The dinosaur feeding apparatus comprised a complicated network of jaw muscles that span an intricate assemblage of intracranial joints that link the bones of the skull together. Data from jaw muscles, such as their homologies, attachments, and sizes, are important to test hypotheses of cranial function and feeding behaviors such as powered cranial kinesis, bite force estimation, chewing behavior, and mechanical optimization. Not surprisingly, jaw muscle reconstruction is a common practice among paleobiologists. However, despite a long history of muscle reconstruction and soft-tissue inferences in dinosaur science, several challenges still impede the development of precise hypotheses of structure, function, and evolution of the jaw muscles in the group. Major cranial structures including the antorbital cavity (Witmer, 1997; Molnar, 2008), ceratopsid frills (Lull, 1908; Dodson, 1996), and buccal emarginations (Galton 1973; Knoll, 2008) have only relatively recently been identified as non-muscular structures. Regardless, the basic anatomy of the adductor chamber of dinosaurs has not been thoroughly reviewed in a comparative context and numerous issues still impact the interpretations of the region's evolution.
Many previous analyses of dinosaur jaw muscle anatomy framed their hypotheses on a broadly comparative framework (e.g., Dollo, 1884; Lull, 1908; Versluys, 1910; Russell, 1935; Janensch, 1936; Ostrom, 1961, 1966; Haas, 1963). With the advent of cladistic methodologies, it is possible to bring increased rigor permitted by phylogenetically constrained methods such as the Extant Phylogenetic Bracket approach (Bryant and Russell, 1992; Witmer, 1995a, 1997). In the past, workers reconstructed muscles based on mammalian anatomy (Dollo, 1884; Russell, 1935) and even included a muscular cheek in their reconstructions (Lull, 1908). Haas (1955), Ostrom (1961, 1964), and later Norman (1984) re-evaluated many earlier descriptions and included their own new insights on several clades of dinosaurs whereas drawing more from lepidosaur anatomy.
Soft-tissue inferences in fossils carry their own pitfalls and caution is necessary when reconstructing the musculature of extinct animals, particularly when their feeding behaviors may differ substantially from their extant relatives (Haas, 1969; Bryant and Seymour, 1990; Witmer, 1995a, b; Carrano and Hutchinson, 2002; Holliday and Witmer, 2008). Parsimony necessitates conservative inferences of jaw muscle anatomy. However, because archosaurs are so morphologically diverse and variation in jaw muscle morphology certainly exists among sauropsids (Haas, 1973; Schumacher, 1973; Buhler, 1981; Busbey, 1989; Vandenberge and Zweers, 1993; Iordansky, 1973; Wu, 2003; Holliday and Witmer, 2007), phylogenetic bracketing may still be inadequate for all soft-tissue inferences and in some cases, compelling evidence may override conservative interpretations.
Interpretations of muscle scars can be problematic (McGowan, 1979; Bryant and Seymour, 1990; Bryant and Russell, 1992; Rieppel, 2002) and correspondence of the muscle-bone relationship in the head has only been tested in a few cases (e.g., Montanucci, 1989, Witmer, 1995b, 1997; Hieronymus, 2002). Osteological correlates can indicate the presence, attachment site, and to some extent, the size of a jaw muscle's attachment. Tendinous or aponeurotic attachments leave more robust muscle scars than fleshy attachments, though fleshy attachments can be equally informative (Bryant and Seymour, 1990; Bryant and Russell, 1992, Carrano and Hutchinson, 2002). In general, osteological correlates of jaw muscles are common among extant sauropsids, although particular clades appear prone to having more well-developed correlates than others have. For example, the temporal bones of extant crocodylians offer a rich tapestry of fossae and ridges that consistently reflect the attachments of particular muscle bellies, their aponeuroses, and adjacent neurovasculature (Iordansky, 1973; Holliday and Witmer, in press). However, the temporal bones of lizards seemingly offer fewer tendinous correlates despite the presence of large aponeuroses, such as the quadrate aponeurosis and the bodenaponeurosis. Osteological correlates are highly susceptible to the vagaries of ontogeny and it is to be expected that juvenile individuals have fewer osteological correlates than more mature individuals do. In addition to these more discrete correlates, Hutchinson (2001a) introduced the concept of bone surface homology to refer to the correspondence of general osteological regions that are continuous through evolution. In the head, examples of homologous bone surfaces may include the lateral surface of the parietal and dorsal surfaces of the pterygoid, which m. adductor mandibulae externus profundus and m. pterygoideus dorsalis consistently attach to, respectively. However, homologous bony surfaces and their corresponding muscles are more difficult to identify on the quadrate, which underwent significant morphological and myological transformations in different sauropsid clades (Holliday and Witmer, 2007). These evolutionary changes can prove to be problematic in soft-tissue reconstruction.
Caveats aside, reconstructing jaw muscle anatomy remains an important task necessary to better understand cranial function and behavior in extinct animals. Osteological correlates have proven useful in reconstructing soft tissues and tracking their evolution in the cephalic (Witmer 1995a, b, 1997; Snively and Russell; 2007) and postcranial anatomy (e.g., Hutchinson, 2001a, b; Carrano and Hutchinson, 2002; O'Connor, 2006) of extinct archosaurs. Although many species may share numerous features with extant taxa, other derived groups, such as hadrosaurs and oviraptorosaurs, present a number of unique challenges. Numerous studies have focused on the jaw muscle anatomy of single taxa though no study has surveyed jaw muscle anatomy across the entire clade. To allay this gap, this article reviews the homologies and phylogenetic support of the different jaw muscles with a focus on those of non-avian dinosaurs. It uses the Extant Phylogenetic Bracket approach (Witmer, 1995a) to constrain inferences of soft tissues in fossil taxa and relies on the archosaur soft-tissue homologies provided by Holliday and Witmer (2007). It presents phylogenetically constrained rules of construction for jaw muscle anatomy and a generalized atlas of dinosaur adductor chamber anatomy, including the osteological correlates and their inferred soft-tissues. Tables 1 and 2 list anatomical and institutional abbreviations.
Table 1. List of anatomical abbreviations
Internal carotid artery
Ascending process of pterygoid
m. adductor mandibulae externus medialis
m. adductor mandibulae externus profundus
m. adductor mandibulae externus superficialis
m. adductor mandibulae posterior
m. depressor mandibulae
m. levator pterygoideus
m. protractor pterygoideus
m. pseudotemporalis profundus
m. pseudotemporalis superficialis
m. pterygoideus dorsalis
m. pterygoideus ventralis
Internal mandibular nerve and artery
Nerve to constrictor internus dorsalis muscles
Hyomandibular ramus of facial nerve
Palatine ramus of facial nerve
Table 2. List of institutional abbreviations
American Museum of Natural History
Academy of Natural Sciences
British Museum of Natural History
Bernard Price Institute
Canadian Museum of Nature
Cleveland Museum of Natural History
Denver Museum of Natural History
Dinosaur National Monument
Field Museum of Natural History
Mongolian Geological Institute
Museum of Comparative Zoology
Museum of the Rockies
Musée National du Niger
Royal Ontario Museum
Ministère de l'Energie et des Mines
Royal Tyrell Museum of Paleontology
Utah Museum of Natural History
United States National Museum
Zigong Dinosaur Museum
MATERIALS AND METHODS
Anatomical data were analyzed within the framework of the Extant Phylogenetic Bracket approach (EPB), which uses the principle of parsimony to infer soft tissues in extinct taxa (see Witmer 1995a, b, 1997; Carrano and Hutchinson, 2002). First, the adductor chambers of extant archosaur and reptile taxa were examined to identify patterns of topological similarity in jaw muscles and surrounding soft-tissues and to suggest hypotheses of soft tissue homology (Fig. 1). These patterns were identified from dissections, computed tomographic (CT) data, and observations of over 100 extant fresh and skeletonized crocodylian, avian, lepidosaurian, and testudine taxa described in Holliday and Witmer (2007). Second, causally associated osteological correlates of these soft tissues were identified in the skeletons of extant taxa (Fig. 1B–D). For example, in mammals, particular muscles (e.g., the temporalis muscle) consistently leave a fossa or crest on the skull (e.g., temporal fossa) that is used to infer the presence of that muscle when cadaveric specimens are unavailable. Last, presumably homologous osteological correlates of jaw muscles and relevant soft tissues were identified across a large assemblage of fossil archosaur material including basal archosaurs, crurotarsans, and non-avian dinosaurs (Fig. 1A, Table 3). These osteological data, which are the focus of this article, serve as proxies for the soft tissues in the fossil taxa and thereby form the congruence test of homology (Patterson, 1982; Witmer, 1997) used to infer the positions and homologies of jaw muscles and relevant neurovasculature using the phylogenetically based scoring procedure of the EPB.
Table 3. List of fossil non-avian dinosaur taxa studied and their identification numbers
CM 41681, DMNH 2818, USNM 4932, USNM 6645, USNM 2274
ROM 20892, ROM 1215
TMP 91.127.1, TMP 97.132.01
CMN 8887, CMN 8889
CMN 8528, MOR 699, MOR 1120, MOR 1157, MOR 1194, UCMP 113697, UCMP 137266, USNM 5740, USNM 4286, USNM 24216
CMN 2288, CMN 2289, MOR 003, ROM 658, USNM 4737, UCMP 130156
ROM 667, MOR 454
ROM 873, ROM 44770
The EPB relies on drawing anatomical inferences from not only fossil taxa, in this case non-avian dinosaurs, but also that clade's closest-related, extant “bracketing” taxa (birds and crocodylians) and finally outgroup taxa (lepidosaurs and testudines) (Fig. 1A). Given the above assumption that causally associated osteological correlates are homologous among extant and fossil taxa (i.e., the congruence test is sound), soft-tissue inferences can be refined further based on the phylogenetic distribution of the osteological correlates among fossil and extant clades. A soft tissue, in this case a muscle, can be interpreted as a level I, II, or III inference depending on the presence of the muscle's correlate in: both extant bracket taxa and the fossil taxon; one extant taxon plus the fossil taxon; or only the fossil taxon, respectively. Additionally, the reconstruction of the correlated soft tissue, along with consistent topological patterns recognized in the extant groups, can subsequently dictate where neighboring structures should have been (i.e., the maxillary nerve, neighboring muscle) and vice versa. If no clear osteological correlates are present among the extant bracket taxa despite the presence of the soft tissue, the levels of inference can then be assigned a “prime” designation (e.g., I′, II′, III′) in which each subordinate prime level is a weaker inference (i.e., it draws less phylogenetic support) than its predecessor in the hierarchy. Following Carrano and Hutchinson (2002), individual muscle attachments may vary in their distribution, therefore, whereas the cranial attachment of a muscle may have a Level I inference, its mandibular attachment may only be a Level 2 inference.
Overview of Adductor Chamber Osteological Correlates
Despite the relative ubiquity of fossae, crests, tuberosities, spurs, flanges, and other muscle-related bony structures in the skulls of dinosaurs and other reptiles (Figs. 2–4), these correlates are quite plastic and can vary among individuals of the same taxon, let alone among different taxa. This phenomenon is in part likely due to ontogenetic differences among individuals of a well-represented taxon. In juvenile Brachylophosaurus, the cranial correlate of m. levator pterygoideus is a fossa (Fig. 2A), whereas in an adult specimen (Fig. 2B), the correlate is a long, striated spur. As in Brachylophosaurus, this spur for m. levator pterygoideus may slightly overhang the ophthalmic groove, or as in Lambeosaurus (Figs. 2C and 5B), it may drape ventrally to then fuse onto the basisphenoid, forming a separate ophthalmic foramen. However, many adult individuals may also have different osteological correlates. Among Tyrannosaurus individuals, only one observed individual (AMNH 5117) possesses a crest in the temporal fossa (Fig. 5C) that marks the separation between m. pseudotemporalis superficialis rostrally, from m. adductor mandibulae externus profundus, caudally. Homologous crests are absent in MOR 555, MOR 1125, CMN 11841, and TMP 83.30.1. On the other hand, related taxa such as Nanotyrannus, individuals of Daspletosaurus (Fig. 3I), as well as other theropods (Fig. 3C,D) possess these crests. Finally, different taxa have arguably homologous, but incredibly morphologically dissimilar osteological correlates that preclude straightforward ranking systems. This problem is best represented by the coronoid attachments of m. adductor mandibulae externus profundus on the mandible. Among theropods, the correlate varies between a shallow depression, a smooth, flat area, or a small, rugose tuberosity, referred to here as the coronoid eminence (Fig. 4J,K). On the other hand, the osteological correlate takes the form of a huge boss in some taxa, such as Panopolosaurus (Fig. 4G). Finally, the coronoid processes of derived ceratopsids and hadrosaurs are characteristically dorsally elongate, may or may not have striations on them (Fig. 4). The loss of bony elements also influences interpretations of muscles and their osteological correlates. Several clades of dinosaurs lost the epipterygoid (Fig. 6), which is the origin of m. pseudotemporalis profundus. Just because the element was lost, does not necessarily mean the muscle was too, because birds also lost the epipterygoid but still have the muscle. However, epipterygoid loss creates a greater challenge when inferring whether the muscle shifted as in Brachylophosaurus (Fig. 2A,B) or if it is absent. Thus, these issues make scoring osteological correlates of muscles as a means of character mapping a complicated, if not largely subjective procedure.
Overview of Adductor Chamber Soft Tissue Anatomy
In general, jaw muscles are divided into groups based on their topological relationships with the branches of the trigeminal nerve (Lakjer, 1926; Holliday and Witmer, 2007). These muscle groups include m. constrictor internus dorsalis, and mm. adductor mandibulae internus, externus, and posterior, most of which are then further divided into smaller muscle bellies. With a few exceptions, such as extant crocodylians, the muscles maintain the same topological relationship with numerous other soft tissues among tetrapods (Holliday and Witmer, 2007; Holliday and Witmer, in press). It is expected that these same topological patterns were present in dinosaurs as well and there are no data to suggest otherwise. Therefore, if the grooves or foramina for the ophthalmic nerve (V1) and stapedial (aST) and internal carotid (aIC) arteries, which, for example, circumscribe the dorsal and caudal margins of the origin of m. protractor pterygoideus on the basisphenoid, reasonable inferences of the muscle's attachment location can be made, regardless of the presence of a clear osteological correlate (Fig. 5). Similarly, the mandibular nerve (V3) and artery (aMN), always pass between mm. adductor mandibulae externus and posterior in the temporal region, and then once they are in the mandibular fossa, they pass lateral and rostral to m. adductor mandibulae posterior. The mandibular neurovascular bundle then gives off lateral and medial (i.e., mylohyoid nerve) mandibular branches. Thus, if the position of the muscle can be inferred (which is consistently the medial mandibular fossa), the location of the neurovascular bundle can also be inferred regardless if they leave a groove or foramen. Finally, in all non-crocodylian sauropsids, the maxillary nerve (V2) passes between mm. pseudotemporalis superficialis and adductor mandibulae externus profundus before turning rostrally toward the suborbital region. In the Triceratops braincase, MOR 699, the maxillomandibular foramen is situated directly ventral to the laterosphenoid buttress, which is rather rostral to where the foramen is in many dinosaurs (Fig. 5A). However, the foramen is angled caudally and has a long groove on its caudal edge, suggesting that the maxillary (and mandibular) nerves are pushed caudally by m. pseudotemporalis superficialis, which otherwise has no clear osteological correlate on the braincase. Thus, nerve foramen morphology can help identify the presence of particular muscles. These examples illustrate how neurovascular tissues, muscles, and their osteological correlates can reciprocally illuminate one another.
In addition to topological patterns, the regions of attachment can also be used to categorize the different muscle groups into the temporal muscles (e.g., mm. pseudotemporalis superficialis and adductor mandibulae externus profundus), palatal muscles (e.g., mm. pterygoideus and adductor mandibulae posterior), and the jaw opening muscle m. depressor mandibulae. Many sauropsids, including most dinosaurs, also maintain m. constrictor internus dorsalis muscles that span the orbitotemporal region between the braincase and the palate (Haas, 1973; Rieppel, 2002; Holliday and Witmer, 2008). This organization will be used in the “Results” section later in which each muscle name will be followed by its abbreviation and general levels of inference for its origin and insertion in the format: Muscle name (abbreviation-origin/insertion). Table 4 lists the hypotheses of muscle homologies from Holliday and Witmer (2007), the muscles' general attachments and their respective EPB inference level for their origin and insertion.
Table 4. The jaw muscles of extant and extinct sauropsids coupled with each muscle attachment's general level of inference in non-avian dinosaurs
Level of inference
m. tensor periorbitae
Rostral border of orbit
m. levator pterygoideus
Laterosphenoid dorsal to ophthalmic foramen/groove
Medial surface of pterygoid and epipterygoid
m. protractor pterygoideus
Ala basisphenoid ventral to ophthalmic foramen/groove
Medial surface of pterygoid and quadrate
m. pterygoideus dorsalis
Dorsal surface of rostral portion of pterygoid and palatine
Medial surface of articular
m. pterygoideus ventralis
Caudoventral surface of pterygoid
Lateral surface of articular and surangular
m. pseudotemporalis profundus
Lateral surface of epipterygoid
Medial surface of surangular/medial mandibular fossa
m. pseudotemporalis superficialis
Rostromedial portion of temporal fossa
Medial surface of coronoid eminence/rostral medial mandibular fossa
m. adductor mandibulae externus profundus
Caudomedial portion of temporal fossa
m. adductor mandibulae externus medialis
Caudolateral portion of temporal fossa
Coronoid eminence/dorsomedial surface of surangular
m. adductor mandibulae externus superficialis
Medial surface of upper temporal bar
Dorsolateral surface of surangular
m. adductor mandibulae posterior
Lateral surface of quadrate
Medial mandibular fossa
M. tensor periorbitae (mTP–Level I/II′).
M. tensor periorbitae (i.e., m. levator bulbi) attaches to the rostral edge of the prootic in lepidosaurs and a prominent crest on the lateral surface of the laterosphenoid in crocodylians and birds (Haas, 1973; Elzanowski, 1987). The muscle then attaches to the rostral portion of the orbital septum in birds and lepidosaurs and to the preotic pillars in crocodylians, forming a sling under the orbital contents in all taxa. All dinosaurs possess the caudal osteological correlate of m. tensor periorbitae: the laterosphenoid buttress, or antotic crest, making its reconstruction a Level 1 inference. This sharp edge of the laterosphenoid is the attachment of the muscle and clearly separates the temporal region from the orbital region (e.g., Fig. 2). In some cases, the muscle creeps onto the rostral surface of the laterosphenoid, as in Majungasaurus (Fig. 2J). On the other hand, correlates for the muscle's rostral attachment are yet to be found.
M. levator pterygoideus (mLPt–III/III).
In lepidosaurs, m. levator pterygoideus originates on the lateral surfaces of the parietal and prootic (e.g., Lakjer, 1926; Haas, 1973). This small muscle passes ventrally to insert on the ventromedial surface of the epipterygoid and is thought to help modulate movements of the palate during feeding, particularly in taxa that express cranial kinesis (Holliday and Witmer, 2008). The muscle is absent in birds and crocodilians, although likely because of different phenomena. Both extant archosaur clades lost the epipterygoid, but crocodilians sutured the palate to the braincase, thereby eliminating the orbitotemporal space altogether. Some avian taxa occasionally possess a short ligament between the braincase and the quadrate that may be a rudiment of m. levator pterygoideus or the epipterygoid (Dzerzhinsky and Yudin, 1982). M. levator pterygoideus does not leave any consistent osteological correlates on the cranium or palate in lepidosaurs.
Reconstructions of m. levator pterygoideus's cranial and palatal attachments are Level III or III′ inferences, the weakest possible soft-tissue inference (Table 4). Regardless, osteological and topological data, and support from lepidosaurian outgroups, indicate many non-avian dinosaurs likely possessed the muscle. Numerous dinosaur taxa possess clear osteological correlates attributable to this muscle on the laterosphenoid. Cranial osteological correlates of m. levator pterygoideus can be identified in ceratopsids, ornithopods, some neosauropods, and tyrannosaurs (Fig. 3). However, clear correlates cannot be easily identified in thyreophorans and deinonychosaurs (Fig. 2L). The muscle's scars are morphologically diverse ranging from small fossae (Fig. 2A) to large tuberosities (Fig. 2K). Despite these differences in shape, the correlates are always found just dorsal to the ophthalmic groove or lateral to the ophthalmic foramen when present, and ventral to the epipterygoid cotyle on the laterosphenoid. The only soft tissue that occupies this region of the braincase in extant taxa is m. levator pterygoideus. Thus, despite weak phylogenetic support from extant bracketing taxa, the topological and osteological support for inferring the presence of the muscle is strong. The levator pterygoideus muscle does not leave any robust osteological correlates in most dinosaur taxa. However, in ceratopsids and hadrosaurs, a large, striated flange arises dorsally from the pterygoid that suggests the presence of a large, tendinous levator pterygoideus attachment (Fig. 3G–I).
M. protractor pterygoideus (mPPt–II/II′).
M. protractor pterygoideus attaches to the ventrolateral surface of the basisphenoid in lepidosaurs or parabasisphenoid in birds ventral to the ophthalmic groove and rostrolateral to the internal carotid foramen. The muscle only occasionally leaves a small crest or spur on the braincase of birds, but in lepidosaurs, the muscle occupies the triangular ala basisphenoid, whose border is excavated by the surrounding neurovasculature such as the stapedial and internal carotid arteries (Fig. 5). Crocodyliforms lost m. protractor pterygoideus once the palate was firmly sutured to the braincase (Fig. 1D). However, rauisuchians, such as Saurosuchus, also have an ala basisphenoid suggesting these taxa possessed the muscle (Holliday and Witmer, in press). Reconstructions of the protractor pterygoideus musculature are Level II inferences for the braincase origin. Like that of lizards, the dinosaur ala basisphenoid is roughly triangular but varies in morphology ranging from small tubercles or flanges in ankylosaurs and deinonychosaurs (Fig. 2L) to large rugose bosses such as those found in tyrannosaurs, hadrosaurs, and ceratopsids (Fig. 2). During maniraptoran evolution, the ala basisphenoid regresses leaving no easily discernable cranial correlate for m. protractor pterygoideus on the braincase, shifting the cranial attachment site inference from a Level II to II′. Whereas the dromaeosaurs Velociraptor and Tsaagan (Fig. 2L) have a small ala basisphenoid, the troodontid Saurornithoides, and basal bird Archaeopteryx both lack the feature. Because extant birds still possess this muscle and some taxa do have small bony spurs indicative of the muscle's attachment, it is reasonable to infer that their closest theropod ancestors also possessed the muscle. The protractor muscle attaches along the medial surfaces of the pterygoid and quadrate in extant taxa and likely does the same in dinosaurs. However, the muscle does not leave any marked scars indicative of its attachment to the pterygoid or quadrate making the palatal insertion a Level II′. The lack of a palatal correlate of m. protractor pterygoideus suggests a fleshy attachment was present. But because the middle ear cavity, which directly abuts the protractor muscle, also leaves similar correlates on the medial surface of the quadrate and pterygoid, it remains difficult to determine the expanse and size of either the air sac or muscle. Finally, the ala basisphenoid is often excavated by a neurovascular groove and is particularly prominent in large taxa such as Triceratops, Brachylophosaurus, and Tyrannosaurus (Fig. 4). Ostrom (1961) inferred this groove in hadrosaurs to be for the maxillary and mandibular nerves as they coursed ventrally from the maxillomandibular foramen. However, when the palate is articulated with the braincase, this groove lies well medial and ventral to the pterygoid, a position quite different from the position where the nerve is found in other sauropsids. These grooves on the ala basisphenoid are most likely for the neurovascular bundle (nCID) that supplies the protractor muscles and other constrictor internus dorsalis muscles.
M. pseudotemporalis profundus (mPSTp–I/I').
M. pseudotemporalis profundus attaches to the lateral surface of the epipterygoid in lepidosaurs (Fig. 1B), the lateral bridge of the laterosphenoid in crocodylians, and the orbital process of the quadrate in birds (Vandenberge and Zweers, 1993; Holliday and Witmer, 2007). The muscle then attaches along the dorsomedial surface of the mandible in lepidosaurs and birds. In crocodylians, the muscle is vestigial and variably merges with the fibers of the medial surfaces of the temporal muscles. There are no clear osteological correlates of the muscle's attachment to the mandible, whereas those to the cranium are fairly robust. Reconstructions of m. pseudotemporalis profundus are level I inferences in dinosaurs that have epipterygoids. However, because crocodylians and birds eliminated the epipterygoid and shifted the position of the muscle (for reviews see Holliday and Witmer, 2007, 2008, in press) interpretations of dinosaur anatomy may not be clear. Among dinosaurs, the muscle likely attached to the lateral surface of the epipterygoid, as in lizards, and the surface of the bony element in large theropod taxa occasionally has a fossa suggesting the attachment of this muscle. However, several clades of dinosaurs also lost the epipterygoid including hadrosaurs, ceratopsids, and sauropods (Fig. 6). The position of the origin of m. pseudotemporalis profundus in ceratopsids and sauropods is unclear and it is possible that they lost the muscle. However, ornithopods (e.g., Figs. 2A,B and 5B) often have a fossa present between the inferred origins of mm. levator pterygoideus and pseudotemporalis superficialis that may be for m. pseudotemporalis profundus. Mandibular attachments of the muscle are ill-defined and it is inferred that, like those attachments of extant taxa, the muscle likely attached along the medial surface of the coronoid process or surangular.
M. pterygoideus dorsalis (mPTd–I/I).
M. pterygoideus dorsalis originates along the dorsal surfaces of the pterygoid and palatine bones in lepidosaurs, crocodylians, and birds (Witmer, 1995b, 1997). The muscle then inserts onto the medial surface of the articular and retroarticular process. The rostral extent of the palatal attachments of m. pterygoideus dorsalis can be difficult to discern from the equally smooth and excavated surfaces left by the nasal passages and paranasal air sinuses in dinosaur taxa (Witmer, 1997). Occasionally, a small crest on the inner surface of the maxilla or lateral surface of the palatine may demarcate the shift from a muscle attachment caudally to the caviconchal recess and antorbital cavity rostrally. However, these structures are rare or often distorted in fossil taxa. The mandibular attachments of m. pterygoideus dorsalis are well supported by the common presence of a smooth excavation along the medial surface of the retroarticular process of the articular in dinosaurs. Therefore, both cranial and mandibular reconstructions of the m. pterygoideus dorsalis are Level I inferences.
M. pterygoideus ventralis (mPTv–I/I).
M. pterygoideus ventralis attaches along the caudoventral edge of the pterygoid in crocodylians and birds. Among crocodylians, the muscle conspicuously wraps around the retroarticular process to attach on the ventrolateral surface of the retroarticular process and surangular. The muscle often attaches along the ventral edge of the retroarticular process but also wraps around the retroarticular process in some birds to attach to the lateral surface of the mandible and even attaches to the jugal in some parrots (Hofer, 1950). It remains unclear if lepidosaurs possess a clear, separate ventral belly of m. pterygoideus. Regardless, the muscle occasionally also wraps around to the lateral surface of the mandible and in some taxa (e.g., Uromastix) it attaches to the lower temporal bar.
Among dinosaurs, the attachment of m. pterygoideus ventralis to the ventral edge of the palate is a phylogenetically supported inference although clear palatal anatomical correlates are rare (e.g., Fig. 3G) making the origin a Level I inference. On the other hand, the mandibular insertion of m. pterygoideus ventralis is well-supported by a smooth fossa on the ventrolateral surface of the mandible making it a clear Level I inference. Like the muscle's attachment in crocodilians and many birds, the muscle also likely wrapped around the retroarticular process to attach to the lateral surface of the mandible (Fig. 4). However, the extent of the muscle's attachment across the mandible is difficult to determine in most dinosaur taxa. Among most ornithischians and sauropods, there is no clear demarcation between the attachment of m. adductor mandibulae externus superficialis and m. pterygoideus ventralis, though a few individuals have a fossa for the former muscle (Fig. 4C) suggesting it does not extend very far ventrally. Among theropods, the ventral surface of the mandibular shelf likely marks the dorsal extent of the pterygoid muscle (Figs. 4J and 7C). Interestingly, both derived hadrosaurs and tyrannosaurs often possess flanges and bony spurs that descend off of the jugal and tend to point toward the retroarticular process. In some cases (e.g., Brachylophosaurus), the spurs are rugose and striated suggesting a tendinous attachment. These morphologies suggest that m. pterygoideus ventralis may actually have attached to the jugal, rather than simply the mandible (Fig. 7).
M. adductor mandibulae posterior (mAMP–I/I).
M. adductor mandibulae posterior is the most phylogenetically and anatomically consistent muscle in the adductor chamber. Among extant taxa, the muscle attaches across the lateral surface of the quadrate. In crocodylians, the muscle leaves a number of crests and fossae (Iordansky, 1973; Holliday and Witmer, 2007) marking its aponeurotic and fleshy attachments whereas among birds and lizards, correlates are rare (Fig. 1). Osteological correlates of the muscle on the quadrate are also rare among dinosaurs (Fig. 3). Regardless, data from extant bracketing and outgroup taxa suggest that the muscle also attached to the quadrate in these fossil taxa. M. adductor mandibulae posterior consistently attaches to Meckel's cartilage within the medial mandibular fossa in lizards, crocodylians, and birds, other than anseriforms, where it attaches to the lateral surface of the mandible. All data from dinosaurs suggest that m. adductor mandibulae posterior also filled the medial mandibular fossa making it a Level I inference (Fig. 4).
The temporal region is dominated by the vertically oriented jaw closing muscles including m. pseudotemporalis superficialis, which is a division of m. adductor mandibulae internus and bellies of m. adductor mandibulae externus including mm. adductor mandibulae externus profundus, medialis, and superficialis. Significant changes occurred in the organization of the temporal muscles during archosaur evolution and the muscles' relationships with the skull in crocodilians, birds, and non-avian dinosaurs are not necessarily similar to those of lepidosaurs. As will be discussed thoroughly later, the organization of the temporal muscles changed in the skull roof as well as on the mandible in archosaurs. These changes create challenges when identifying homologies as well as determining the attachments, function, and evolution of the muscles in fossil taxa. The following passages will document the temporal muscles from superficial (i.e., m. adductor mandibulae externus superficialis) to deep (i.e., m. pseudotemporalis superficialis; Figs. 2, 3, 8).
M. adductor mandibulae externus superficialis (mAMES–I/I).
M. adductor mandibulae externus superficialis invariably attaches across the upper temporal bar in lepidosaurs via a fleshy attachment that does not leave any specific osteological correlate. The muscle then attaches to the dorsolateral surface of the mandible and only occasionally leaves a smooth, ovate region that may be bordered laterally by a faint ridge marking the transition from muscle to integument attachment. In crocodilians, the muscle attaches along the ventral surface of the quadratojugal by means of a fleshy attachment laterally and aponeurotic attachment medially, which leaves a rostroventrally oriented ridge that it shares with m. adductor mandibulae posterior (Iordansky, 1973; Schumacher, 1973). The mandibular attachment of the crocodylian m. adductor mandibulae externus superficialis is well-emarginated by a smooth region on the dorsal surface of the surangular that is bounded rostrally by the coronoid eminence left by m. adductor mandibulae externus profundus. Because birds lost the upper temporal bar, m. adductor mandibulae superficialis generally attaches across the lateral surface of the squamosal and when present, the ventral surface of the postorbital process (e.g., Anas), but also across the suprameatal shelf in many others (e.g., Gallus, Larus, Phalacrocorax; Fig. 1B). In ratites (e.g., Struthio), m. adductor mandibulae externus superficialis is greatly reduced and attaches along the temporal fascia and only slightly to the lateral edge of the postorbital process. The muscle often leaves small crests marking its aponeurotic attachments on the cranial surface in most birds.
As in extant taxa, the temporal bar is arguably the best indicator for the attachment of m. adductor mandibulae externus superficialis. Taxa with long upper temporal bars likely had rostrocaudally broader muscles; those with shorter bars have smaller muscles. Despite this basic anatomical relationship with the skull, the muscle does not leave clear osteological correlates. In taxa that have rugose bone textures on the skull surface (e.g., tyrannosaurs, abelisaurids), the muscle attachment is easily recognizable because of its contrasting smooth texture. Thus, its reconstruction is a well-supported, Level I inference. However, the medial, or deep, extent of the muscle is difficult to estimate because it does not usually attach to bony structures that mark its boundaries with deeper muscles such as mm. adductor mandibulae externus medialis or profundus caudomedially, or m. pseudotemporalis superficialis rostromedially.
The most problematic concern of reconstructions of m. adductor mandibulae externus superficialis's origin is that despite the muscle's likely attachment across the medial surface of the squamosal and postorbital, basically all of the medial surfaces of the postorbital, postfrontal, squamosal, jugal, and quadratojugal are smooth because of the numerous soft-tissues that emarginate the region including fascia, neurovascular bundles, periorbital structures, and pneumatic diverticulae. Haas (1963) interpreted m. adductor mandibulae externus superficialis to attach along the postorbital bar, rather than the upper temporal bar in Diplodocus based on his evaluation of the smooth, slightly grooved medial surface of the bone. However, this groove is more likely excavated by postorbital or jugal vessels from the stapedial artery that pass along the medial surfaces of the postorbital and jugal. Among theropod dinosaurs that have a prominently excavated squamosal recess, the interface between the pneumatic sinus and muscle on the squamosal may mark the muscle's caudal boundary (Witmer and Ridgely, 2008). However, it is also possible that m. adductor mandibulae externus superficialis simply enclosed the medial portions of the sinus, which likely derives from the suborbital diverticulum (Witmer and Ridgely, 2008). In taxa that do not have well-developed diverticular correlates, although their phylogenetic relationships support the structure's reconstruction, the inference of a muscle attachment versus pneumatic structure is equivocal. Finally, many lepidosaurs also have a small superficial muscle belly, m. levator anguli oris, which is often associated with m. adductor mandibulae externus superficialis (Haas, 1973). The muscle originates on the pretemporal and lateral temporal fasciae and then inserts on the rictus (i.e., corner of the mouth; Haas, 1973). Extant archosaurs do not possess this muscle nor does the muscle leave osteological correlates on the skeleton. Thus, the reconstruction of m. levator anguli oris is a Level III′ inference in dinosaurs.
In most lepidosaurs, the m. adductor mandibulae externus superficialis attaches along the dorsolateral surface of the surangular, caudal to the coronoid process. The muscle is typically a fleshy, parallel-fibered belly and occasionally leaves a fossa or a slight bony ridge (Fig. 1B). In crocodylians, the muscle occupies the majority of the dorsal surface of the surangular, caudal to the coronoid eminence and leaves a faint ridge marking its extent laterally. The muscle's mandibular attachment varies among birds and generally attaches along the lateral surface of the mandible between the jaw joint caudally, the attachment of m. adductor mandibulae externus profundus rostrally and the attachment of m. pterygoideus ventralis ventrally.
As in extant taxa, the dorsolateral surface of the surangular is the most likely attachment of m. adductor mandibulae externus superficialis among non-avian dinosaur taxa and the muscle does not lend many specific identifiable correlates other than a smooth region of the surangular. Haas (1955) reconstructed the muscle in Protoceratops attaching across the lateral surface of the surangular based on the smooth, shallow fossa on the element and this is a reasonable inference. In basal (e.g., Leptoceratops, CMN 8889) and derived ceratopsids (e.g., Styracosaurus CMN 344), the smooth surangular becomes incorporated into the caudal portion of the large, rugose coronoid process suggesting a marked difference in muscle attachment type between the two structures (Fig. 4G). The caudal, smooth surface is most likely the attachment for m. adductor mandibulae externus superficialis (e.g., Ostrom, 1964). The dorsal surface of the surangular forms the caudal portion of the coronoid process in basal ornithopods such as Thescelosaurus (ROM 3587; Fig. 4A) and Dryosaurus (CM 3392), and there is occasionally a slight shallow fossa long the dorsal surface of the element that is the correlate for m. adductor mandibulae externus superficialis. However, in derived ornithopods, the attachment of m. adductor mandibulae externus superficialis is rarely demarcated (e.g., Fig. 4C), but still most likely attaches along caudodorsal surface of the surangular, just rostral to the jaw joint (e.g., Ostrom, 1961; Rybczynski et al., 2008). In sauropods, the dorsal edge of the surangular is often smooth and acutely angled suggesting a very thin mandibular attachment of m. adductor mandibulae externus superficialis. There are no clear, consistent osteological correlates of the muscle on the lateral surface of the mandibulae in Camarasaurus or Diplodocus (Fig. 5B). Non-avian theropods have the most clearly defined mandibular attachments for m. adductor mandibulae externus superficialis because the surangular has a prominent shelf on its lateral edge marking the lateral extent of the muscle (Figs. 5 and 8).
M. adductor mandibulae externus medialis (mAMEM–I/I′).
M. adductor mandibulae externus medialis is the most problematic of the temporal muscles in several ways. For reasons discussed later in “Discussion” section, it is difficult to interpret in the adductor chamber of extant reptiles and its reconstruction in fossil taxa is equally ambiguous. The muscle is large and well-differentiated in lepidosaurs and attaches to the lateral surface of the large temporal aponeurosis, the bodenaponeurosis (Haas, 1973). The muscle attaches along the caudal surface of the dorsotemporal fossa and may leave a shallow fossa on the surface of the posttemporal bar, lateral to the spur often left by the bodenaponeurosis. However, its attachment to the mandible is unclear because it may share a common attachment with mm. pseudotemporalis superficialis and adductor mandibulae externus profundus. In birds and crocodylians, the muscle is often anatomically and always topologically indistinguishable from mm. adductor mandibulae externus profundus and superficialis (Holliday and Witmer, 2007). In crocodylians, m. adductor mandibulae externus medialis is a small, quadrangular muscle that occupies a smooth region of the quadrate between the trigeminal foramen and m. adductor mandibulae posterior, with which it shares aponeurotic attachments. However, the muscle melds with parts of mm. adductor mandibulae externus profundus and superficialis as it attaches to the mandible leaving no specific osteological correlate. Among birds, many of which have divided m. adductor mandibulae externus into more than three, simple bellies, a discrete m. adductor mandibulae externus medialis is difficult to separate from deep or superficial portions without corroborating developmental evidence. That said, these intermediate bellies of m. adductor mandibulae externus typically attach to the temporal fossa, or to fasciae or aponeuroses of the other temporal muscles cranially and then to the coronoid region of the mandible.
Reconstructions of the cranial and mandibular attachments of m. adductor mandibulae externus medialis are Level I′ inferences among non-avian dinosaurs. However, these inferences are problematic, because interpretations of attachments of m. adductor mandibulae externus medialis depend on interpretations of where the surrounding muscles attach, which typically leave correlates that are more consistent. The mandibles of hadrosaurs and ceratopsids (Fig. 4B, D, E) occasionally have shallow fossae that pass between the coronoid process and the jaw joint, a region that was certainly bounded by mm. adductor mandibulae externus profundus and superficialis, respectively. It could be expected that m. adductor mandibulae externus medialis occupied this intermediate position. However, this inference is equal to the hypothesis that the muscle attached to the coronoid process with m. adductor mandibulae externus profundus.
M. adductor mandibulae externus profundus (mAMEP–I/I).
Musculus adductor mandibulae externus profundus is the deepest portion of the adductor mandibulae externus muscle group. In lepidosaurs, it is relatively small compared with the other temporal muscles and attaches to the caudomedial corner of the dorsotemporal fossa and a portion of the prootic, somewhat deep to mm. pseudotemporalis superficialis and adductor mandibulae externus medialis. It attaches laterally to the bodenaponeurosis and then prominently attaches to the coronoid process. In crocodylians, m. adductor mandibulae externus profundus is the only muscle that occupies the dorsotemporal fossa, though temporal osteological correlates of basal crocodylomorphs show evidence of there being multiple muscles in the fossa (Holliday and Witmer, 2007, in press). In crocodylians, the muscle attaches to the caudomedial corner of the dorsotemporal fossa and then attaches as a tendon to a characteristic rugosity on the dorsal surface of the surangular rostral to m. adductor mandibulae externus superficialis. In ratites and anseriforms, the muscle attaches to the postorbital process and in galliforms and most other birds, the muscle attaches to the temporal fossa. The muscle then consistently attaches to the coronoid process in all birds.
Among dinosaurs, m. adductor mandibulae externus profundus most likely occupied most of the caudomedial portion of the dorsotemporal fossa and likely attached to the sagittal and nuchal crests when present. However, as noted earlier, its caudolateral border, which would have been shared with mm. adductor mandibulae externus medialis or superficialis, is unclear. The muscle's rostral bony attachment is occasionally marked by a small spur or vertical crest in a few taxa (e.g., Carcharodontosaurus, Daspletosaurus; Fig. 3D,F), or a subtle shift in curvature of the dorsal edge of the fossa (e.g., Herrerasaurus, Hypsilophodon, Brachylophosaurus; Figs. 2 and 3). However, in most taxa, there are no definable osteological correlates other than the dorsotemporal fossa itself that indicate the rostral extent of m. adductor mandibulae externus profundus. In Edmontosaurus (CMN 2289; Fig. 3A), there is a shallow fossa on in the ventral portion of the temporal fossa that may correspond to this muscle, which would then suggest that m. adductor mandibulae externus medialis may extend across the dorsal edge of the dorsotemporal fossa, as in lizards. However, this is the only specimen observed that bears this feature. That said, cranial reconstructions of the m. adductor mandibulae externus profundus are Level I inferences. Most non-avian dinosaurs also possess mandibular osteological correlates of the muscle including coronoid processes in most ornithischians and smaller coronoid eminences in theropods and thus, these the muscle's mandibular attachments are also Level I inferences.
M. pseudotemporalis superficialis (mPTs–I/I′).
The deepest and most rostral temporal muscle is m. pseudotemporalis superficialis. In lepidosaurs, m. pseudotemporalis superficialis attaches to the medial surface of the dorsotemporal fossa. The muscle then attaches to the medial portion of the coronoid region. During crocodyliform evolution, m. pseudotemporalis superficialis shifted from a position similar to that present in lepidosaurs to the caudal surface of the laterosphenoid and does not attach within the dorsotemporal fossa proper in extant crocodylians (Holliday and Witmer, 2007, in press). The muscle then inserts on the rostral portion of the medial mandibular fossa where it is compressed by the pterygoid buttress and develops a sesamoid cartilage, the cartilago transiliens. In ratites other than Apteryx, m. pseudotemporalis superficialis solely occupies the dorsotemporal fossa and also attaches in the medial mandibular fossa via an intertendon. In virtually all other birds, the muscle is miniscule relative to other jaw muscles and attaches along caudoventral edge of the laterosphenoid buttress of the temporal region and then to the medial surface of the coronoid region of the mandible.
Reconstructions of m. pseudotemporalis superficialis are Level I or I′ inferences (Table 4). Despite this strong phylogenetic support, the position of m. pseudotemporalis superficialis is difficult to clearly identify among many dinosaurs. Drawing from the muscle's position in lepidosaurs, many early studies considered it to be the dominant muscle of the dorsotemporal fossa. Lull (1908) reconstructed the muscle on the rostral surface of the laterosphenoid, inside of the orbit, in Protoceratops. However, as noted by Haas (1955) this was a novel, and unsupported interpretation and the muscle likely occupied part of the dorsotemporal fossa. Several authors have suggested that m. pseudotemporalis superficialis attached to a large excavation on the dorsal surface of the frontal in many theropod taxa (e.g., Coria and Currie, 2002; Molnar, 2008). However, the shape, horizontal orientation, surrounding bony structures, and the phylogenetic distribution of this structure suggest that a muscle did not likely attach to it (Holliday, 2008; unpublished data) and phylogenetic bracketing further supports that m. pseudotemporalis superficialis more likely attached within the dorsotemporal fossa proper. However, like m. adductor mandibulae externus profundus, m. pseudotemporalis superficialis does not commonly leave specific osteological correlates on the skull other than the characteristic smooth area of rostrolateral portion of the dorsotemporal fossa (Fig. 3). As noted in the earlier section, some individuals may have a crest or curvature shift on the temporal fossa that indicates the separation between m. adductor mandibulae externus profundus and m. pseudotemporalis superficialis (Fig 3). The rostral extent of the muscle is bounded by the laterosphenoid buttress and the attachment of m. tensor periorbitae.
The mandibular attachment of m. pseudotemporalis superficialis is difficult to infer. The extant bracketing taxa suggest that the muscle probably attached to the rostral portion of the medial mandibular fossa in ankylosaurs, stegosaurs, pachycephalosaurs, and basal ceratopsids and iguanodontians. However, m. pseudotemporalis superficialis likely attached to the medial surface of the coronoid process in many derived iguanodontians and ceratopsids. The derived ornithischian condition—a coronoid attachment—is hypothesized to be similar to that present in lepidosaurs because in hadrosaurs and ceratopsids, the medial mandibular fossa is very small and significantly caudal to the coronoid process thereby giving the muscle a rostral moment, rather than the caudovertical moment it provides in other taxa. Furthermore, the coronoid processes of these taxa tend to have different domains of striations on them, which suggest the presence differently oriented muscle attachments (Fig. 4B,E). In theropods and sauropods, which have large, rostrally extended medial mandibular fossae that are very similar to those found in crocodylians and ratites, inferring the muscle to attach in the rostral portion of the fossa is a Level I inference.
Jaw muscle inferences offer a potential wealth of phylogenetic, anatomical, and functional information. However, there are a number of vagaries associated with them and the distribution of osteological correlates on the skull can differ among individuals as well as clades. On the other hand, specific osteological correlates, and their soft-tissue inferences, may be phylogenetically robust. For example, with the exception of extant crocodylians, which have significantly reorganized their jaw muscles, m. adductor mandibulae posterior is a relatively simple, parallel-fibered muscle among extant sauropsids. Therefore, from a phylogenetic standpoint, it is reasonable to infer that the muscle was also parallel-fibered among non-avian dinosaurs. This inference is anatomically supported by the lack of crests and other osteological correlates that would suggest tendinous attachments on the quadrates. The temporal muscles are more pinnate than m. adductor mandibulae posterior, and thus, a pinnate m. adductor mandibulae externus profundus or m. pseudotemporalis superficialis is a relatively strong inference that is supported by the extant phylogenetic bracket as well as at least a few individuals that have crests suggesting strong aponeuroses. However, these inferences also rely on a consistent distribution of correlates, which given the fragmentary nature of skull specimens, is still a challenge. These issues aside, there are several trends in dinosaur jaw muscle anatomy and evolution that can be explored.
Protractor muscles are key structures involved in cranial kinesis, or the ability of an animal to move the palate independent of the braincase and mandible. However, to what extent dinosaurs could display this complicated behavior remains debatable (Holliday and Witmer, 2008). Regardless as to whether the protractor muscles are functional or evolutionary relics, they are ubiquitous among dinosaurs. For example, compared with those of basal ornithopods, the ala basisphenoid of hadrosaurs have a greatly expanded tripartite morphology suggesting an enlarged and modified muscle that had pronounced tendinous attachments to the braincase that radiated not only caudoventrally, but also rostrolaterally onto the palate. These data suggest m. protractor pterygoideus may have become hypertrophied to resist laterally oriented forces generated by the medially inset dental batteries that derived ornithopods evolved (Rybczynski et al., 2008). Similarly, the alae basisphenoid of many non-coleurosaurian theropods, such as Herrerasaurus, Allosaurus, and Majungasaurus are smooth, triangular pendants that are attachments for a modest, fleshy, pinnate m. protractor pterygoideus. However, those of large tyrannosaurs become dorsoventrally elongate, highly textured, and covered with numerous crests that point ventrally or caudoventrally suggesting an increase in aponeurotic attachment for m. protractor pterygoideus.
These changes in morphology of the ala basisphenoid may be functionally adaptive, but they may also simply be the results of the evolution of large head size in tyrannosaurs, ceratopsids and hadrosaurs. Similarly, the disappearance of the ala basisphenoid during maniraptoran evolution may be associated with trends in miniaturization displayed in the clade [e.g., Sinosauropteryx, (Currie and Chen, 2001); Bambiraptor, (Burnham, 2004); Sinovenator (Xu et al., 2002); Microraptor, (Xu et al., 2003); Mei, (Xu and Norell, 2004), and Mahakala, (Turner et al., 2007)]. Whereas adductor muscles are known to scale positively with body size in lizards (Herrel and O'Reilly, 2005; Herrel et al., 2007), it remains to be determined if protractor muscles scale similarly with other parts of the feeding apparatus among extant and fossil taxa. Neosauropods attained huge body sizes without the accompanying increase in head size and reduced their alae basisphenoid to thin flanges of bone that suggest they had a simple soft-tissue septum that separated the orbit from the ear. However, these taxa possess also some of the most extreme head and feeding structures found in dinosaurs (e.g., Nigersaurus, Diplodocus). Thus, teasing out the differences between allometry and functional significance in the system is challenging, but it offers numerous new directions to explore.
Muscles that directly attach to the bony surfaces of the temporal region, such as mm. pseudotemporalis superficialis and adductor mandibulae externus profundus, are more likely to leave discernable osteological correlates that support anatomical inferences. However, muscles that instead primarily attach to soft tissues such as skin that covers the lateral temporal fenestrae, such as m. levator anguli oris, or large aponeuroses, such as m. adductor mandibulae externus medialis, will not leave bony evidence of their attachments. This presents a major problem with reconstruction of structure in the temporal region because m. adductor mandibulae externus medialis was likely an important muscle in the temporal region of dinosaurs. Despite these rather robust inferences, the anatomical and phylogenetic vagaries of the muscle preclude estimations of the location and potential size of the muscle in fossil archosaurs. Whereas lepidosaurs have prominent m. adductor mandibulae externus medialis bellies, extant archosaurs do not. This is problematic because, like all tetrapods, non-avian dinosaurs probably had at least rudimentary versions of this muscle and phylogenetic bracketing is supportive of the muscle's reconstruction. However, from an anatomical perspective, it is simply unclear where and to what extent the muscle attached on the dorsotemporal fossa or the coronoid region of the mandible.
The identification of evolutionary patterns of jaw muscles is clouded by the difficulties in diagnosing homologies of specific bellies among fossil taxa. Although all dinosaurs had temporal muscles, exactly which muscles occupied the dorsotemporal fossa are difficult hypotheses to test. The most consistent inferences are that m. adductor mandibulae externus superficialis attached to the upper temporal bar and m. pseudotemporalis superficialis attached to the caudal surface of the laterosphenoid. The region's organization in extant taxa brackets inferences to a point: lepidosaur temporal fossae have three muscles (mm. pseudotemporalis superficialis, adductor mandibulae externus profundus and medialis); crocodylians and neognath birds have one (m. adductor mandibulae externus profundus). However, it is clear that the extant conditions in crocodylians and birds are highly derived and numerous data suggest fossil crocodylomorphs and maniraptoran dinosaurs certainly had multiple muscles in the dorsotemporal fossa. The few osteological correlates of temporal muscles that do exist among non-avian dinosaurs indicate that the mm. adductor mandibulae externus profundus/medialis complex was the dominant muscle of the dorsotemporal fossa whereas m. pseudotemporalis superficialis was limited to the caudal surface of the laterosphenoid. However, if one considers that ratites (e.g., Struthio, Eudromia) may represent the primitive avian condition, rather than a secondarily derived clade, then one would infer that m. pseudotemporalis superficialis was the dominant temporal muscle among, at least, theropod dinosaurs. Clear data that indicate shifts in muscle attachments among different clades of non-avian dinosaurs, such as the envisioned muscle shifts between basal and derived ceratopsians and ornithopods, would certainly benefit systematic analyses.
Inferences of jaw muscles reciprocally illuminate inferences of other cranial soft tissues. As they exit the braincase, the mandibular, and more often, the maxillary nerves excavate portions of the laterosphenoid and prootic. It is assumed that non-avian dinosaur jaw muscles and nerves follow the same topological patterns as other sauropsids, thus if a groove for the maxillary nerve is identified on the laterosphenoid, marking its course rostrally, it can be inferred that m. pseudotemporalis superficialis, which lies medial to the nerve was small or attached rostrally on the braincase. In hadrosaurs, the maxillomandibular nerves exit directly laterally and a bit caudally, often leaving a shallow groove or lip on the prootic (Figs. 2 and 5). This suggests that m. pseudotemporalis superficialis did attach to the rostral portion of the dorsotemporal fossa even though there are no direct correlates that indicate the muscle's attachment. Therefore, given adequate knowledge of jaw muscle anatomy and using neurovascular topological rules, a portion of the other soft tissues of the adductor chamber can be reconstructed with relative confidence. Interestingly, in hadrosaurs, when the temporal muscles are reconstructed from the temporal fossa to the coronoid process, around the laterosphenoid buttress, the muscles pass through much of the orbit. Therefore, interpretations of muscle anatomy may directly impact inferences of eyeball size. Thus, as in other vertebrates, orbit size may not be a complete index of eyeball size (Ross and Kirk, 2007). These packing issues can be further explored not only in the orbit, but also the nasal cavity, pharynx, and middle ear cavities, which are bounded by large portions of the pterygoideus dorsalis and ventralis muscles, and protractor muscles, respectively.
Finally, feeding behavior and connective tissue adaptive plasticity are major factors involved in the structure and function of jaw muscles and the skull (Ravosa et al., 2007). These phenomena make interpreting jaw muscle functional anatomy extremely difficult over ontogenies of animals and may even manifest themselves among different populations of the same taxon (Erickson et al., 2004). These anatomical features must be well-understood before the testing of functional and systematic hypotheses. The best solution to investigating feeding behaviors, such as chewing or bite force, among non-avian dinosaurs may be to test hypotheses using gross as well as more exact muscle inferences (i.e., “temporal” muscles versus specific muscle bellies) while also testing a variety of virtual physiological cross-sectional areas, recruitment levels, and other behavioral or biomechanical parameters (e.g., Rayfield, 2005; Wroe et al., 2005; Rayfield and Milner, 2008). Anatomical inferences can only form a general framework for functional analysis when dealing with soft tissues such as jaw muscles.
In conclusion, jaw muscles offer a variety of bony structures that support inferences of their reconstruction in fossil taxa such as non-avian dinosaurs. Even individuals of the same taxon may have more or fewer osteological correlates that support one muscle's inference versus another. When anatomy does not offer necessary insight for soft-tissue reconstruction, phylogenetic bracketing may lend inferential support. Among the muscles described above, the orbitotemporal muscles have the weakest phylogenetic support, but generally have the strongest anatomical support: m. levator pterygoideus is a Level III inference, but has excellent osteological correlates that support its reconstruction. On the other hand, temporal muscles such as m. adductor mandibulae externus medialis or m. pseudotemporalis superficialis have strong phylogenetic support, and it is likely all fossil reptiles had the muscle bellies, but anatomical structures that explicitly support identifications of their attachments are rare. It is relatively easy and straightforward to make conservative interpretations of jaw muscles and caution is suggested when using muscle data in evolutionary and functional analyses. Regardless, jaw muscles are a critical component to understanding head anatomy, function, and evolution in non-avian dinosaurs.
Many thanks to Lawrence Witmer, Natalia Rybczynski, David Dufeau, Tobin Hieronymus, Ryan Ridgely, Emily Rayfield, Mark Young, and others for providing advice and assistance during the development of this project. Thanks to numerous staff and curators at museums for access to specimens. Comments from Peter Dodson and two anonymous reviewers greatly improved the manuscript.