, this muscle inserts on the distal end of the paroccipital process (Fig. 3). In Crocodylia, m. rectus capitis lateralis is incorporated in “m. iliocostalis capitissensu Seidel (1978), or m. capitisternalis sensu Fürbringer (1876), that inserts on the distal end and ventral margin of the paroccipital process, medial to the origin of m. depressor mandibulae (Fig. 3). By phylogenetic bracketing using these two extant archosaurian taxa, it is safe to infer that a subvertebral muscle, m. rectus capitis lateralis, would have inserted on the distal end of the paroccipital process in marginocephalians (Figs. 4, 5, 6, 7, 8, 4–8). This insertion was not reconstructed in the past studies on marginocephalians by Lull (1908), Maryańska and Osmólska (1974), and Sues and Galton (1987) because these studies used the lepidosaurian condition as a model (Fig. 1A–C). Russell (1935) was an exception as he recognized the insertion of this muscle (his “m. rectus capitis anticus minor”; Fig. 1D). Osteological correlates of this insertion are not always clear in marginocephalians. In Triceratops horridus (YPM 1820), however, a strong rugosity is developed at the distal end of the paroccipital process that may represent the insertion of this muscle (Fig. 7C).

M. depressor mandibulae (Level I or I′ inference).

Although m. depressor mandibulae, which inserts on the retroarticular process or posterior fossa of the lower jaw, is not an axial muscle, it arises from the occipital region and its origin has been reconstructed in marginocephalians in the past (Fig. 1). In Lepidosauria, the origin of this muscle in the occipital region includes the parietal, squamosal, and/or distal end of the paroccipital process (Figs. 3 and 13). In both Crocodylia and Aves, the origin of this muscle includes the distal edge of the paroccipital process and further extends onto the adjacent part of the squamosal (Figs. 3 and 13). In addition, the origin of this muscle extends to the quadrate and lateral aspect of the squamosal in Crocodylia and to the parietal in some birds (e.g., Struthio camelus; Figs. 3 and 13). In past reconstructions of marginocephalians, the origin of this muscle was inferred to lie on the posterior surface of the quadrate and quadratojugal (in Triceratops; Lull,1908, Fig. 1C), on a concave area on the ventral part of the lateral surface of the squamosal (in Protoceratopsandrewsi; Haas,1955), on a large area on the posterior surfaces of the squamosal and parietal near the distal end of the paroccipital process (in Chasmosaurus belli and other ceratopsids; Russell,1935, Fig. 1D), on the ventrally deflected paroccipital process (in Stegoceras validum; Sues and Galton,1987, Fig. 1B), and on a rugosity on the distal end of the paroccipital process, or possibly the posterolateral margin of the squamosal (in pachycephalosaurians such as Homalocephale calathocercos and Prenocephale prenes; Maryańska and Osmólska,1974). Phylogenetic bracketing based on conditions in extant diapsids, especially those in Crocodylia and Aves, suggests that the origin of m. depressor mandibulae in these dinosaurs would have included the distal end of the paroccipital process and the posterior surface of the squamosal (Figs. 4, 5, 6, 7, 8, 4–8). In H. calathocercos (MPC-D 100/1201) and Dracorex hogwartsia (TCMI 2004.17.1), for example, the posterior aspect of the lateral margin of the squamosal bears scarring that likely represents the origin of this muscle (Fig. 4F). The origin would have likely extended ventrally and included the ventrally deflected part of the distal end of the paroccipital process in these and other pachycephalosaurians (Figs. 4D,E and 5A) as Maryańska and Osmólska (1974) inferred. In Protoceratops andrewsi (AMNH 6466 and HMNS uncataloged cast of MPC-D 100/533), there is a narrow groove on the ventrolateral edge of the squamosal dorsal to the distal end of the paroccipital process, which would have served as at least a part of the origin of this muscle. The origin may also have extended onto the lateral surface of the squamosal as Haas (1955) inferred, and/or dorsally onto the posterior surface of the squamosal as reconstructed in Fig. 6E. In Triceratops horridus (YPM 1823), a smooth depression lies on the ventromedial part of the posterior surface of the squamosal (Fig. 7D). In another specimen of T. horridus (YPM 1820), this part of the squamosal bears rugosity (Fig. 7C). These depression and rugosity, as well as the adjacent, distal end of the paroccipital process, would have been the origin of m. depressor mandibulae (Fig. 7B).

Figure 13.

Origins of m. depressor mandibulae in extant diapsids on the occiputs in posterior view. The distal end of the paroccipital process is a part of the origin of this muscle in all diapsid clades, and the squamosal also serves as its origin both in Crocodylia and in Aves. Also note that the origin of this muscle extends onto the parietal in Struthio camelus, as in Lepidosauria. Not to scale. Abbreviations: fm, foramen magnum; oc, occipital condyle; p, parietal; parocc, paroccipital process; q, quadrate; sq, squamosal.

M. episternocleidomastoideus (part of the m. cucullaris complex; Level II′ inference).

Another muscle group that has an attachment on the occiput is the m. cucullaris complex, which generally arises from the shoulder girdle and/or sternum.

Although this complex consists of a single muscle, m. cucullaris, in Sphenodon punctatus and some squamates, it divides into two muscles, m. trapezius and m. episternocleidomastoideus, in the majority of squamates, as well as in Crocodylia and Aves (e.g., Fürbringer,1876,1900,1902). In lepidosaurians, this muscle complex (consisting of either a single m. cucullaris or separate m. trapezius and m. episternocleidomastoideus) has the insertion on the parietal, squamosal, and/or distal end of the paroccipital process medial to the origin of m. depressor mandibulae (Fig. 3). In Crocodylia and Aves, only m. episternocleidomastoideus inserts on the skull. In Crocodylia, this muscle (as a part of the merged “m. iliocostalis capitis”) inserts on the distal end and ventral margin of the paroccipital process, medial to the origin of m. depressor mandibulae (Fig. 3). In Aves, the insertion of m. episternocleidomastoideus varies, including the squamosal, frontal, or parietal (e.g., Vanden Berge,1975; Vanden Berge and Zweers,1993). In Struthio camelus, for example, the insertion of this muscle is near the posterior end of the parietal and lies dorsomedial to the origin of m. depressor mandibulae (Fig. 3). In Casuarius casuarius, on the other hand, the insertion of m. episternocleidomastoideus extends from the posterior edge of the parietal to the dorsolateral margin of the paroccipital process, partly lying medial to the origin of m. depressor mandibulae (Fig. 3).

By using the EPB approach based on the conditions in Crocodylia and Aves, it can be inferred that m. episternocleidomastoideus would have inserted on the skull in marginocephalians. The position of its insertion, however, cannot be inferred unambiguously because it is different among extant diapsids, especially between Crocodylia and Aves. In this study, the insertion of this muscle is tentatively reconstructed on the distal end of the paroccipital process, medial to the origin of m. depressor mandibulae using the crocodylian condition as a model (Figs. 4, 5, 6, 7, 8, 4–8). However, the posterior surface of the parietal or squamosal is an equally plausible position for its insertion based on the avian condition. Lull (1908) in fact illustrated the insertion of his “levator claviculae” muscle in Triceratops on the lateral part of the parietal near the suture with the squamosal (Fig. 1C). This part of the parietal is reconstructed as a part of the insertion of m. longissimus capitis,pars articuloparietalis in this study (Fig. 7A,B). Similarly, a tab-like projection on the parietal that is reconstructed as a part of the insertion of this muscle in Protoceratops andrewsi above (Fig. 6E) may instead represent the insertion of m. episternocleidomastoideus if the latter muscle inserted on the parietal as it does in many extant birds.

Russell (1935), on the other hand, inferred his “capiti-dorsi-clavicularis” muscle in Chasmosaurus belli as inserting on a median concavity on the parietal (Fig. 1D), which is identified as the insertion of m. spinalis capitis in this study (Fig. 8). M. capitidorsoclavicularis (= m. trapezius in many other studies) is a muscle in lepidosaurians, and its crocodylian and avian homologs hypothesized by Fürbringer (1876,1900,1902), m. dorsoscapularis and m. cucullaris cervicis, respectively, do not insert on the skull. In addition, the insertion of this muscle in lepidosaurians does not include the very median part of the parietal in the taxa examined in Tsuihiji (2007), such as Sphenodon punctatus and Iguana iguana (Fig. 3). Accordingly, it is unlikely that m. capitidorsoclavicularis or its homolog would have inserted on the median concavity on the parietal in Chasmosaurus contrary to Russell (1935).

Muscle Attachments in the Occiput in Tyrannosauridae

M. rectus capitis posterior (Level I or I′ inference).

Because the all extant diapsids examined have the insertion of m. rectus capitis posterior on the supraoccipital, the most parsimonious inference based on the EPB approach is that the posterior surface of this bone, ventral to a prominent process on its dorsal end, serves as at least a part of the insertion of this muscle (Fig. 14A,B). The midline of the supraoccipital, deeply concave in some tyrannosaurid specimens (e.g., Daspletosaurus torosus CMN 8506 and Daspletosaurus sp. MOR 590; Fig. 14C,E), would have been the insertion of the supraspinal ligament. The insertion of m. rectus capitis posterior would have been the area lateral to it although this area tends to lack a prominent osteological correlate of the muscle insertion. It is possible that the attachment site of this muscle extended laterally onto the adjacent, concave area of the parietal. Bakker et al. (1988) similarly inferred that the insertion of this muscle is on the supraoccipital. These authors, however, also inferred that the prominent process on the dorsal end of this bone represents a tendinous insertion of this muscle. This process (Fig. 14C–E) is instead inferred as a tendinous insertion of m. spinalis capitis in this study as discussed below. Snively and Russell (2007a,b), in contrast, identified a large, concave area of the parietal as the insertion of m. rectus capitis posterior (their “medial portion of m. splenius capitis”). This area of the parietal is inferred to be the insertion of m. longissimus capitis,pars articuloparietalis in this study as described below.

Figure 14.

Occipital regions of tyrannosaurids, Daspletosaurus torosus [CMN 8506, (A) and (C)], Tarbosaurus bataar [MPC-D 107/14, (B) and (D)], and Daspletosaurus sp. [MOR 590, (E)]. (A,B) Muscle attachments reconstructed on the occiputs in posterior view. The insertion of m. episternocleidomastoideus indicated by “?” is only tentatively reconstructed using the crocodylian condition as a model. An alternative position of the insertion of this muscle is a tab-like lateral expansion on the parietal, which is regarded as a part of the origin of m. depressor mandibulae (as reconstructed with “?”) in the present figure. (C–E) Prominent osteological correlates of muscle attachments (1–3) on the nuchal crests in posterior view. (1) Process on the dorsal end of the supraoccipital representing the tendinous insertion of m. spinalis capitis; (2) scar and/or depression on the parietal representing the fleshy insertion of m. spinalis capitis. This scar is especially well-demarcated in Daspletosaurus sp. as shown in (E); and (3) tab-like expansion putatively representing the parietal part of the origin of m. depressor mandibulae (or alternatively representing the insertion of m. episternocleidomastoideus). Not to scale. Abbreviations: fm, foramen magnum; oc, occipital condyle; p, parietal; parocc, paroccipital process; so, supraoccipital; sq, squamosal.

M. obliquus capitis magnus (Level I′ inference).

As reconstructed in marginocephalians above, the insertion of m. obliquus capitis magnus can be inferred to be on the paroccipital process in tyrannosaurids based on the EPB approach (Fig. 14A,B), as previously reconstructed in Allosaurus by Bakker (2000). Snively and Russell (2007a,b) inferred that this muscle (their “lateral division of m. splenius capitis ”) may have inserted on the squamosal. M. obliquus capitis magnus, however, does not insert on the squamosal in extant Lepidosauria, Crocodylia, or Aves (Fig. 3), making the inference in this study more plausible than the one by Snively and Russell (2007a,b).

M. spinalis capitis and m. longissimus capitis, pars articuloparietalis (Level I inference).

Because the dorsal end of the supraoccipital serves as at least a part of the insertion of m. spinalis capitis in Iguana iguana, Crocodylia, and some birds as described above (Fig. 3), the prominent process on the dorsal end of this bone in tyrannosaurids is identified here as the tendinous insertion of this muscle (Fig. 13C–E). The part of the parietal adjacent and dorsal to this process of the supraoccipital is marked by a depression in Tyrannosaurus rex (MOR 555) and Daspletosaurus torosus (CMN 8506; Fig. 14C) and a scar in Tarbosaurus bataar (MPC-D 107/14; Fig. 14D) and Daspletosaurus sp. (MOR 590; Fig. 14E). The scar in Daspletosaurus sp. (MOR 590) is especially well demarcated: it is dorsoventrally elongated and ovoid-shaped, bounded laterally by a low ridge and medially by the median ridge for insertion of the supraspinal ligament (Fig. 14E). These scars or depressions likely represent the fleshy insertion of m. spinalis capitis, continued dorsally from the putative, tendinous insertion on the dorsal process on the supraoccipital (Fig. 14A,B). The insertion of m. longissimus capitis,pars articuloparietalis, on the other hand, is inferred here to be on the concave, posterior surface of the nuchal crest of the parietal in tyrannosaurids (Fig. 14A,B) based on observations that this muscle inserts mainly on the parietal in extant diapsids except for Crocodylia as described above (Fig. 3).

Snively and Russell (2007a,b), as well as Bakker et al. (1988), inferred that the insertion of m. spinalis capitis (their “m. transversospinalis capitis”) in tyrannosaurids is the dorsal margin of the parietal that tends to be marked by rugosities. These authors interpreted such rugosities as representing a strong tendinous insertion of m. spinalis capitis. In Iguana iguana and Alligator mississippiensis, however, the skin attaches to the rugose, posterodorsal margin of the parietal (personal observations). This suggests that at least part of rugosities on the parietal in tyrannosaurids may similarly represent the skin attachment. As for m. longissimus capitis,pars articuloparietalis, Snively and Russell (2007a,b) inferred that this muscle (their “m. complexus”) would have inserted on the posterior aspect of the squamosal in tyrannosaurids. As described above, the insertion of this muscle includes the squamosal only in Sphenodon punctatus among extant diapsids examined in Tsuihiji (2005; Fig. 3). Although Snively and Russell (2007b) described the insertion of m. longissimus capitis,pars articuloparietalis includes the posterior surface of the squamosal in some birds, the main insertion of this muscle in birds is on the parietal, as demonstrated in an embryonic Struthio camelus by Tsuihiji (2005; Fig. 3). Another reason for which Snively and Russell (2007b) reconstructed the insertion of this muscle on the squamosal in tyrannosaurids is their observations that the crocodylian m. epistropheo-capitis lateralis and lepidosaurian m. obliquus capitis magnus, both of which these authors hypothesized to be homologous with m. longissimus capitis,pars articuloparietalis (their avian “m. complexus”), insert on the squamosal. However, those crocodylian and lepidosaurian muscles, both of which are called m. obliquus capitis magnus in this study, are not homologous with m. longissimus capitis,pars articuloparietalis according to Tsuihiji (2005). In addition, m. obliquus capitis magnus in lepidosaurians and crocodylians inserts on the paroccipital process (otoccipital), not on the squamosal, as described above (Fig. 3). These observations demonstrate that the parietal is a more likely insertion for m. longissimus capitis,pars articuloparietalis than the squamosal is. Furthermore, unlike in marginocephalians, the posterior surface of the squamosal does not from a continuous surface with the parietal in tyrannosaurids. Accordingly, it is likely that the insertion of this muscle did not expand onto the squamosal in tyrannosaurids unlike in marginocephalians. Instead, the posterior surface of the squamosal would have more likely served as the origin of m. depressor mandibulae as discussed below.

M. longissimus capitis, pars transversalis capitis (Level II′ inference).

Because this muscle is absent in Aves and does not leave a definite and specific osteological correlate on skeletal elements even when it is present in Lepidosauria or Crocodylia, it is not possible to determine the presence or absence of this muscle in tyrannosaurids based on their osteological features and thus its insertion on the occiput is not reconstructed in this study. Snively and Russell (2007a,b), on the other hand, reconstructed the insertion of this muscle on the distal end of the paroccipital process by using the crocodylian condition as a model.

M. longissimus capitis, pars transversalis cervicis (Level I inference), m. iliocostalis capitis (Level II inference), and m. rectus capitis anterior (Level I inference).

The insertion of m. longissimus capitis, parstransversalis cervicis and m. iliocostalis capitis combined, as well as that of m. rectus capitis anterior, can be inferred to be on the basal tubera in tyrannosaurids as is the case with marginocephalians (Fig. 14A,B). In tyrannosaurids such as Daspletosaurus torosus (CMN 8506), Daspletosaurus sp. (MOR 590), Tarbosaurus bataar (MPC-D 107/14), and Gorgosaurus sp. (TCMI 2001.89.1), a scar (“ascending scar” in Bakker et al.,1988) is present on the posterior or posteromedial aspect of the medial ridge of the basal tubera. This ridge extends ventrolaterally from the occipital condyle and medially bounds a concavity on the posterior aspect of the basal tubera. The scar extends ventrolaterally and surrounds the ventral margin of the concavity. This scar likely represents the insertion of m. longissimus capitis, parstransversalis cervicis and m. iliocostalis capitis combined. The concavity is generally considered as a pneumatic feature (e.g., Currie,2003). However, although its surface is generally rather smooth, it bears a weak scar in specimens such as Daspletosaurus torosus (CMN 8506), suggesting that a part of the concavity may also have served as the insertion of fleshy fibers of these muscles. The scar on the medial ridge extends further ventrally and anteriorly onto the ventral end of the basioccipital and the adjacent, posteroventral part of the basisphenoid. This part of the scar is inferred here to represent the insertion of m. rectus capitis anterior, one of the two subvertebral muscles inserting on the skull (Fig. 14A,B).

Bakker et al. (1988) inferred that m. rectus capitis anterior inserts on the “ascending scar” whereas m. iliocostalis capitis inserts on the scar on the basisphenoid. The topological relationship between the insertions of these muscles inferred by Bakker et al. (1988), therefore, is basically reversed from the one inferred in this study. Snively and Russell (2007a,b), on the other hand, inferred that m. longissimus capitis, parstransversalis cervicis (their “m. longissimus capitis profundus” or “m. rectus capitis dorsalis”) would have inserted on the lateral margin of the basal tubera “as in crocodilians” (Snively and Russell,2007b, p 795). In Alligator mississippiensis, Caiman crocodilus (Fig. 11), and Osteolaemus tetraspis examined in Tsuihiji (2007), however, the insertion of m. longissimus capitis, parstransversalis cervicis included the median, vertical ridge on the basal tubera. Therefore, this crocodylian condition is in accordance with the present inference that this muscle would have inserted on the medial ridge or “ascending scar” of the basal tubera. The insertion of m. rectus capitis anterior reconstructed by Snively and Russell (2007a,b), on the other hand, generally agrees with the one inferred in this study.

M. rectus capitis lateralis (Level I or I′ inference).

As in marginocephalians, m. rectus capitis lateralis can be inferred to have inserted on the distal end of the paroccipital process in tyrannosaurids (Fig. 14A,B). In some specimens such as Daspletosaurus sp. (MOR 590) and Gorgosaurus sp. (TCMI 2001.89.1), the posterior surface of the distal end of the paroccipital process bears a scar consisting of short, horizontal striations that likely represents the insertion of this muscle.

Snively and Russell (2007a,b) inferred the insertion of m. rectus capitis lateralis to be on the ventral portion or ventral edge of the paroccipital process in tyrannosaurids. The insertion of this muscle, however, includes the distal end of the paroccipital process in both Crocodylia and Aves (Fig. 3), making this part of the paroccipital process a more likely site of insertion of this muscle in tyrannosaurids based on the EPB approach (Fig. 14A,B).

M. depressor mandibulae (Level I or I′ inference) and m. episternocleidomastoideus (Level II′ inference).

Phylogenetic bracketing based on conditions in Crocodylia and Aves (Figs. 3 and 13) suggests that the origin of m. depressor mandibulae in tyrannosaurids would have included the posterior surface of the squamosal, which bears scarring, and the distal end of the paroccipital process (Fig. 14A,B) as previously reconstructed in Allosaurus by Bakker (2000).

As is the case with marginocephalians, the position of the insertion of m. episternocleidomastoideus in tyrannosaurids cannot be determined based on the EPB approach because this position is different between Crocodylia and Aves. In this study, the insertion of this muscle is tentatively reconstructed on the distal end of the paroccipital process medial to the origin of m. depressor mandibulae using the crocodylian condition as a model (Fig. 14A,B).

In tyrannosaurids such as Daspletosaurus torosus (CMN 8506) and Tarbosaurus bataar (MPC-D 107/14), the lateral margin of the parietal bears a distinct, tab-like expansion that is reflected slightly anteriorly (Fig. 14A–C). In Struthio camelus and Lepidosauria, the origin of m. depressor mandibulae extends from the squamosal onto the lateral part of the parietal (Figs. 3 and 13). Therefore, this tab-like expansion in tyrannosaurids may represent the parietal part of the origin of this muscle as reconstructed in Fig. 13A,B. Another possibility is that this expansion served as the insertion of m. episternocleidomastoideus if this muscle did not insert on the distal end of the paroccipital process as reconstructed in this study. Such a reconstruction would be similar to the condition observed in many extant birds, in which m. episternocleidomastoideus inserts on the parietal (Fig. 3).

DISCUSSION

Muscle Attachments in the Occiput in Marginocephalia

Application of the EPB approach to the marginocephalian occiput in this study resulted in inferred attachments of several muscles greatly different from those in past reconstructions. First, as for pachycephalosaurians, the most conspicuous differences in inferred muscle attachments between this study and studies by Maryańska and Osmólska (1974) and Sues and Galton (1987) are derived from these authors' choice of the lepidosaurian or squamate condition, which includes some apomorphies that are not present in archosaurians, as a model for their reconstructions. Such lepidosaurian or squamate apomorphies include the absence of a subvertebral muscle, m. rectus capitis lateralis, inserting on the paroccipital process (apomorphic to Lepidosauria) and the insertion of m. rectus capitis anterior on the basal tubera lying proximal to those of m. longissimus capitis, pars transversalis cervicis and m. iliocostalis capitis (apomorphic to Squamata). Thus, application of such lepidosaurian and squamate conditions to pachycephalosaurians by Maryańska and Osmólska (1974) and Sues and Galton (1987) resulted in reconstructions of muscle attachments (Fig. 1A,B) that are not congruent with the archosaurian conditions. Based on the EPB approach, this study corrected such aspects of these past reconstructions by inferring an insertion of m. rectus capitis lateralis on the distal end of the paroccipital process and reversing the relative positions of the insertions of m. rectus capitis anterior and m. longissimus capitis, pars transversalis cervicis + m. iliocostalis capitis (Figs. 4D,E, 5A, and 12C).

Other muscles of which insertions were reconstructed differently between Maryańska and Osmólska (1974) and Sues and Galton (1987) on the one hand and this study on the other include m. obliquus capitis magnus and m. longissimus capitis,pars articuloparietalis. This study regarded the site of insertion of m. obliquus capitis magnus identified by Maryańska and Osmólska (1974) and Sues and Galton (1987) as the site of attachment of the atlanto-occipital capsular membrane/ligament (Fig. 9), and instead reconstructed the insertion of this muscle more dorsally on the paroccipital process (Figs. 4D,E and 5A). The insertion of m. longissimus capitis,pars articuloparietalis, on the other hand, was not recognized by these authors. This study reconstructed this muscle as inserting on the squamosal (Figs. 4D,E and 5A,C), with its insertion taking up the lateral part of that of m. spinalis capitis reconstructed by Maryańska and Osmólska (1974) and Sues and Galton (1987; Fig. 1A,B). As a result, the insertion of the latter muscle reconstructed in this study is not nearly as extensive as the one reconstructed by these authors. Another potential difference between the past and present reconstructions of the marginocephalian occiput concerns m. longissimus capitis, pars transversalis capitis. The present application of the EPB approach could not determine the presence or absence of the insertion of this muscle as described above. This muscle may have been retained in pachycephalosaurians with its insertion on the paroccipital process as reconstructed by Maryańska and Osmólska (1974) and Sues and Galton (1987; Fig. 1A,B), or it may have already been absent as in extant Aves.

Second, the two past reconstructions on ceratopsians by Lull (1908) and Russell (1935) are fairly different from each other. Between them, the reconstruction by Russell (1935) on Chasmosaurus belli is closer to the present reconstruction based on the EPB approach probably because the former study was based partly on the anatomy of a bird (raven). Muscles of which insertions were reconstructed differently between Russell (1935) and this study include m. spinalis capitis and m. longissimus capitis,pars articuloparietalis. The insertions of these two muscles were reconstructed in more dorsal positions on the parietal in this study (Fig. 8) than in Russell (1935; Fig. 1D). In addition, this study agrees with Russell (1935) on the insertion of m. rectus capitis posterior only partially. That is, in addition to the depression on the dorsal part of the paroccipital process that Russell (1935) inferred as the insertion of this muscle, this study regarded a second depression on the parietal continuous with this depression (reconstructed as the insertion of the “longissimus capitis” by Russell,1935) and another depression dorsal to the foramen magnum (reconstructed as the insertion of the “biventer cervicis” by Russell,1935) as other parts of the insertion of this muscle (Fig. 8). Furthermore, unlike this study, Russell (1935) did not reconstruct the insertion of m. obliquus capitis magnus.

Lull's (1908) reconstruction on Triceratops (Fig. 1C), on the other hand, differs from the present reconstruction in most respects. The insertions of m. spinalis capitis, m. obliquus capitis magnus, and m. rectus capitis lateralis were not reconstructed in Lull (1908). The insertion of m. longissimus capitis,pars articuloparietalis reconstructed by Lull (1908; his “complexus major”) was interpreted as those of m. spinalis capitis and the medial part of m. rectus capitis posterior in this study (Fig. 7A–C). The insertion of m. longissimus capitis,pars articuloparietalis in this study was instead reconstructed as occupying a large area on the parietal and squamosal. Lull (1908) reconstructed a suboccipital muscle (“rectus posticus minor”) inserting on a depression lateral to the base of the occipital condyle (Fig. 1C). This depression, however, appears to represent exits of cranial nerves. In addition, Lull (1908) reconstructed a large insertion of m. longissimus capitis, pars transversalis capitis (his “complexus minor”) on the distal part of the paroccipital process and the adjacent, medial part of the squamosal (Fig. 1C). Although the presence or absence of this muscle in marginocephalians could not be determined in this study, it is possible that this muscle was retained in these dinosaurs as discussed above. Even if it was present in Triceratops, however, the insertion of this muscle would unlikely have expanded to the squamosal because the distal end of the paroccipital process probably served as the attachments of m. rectus capitis lateralis and m. depressor mandibulae (Fig. 7A–C).

To summarize, reconstructions of muscle attachments in the marginocephalian occiputs proposed in this study are substantially different from those in past studies. As discussed above, this is partly due to the fact that many such past reconstructions were based on the lepidosaurian or squamate conditions, which include features not present in archosaurians, while this study mainly used the archosaurian conditions based on the EPB approach. In addition, this study emphasized on correct recognition of osteological correlates of muscle attachments in extant animals based on detailed dissections while some of the past reconstructions relied on published accounts in identifying muscle attachments. The differences between the past and present reconstructions therefore emphasize the importance of collecting data on the muscular anatomy through detailed dissections of various extant diapsids, especially Aves and Crocodylia, for reconstructing the anatomy of non-avian dinosaurs. It should be noted, however, that even this study could not unambiguously determine the conditions in marginocephalians for a couple of muscles, that is, presence or absence of the insertion of m. longissimus capitis, pars transversalis capitis and the position of the insertion of m. episternocleidomastoideus. Conditions of these muscles are different between Aves and Crocodylia, and the insertions of these muscles also do not have clear and specific osteological correlates that can be identified in marginocephalians. Reconstructions of conditions of these muscles, therefore, require higher degrees of speculation than those of other muscles inserting on the occiput.

The present reconstructions show that insertions of axial muscles are generally enlarged in derived marginocephalians, apparently correlated with expansion of the parietosquamosal shelf/frill in these dinosaurs. In particular, the osteological correlates for the insertion of m. rectus capitis posterior are two pairs of large depressions in Ceratopsidae and are especially enlarged in Triceratops (Fig. 7C). The main origin of this muscle is the dorsal or anterior edge of the neural spine of the axis in extant diapsids. In Ceratopsidae, the neural spine of the axis is longitudinally elongated (e.g., Hatcher et al.,1907; Tsuihiji and Makovicky,2007), thus indicating that the origin of m. rectus capitis posterior would have been enlarged. Accordingly, the morphology of the both origin and insertion suggests massive development of m. rectus capitis posterior in these dinosaurs. The pachycephalosaurian Stygimoloch spinifer also has a deep depression marking the insertion of this muscle (Fig. 5D). These observations suggest the convergent hypertrophy of m. rectus capitis posterior in derived members of Pachycephalosauria and Ceratopsia, apparently reflecting a functional demand of supporting their heavy heads.

In addition to accommodating large muscle masses, expansion of the parietosquamosal shelf contributes to increasing the lever arm lengths of in-forces of some epaxial muscles from the fulcrum (= occipital condyle) to increase out-forces for supporting or moving the heavy head. For example, m. longissimus capitis,pars articuloparietalis has an especially long lever arm length in Triceratops (Fig. 7A,B). This muscle also has a large insertion area, suggesting that it would have pulled with great force. Together, this morphology suggests that m. longissimus capitis,pars articuloparietalis in Triceratops would have exerted a large torque for flexing the head dorsolaterally. Farlow and Dodson (1975) hypothesized that Triceratops engaged in intraspecific combat in which in horns were locked in shoving and wrestling matches involving twisting cranial movements. Powerful dorsolateral flexion of the head enabled by this muscle, therefore, could have been used in such shoving behavior while horns were interlocked in positions such as those hypothesized by Farke (2004).

Muscle Attachments in the Occiput in Tyrannosauridae

Despite the fact that both Snively and Russell (2007a,b) and this study used the phylogenetic bracketing method based on the anatomy of extant diapsids, positions of insertions of several muscles, particularly those of m. obliquus capitis magnus, m. spinalis capitis, and m. longissimus capitis,pars articuloparietalis, were inferred differently in the tyrannosaurid occiput between these studies. Such different reconstructions resulted mainly from differences in identification of the insertion areas of the same muscles or different hypotheses on muscle homologies in extant diapsids between Snively and Russell (2007a,b) and this study as discussed above. Another factor contributing to different reconstructions of some muscle insertions, especially that of m. longissimus capitis,pars articuloparietalis, is the inferred position of the origin of m. depressor mandibulae. Although Snively and Russell (2007a,b) did not specifically address the origin of m. depressor mandibulae, one of their figures (fig. 8A in Snively and Russell,2007a) shows that these authors considered it as being restricted to the ventrolateral corner of the paroccipital process. In this study, in contrast, the origin of m. depressor mandibulae was reconstructed as including not only the distal end of the paroccipital process but also the posterior surface of the squamosal based on phylogenetic bracketing. This reconstruction, therefore, made the insertion of m. longissimus capitis,pars articuloparietalis on the squamosal proposed by Snively and Russell (2007a,b) unlikely. In other words, the reconstructed origin of m. depressor mandibulae was able to constrain the possible insertion area of m. longissimus capitis,pars articuloparietalis in the present reconstruction. This demonstrates the importance of detailed examinations of not only the structures of interest but also those surrounding them, in reconstructing the anatomy of extinct animals.

Differences in reconstructed positions of axial muscle insertions affect functional inferences such as those made by Snively and Russell (2007a,b). For example, Snively and Russell (2007a), based on their reconstruction of the insertion of m. longissimus capitis,pars articuloparietalis (their “m. complexus”) on the squamosal, inferred that the morphology of this bone influences the capacity of this muscle for dorsal and lateral flexion of the head in large theropods. They inferred that this muscle played a significant role in lateral flexion of the head in tyrannosaurids because of its putatively long lateral lever arm from the occipital condyle. They further inferred that the great width across the squamosals in tyrannosaurids suggests this muscle having “stronger leverage for lateroflexion than in other theropods” (p 953). In contrast, the present reconstruction of this muscle inserting on the parietal nuchal crest (Fig. 14A,B) would suggest that this muscle would have had a longer lever arm for dorsal flexion than for lateral flexion from the occipital condyle and would not have been an effective lateral flexor of the head. Furthermore, this study recognized the squamosal as the origin of m. depressor mandibulae as discussed above and not as an insertion of an axial muscle, making the morphology of this bone mostly irrelevant in discussing functions of axial muscles in moving the head. Another example is m. obliquus capitis magnus, of which insertion was reconstructed as a small area on the squamosal by Snively and Russell (2007a). This study, in contrast, reconstructed the insertion of this muscle as occupying a large area on the posterior surface of the paroccipital process (Fig. 14A,B). The present reconstruction indicates that this muscle would have played a much more significant role in lateral flexion of the head than the reconstruction by Snively and Russell (2007a) would suggest.

These results demonstrate that initial differences or errors in observations and interpretations of the anatomy of extant animals are amplified in subsequent paleobiological inferences such as anatomical reconstructions and functional analyses based on them. As is the case with the studies on the marginocephalian occiput discussed above, this emphasizes the importance of accurate and detailed observations on the anatomy of extant animals as the basis for such paleobiological analyses.

CONCLUSION

With an advent of parsimony-based methods of Bryant and Russell (1992) and Witmer (1995), past reconstructions of the soft tissue anatomy in extinct taxa can now be evaluated in terms of phylogenetic distributions of characters in extant bracketing taxa. This study applied such methods to muscle attachments on the occiput of marginocephalian dinosaurs, and effectively confirmed or corrected aspects of past reconstructions, most of which were based on a single animal model. The present reconstructions of occipital muscle attachments in these dinosaurs, therefore, are phylogenetically justified inferences and will provide the basis for future functional inferences on the use of their heads, such as possible combat behaviors both in pachycephalosaurians (e.g., Sues,1978; Carpenter,1997; Bakker et al.,2006) and in ceratopsians (e.g., Farlow and Dodson,1975; Farke,2004; Farke et al.,2009).

In addition, occipital muscle attachments in tyrannosaurids were reconstructed and the results were compared with recent reconstructions in these dinosaurs by Snively and Russell (2007a,b). Several aspects of reconstructions were fairly different between Snively and Russell (2007a,b) and this study despite the fact that both studies used the same, phylogenetic and parsimony-based method based on the anatomy of extant diapsids. This is due to differences in initial identifications of muscle attachment areas or different hypotheses on muscle homologies in extant diapsids between these studies. This result demonstrates that, even with the methodological advance, the most critical step in soft tissue reconstructions in extinct animals is still collecting accurate and detailed data in extant taxa. Differences in reconstructions between Snively and Russell (2007a, b) and this study, therefore, may be resolved by examining more extant taxa, especially crocodylians that were not extensively sampled in either study, for data on occipital muscle attachments.

Acknowledgements

A part of the research presented in this article is based on the author's doctoral dissertation studies undertaken at the Department of Geology and Geophysics, Yale University, under the direction of J. Gauthier, whose comments on the relevant portion of the dissertation greatly improved the clarity of the manuscript. The author thanks the following people who helped to obtain extant specimens dissected for the study: R. Elsey (Rockefeller Wildlife Refuge, Louisiana Department of Wildlife and Fisheries), M. Calder, C. Marshall, J. Culwell, K. Culwell, P. Warney, J. Gauthier, M. Dickman, and W. Joyce. The author also thanks K. Zyskowski and G. Watkins-Colwell (YPM), M. Oshima, Y. Kato, and H. Taru (KPM), and J. Vindum (California Academy of Sciences) for allowing to dissect specimens under their care. I. Nishiumi (National Museum of Nature and Science, Tokyo) allowed to study a Casuarius casuarius skeleton under his care. The author is grateful to the following people for access to fossil specimens under their care: M. Norell and C. Collins (AMNH), A. Milner, P. Barrett, and S. Chapman (BMNH), H. Mayr (BSP), X.-C. Wu and K. Shepard (CMN), M. Maisch, A. Seilacher, and H. Lüginslands (GPIT), M. Watabe and S. Suzuki (HMNS), C. Schaff (MCZ), J. Horner, and P. Leiggi (MOR), R. Barsbold and K. Tsogtbaatar (MPC), P. Sheehan (MPM), R. Cifelli and J. Person (OMNH), R. Schoch (SMNS), D. Evans and V. Porter (TCMI), J. Gauthier, L. Murray, W. Joyce, D. Brinkman, and M. Fox (YPM).

Ancillary