palaeontologia electronica · multiple sets of elements. in hadrosaurids the pleu-rokinetic hinge...

Palaeontologia Electronica http://palaeo-electronica.org

PE Article Number: 11.2.9ACopyright: Society of Vertebrate Paleontology July 2008Submission: 24 September 2007. Acceptance: 4 February 2008

Rybczynski, Natalia, Tirabasso, Alex, Bloskie, Paul, Cuthbertson, Robin, and Holliday, Casey. 2008.A Three-Dimensional Animation Model of Edmontosaurus (Hadrosauridae) for Testing Chewing Hypotheses. Palaeontologia Electronica Vol. 11, Issue 2; 9A:14p;http://palaeo-electronica.org/2008_2/132/index.html

A THREE-DIMENSIONAL ANIMATION MODEL OF EDMONTOSAURUS (HADROSAURIDAE) FOR TESTING CHEWING HYPOTHESES

Natalia Rybczynski, Alex Tirabasso, Paul Bloskie,Robin Cuthbertson, and Casey Holliday

ABSTRACT

Here we describe a 3-D animated model of the craniodental system of a hadro-saur, developed for testing hypotheses of feeding kinematics. The model was createdfrom scanned cranial elements of an Edmontosaurus regalis paratype (CMN 2289).Movements within the model were created in animation software using inverse kine-matics and a wiring system composed of cranial elements. The model was used toreproduce the pleurokinetic hypothesis of hadrosaur chewing. The pleurokinetichypothesis, formally developed in the 1980s, proposed that hadrosaurs employedtransverse chewing movements via cranial kinesis. Specifically during the powerstrokethe maxillae were abducted. This is the first model to allow investigation into secondaryintracranial movements that must have occurred in order for the skull to accommodatethe primary, pleurokinetic movements. This study found secondary movements to beextensive among the joints of the palate and face. Further refinement and developmentof the model, including the integration of soft-tissue structures, will allow for a morein-depth examination of the pleurokinetic hypothesis and comparison with alternativefeeding hypotheses.

Natalia Rybczynski. Research Services (Palaeobiology), Canadian Museum of Nature, PO Box 3443 STN “D” Ottawa, ON, K1P 6P4, Canada [email protected] Tirabasso. Information Management and Technology Services, Canadian Museum of Nature, PO Box 3443 STN “D” Ottawa, ON, K1P 6P4, Canada [email protected] Paul Bloskie. Arius3D, Canadian Museum of Nature, PO Box 3443 STN “D” Ottawa, ON, K1P 6P4, Canada [email protected] Cuthbertson. Dept. of Biological Sciences/Faculty of Veterinary Medicine, University of Calgary, 3330 Hospital Dr. NW, Calgary, Alberta, Canada, T2N 4N1, Canada [email protected] Holliday. Dept. Anatomy & Pathology, Joan C. Edwards School of Medicine, 1542 Spring Valley Dr., Marshall University, Huntington, West Virginia 25704-9398, USA [email protected]

KEYWORDS: Animation, chewing, dinosaur, feeding, Inverse Kinematics, pleurokinesis wiring system

RYBCZYNSKI ET AL: Edmontosaurus MASTICATION

2

INTRODUCTION

Reconstructing feeding behavior in extinctanimals is fundamental to gaining insight into theirecology and evolution. Previous investigations intodinosaur feeding have utilized multiple approachesincluding, soft-tissue reconstruction (Ostrom 1961;Haas 1963; Galton 1984; Norman 1984; Molnar1991) and study of toothwear patterns (e.g., Nor-man and Weishampel 1985; Weishampel and Nor-man 1989; Fiorillo 1998; Barrett 2000).Additionally, consideration of joint surfaces haveserved to help reconstruct kinematics of the feed-ing system, and more recently finite-element mod-eling has been used to investigate how the headskeleton may resist strains associated with feeding(e.g., Ostrom 1964; Weishampel 1984; Upchurchand Barrett 2000; Rayfield et al. 2001; Rayfield2004; Therrien et al. 2006). A major challenge inreconstructing dinosaurian feeding behaviorrelates to our incomplete understanding of soft tis-sues (e.g., musculature, ligaments) (Witmer 1995).Even if the gross morphology of the soft tissues isproperly reconstructed, it is still difficult to infer howthe soft tissues would have interacted with thehead skeleton to produce feeding movements(Lauder 1995). The challenge appears most acutefor lineages where the head skeleton is highlyderived, as seen in hadrosaur dinosaurs. Nonethe-less, the architecture of the head skeleton doesoffer numerous data that can be used to identifyconstraints in mobility, an important first step in anykinematic reconstruction.

Within Dinosauria, feeding behavior of hadro-saurs has garnered particular attention. Hadro-saurs (Ornithopoda, Iguanodontia, Hadrosauridae)are a globally successful group of large-bodied,Late Cretaceous herbivores (Ostrom 1961; Nor-man and Weishampel 1985). A possible key totheir success is a specialized feeding system thatincludes a transverse chewing stroke (Norman andWeishampel 1985). Transverse chewing in combi-nation with other specializations such as a robustoccluding dentition, may have allowed hadrosaursto process tough foods more effectively than theirless specialized herbivorous counterparts. Amongtetrapods, transverse chewing appears to havearisen within only two lineages: mammals (Weijs1994) and hadrosaurs. In mammals, transversechewing is achieved through mandibular move-ments, whereas in hadrosaurs it has been hypoth-esized to involve a unique, complicated set ofintracranial movements (i.e., pleurokinesis)(Weishampel 1984; Norman and Weishampel1985). Despite many previous studies into the had-

rosaur feeding apparatus (Ostrom 1961; Heaton1972; Norman 1984; Weishampel 1984) elucidat-ing the mechanisms of cranial kinesis has been dif-ficult in part because of conflicting interpretations ofthe soft tissues (Ostrom 1961; Weishampel 1984),but also because of the complex geometry associ-ated with the proposed cranial kinesis. This studyis the first to investigate hadrosaur feeding by com-bining three-dimensionally scanned, cranial ele-ments with animation techniques. The advantageof this 3-D animation approach is that it allows usto take into account the geometry of each cranialelement and investigate how the architecture of theskull and shapes of the intracranial joints con-strains feeding movements.

Pleurokinesis and the evolution of the hadrosaur feeding apparatus

One of the most specialized components ofthe hadrosaur (cranial) feeding apparatus is thedental system. Primitively, the ornithopod dentitioncomprised separated, cusped teeth; however in theevolution of the lineage the teeth within the toothrows became progressively more tightly packedtogether. In the most highly derived ornithopods(e.g., hadrosaurs) the teeth are interlocking andeach tooth row forms a single occlusal surface.The resulting toothrow is referred to as a “dentalbattery.” The occlusal surfaces of the dental batter-ies tend be heavily worn, suggesting that the ani-mals employed a high degree of oral processing.Moreover, tooth wear evidence indicates that thepowerstroke phase of the chewing cycle, when theupper and lower teeth are mechanically processingfood, involved a large transverse component(Weishampel 1984). In early ornithopods, thetransverse powerstroke was probably achieved byrotating the mandibles about their long axis, so thatthe dentaries pivoted against the predentary(Weishampel 1984; Crompton and Attridge 1986).In contrast, more derived ornithopods (i.e., Iguan-odontia) were hypothesized to have abandonedmandibular rotation in favor of maxillary rotation(Weishampel 1984; Norman and Weishampel1985; You et al. 2003). Maxillary rotation wouldhave occurred about the pleurokinetic hinge, whichis a complex set of articulations formed betweenmultiple sets of elements. In hadrosaurids the pleu-rokinetic hinge is formed along the maxilla-premax-illa, lacrimal-jugal, postorbital-jugal, andquadrate-squamosal contacts (Weishampel 1984)(Figures 1, 2).

The pleurokinetic hypothesis of iguanodontidchewing assumes bilateral occlusion (i.e., isog-

PALAEO-ELECTRONICA.ORG

3

nathy) and a unique system of mobile intracranialjoints (Norman and Weishampel 1985). To investi-gate how this complex system may have func-tioned during chewing Weishampel (1984) used an“Integrated Mechanisms Program.” This computerprogram allowed him to model the skull as a rigid-linked, closed-loop system with multiple degrees offreedom. Using an iterative approach, he resolvedhow a bilaterally occluding dentition, coupled with apleurokinetic hinge system could produce a trans-verse powerstroke. The model captures the pri-mary movements of the feeding mechanism. Themovements involved in the powerstroke are: (1)mandibular adduction (2) abduction of the maxillae,and (3) abduction and retraction of the quadrates(Figure 3). Note that although abduction of themaxillae, and abduction and retraction of the quad-rates are referred to as primary movements, theyarise as a consequence of mandibular adduction,due to the geometry of the cranial linkage system.

Though sophisticated for its time, the originalcomputer model could not easily account for the 3-D morphology of the joint surfaces (Figure 3) andso did not allow researchers to investigate second-ary movements. Secondary movements are intrac-ranial movements that occur beyond thepleurokinetic hinge (see above). They can also bethought of as movements that would be necessaryin order for the primary movements (e.g., maxillaryabduction) to occur. Regions of potential second-ary movement include the pterygoid-palatine,pterygoid-ectopterygoid-maxilla, pterygoid-quad-rate, pterygoid-basisphenoid, and jugal-quadrato-jugal contacts (see Weishampel 1984). Secondarymovements are important as they may representadditional constraints within the linkage system.Additionally, some of the elements involved in thesecondary movements (e.g., pterygoid) wouldhave served as muscle attachment sites (e.g., M.protractor pterygoideus, M. pterygoideus dorsalis),and so are implicated in aspects of the feedingmechanism associated with force production.

This study represents the first step towarddeveloping a new animation model that will beused to investigate and test chewing hypotheses inEdmontosaurus. Here we describe how the anima-tion model was created from scanned fossil ele-ments recovered from a single individual, and thenmobilized using inverse kinematics and a wiringsystem. We finish with a discussion of the prelimi-nary results and an outline of future directions.Notably, an additional objective of this study was toproduce an animation, which could illustrate thepleurokinetic hypothesis of hadrosaur chewing for

Figure 1. Edmontosaurus cranial anatomy. 1, Left, lat-eral view; 2, Left, lateral view with quadratojugal andjugal removed from relevant skeletal elements; 3,Medial view of left half of skull with relevant skeletal ele-ments labeled; 4, Caudal view of skull with relevantskeletal elements. Abbreviations: bj, basal joint; bpt,basipterygoid process; cp, coronoid process; dn, den-tary; ect, ectopterygoid; fr, frontal; jj, jaw joint; ju, jugal;la, lacrimal; mx, maxilla; na, nasal; oj, otic joint; pal,palatine; po, postorbital; prdn, predentary; prmx, pre-maxilla; prp, preotic pendant; pt, pterygoid; pt, ptery-goid; qj, quadratojugal; qu, quadrate; sa, surangular;sp, splenial; sq, squamosal.


4

museum visitors. The model described herein iscurrently on exhibit at the Canadian Museum ofNature, Talisman Fossil Gallery in Ottawa.

METHODS

Edmontosaurus material

The cranial material used for scanning is fromCanadian Museum of Nature (CMN) specimennumber 2289, a paratype of Edmontosaurus rega-lis. The specimen was recovered by C.H. Stern-berg in 1916 from the Edmonton Formation alongthe Red Deer River, “seven miles west and north ofMorrin” (Lambe 1920, p. 2). The head skeleton ismostly complete and exceptionally well-preserved.The skull is relatively undistorted and is only miss-ing the predentary and premaxillae. The suspenso-rium (e.g., quadrates, palatal elements) were founddisarticulated from the braincase. For this study,the braincase and available left-side elementswere selected for scanning. The left-side elementswere favored because they were more completethan those of the right side.

Arius 3-D laser scanner

The elements were scanned using an Arius 3-D laser scanner. The laser system characterizeseach scanned point from the surfaces of the objectaccording to its color and location in three-dimen-sional space. It does this by scanning the surface

of an object using one focused laser beam com-prising three wavelengths (red, green, and blue),and recording the reflected light using a chargecouple device. Each point on the object isdescribed by six numeric values; three positionalvalues X, Y, and Z, and three surface color valuesR, G, and B. The X coordinate of each point on theobject is calculated from an accurate measurementof the position of the scanning mirror in the cam-era. The Y coordinate is calculated from an accu-rate measurement of the camera motion system.

Figure 2. Location of the pleurokinetic hinge in Edmontosaurus.

Figure 3. Animation showing pleurokinetic hypothesis,illustrated using Corythosaurus casuarius (CMN 8676).Animation is based on Weishampel (1984, figure 20). Alimitation of this model is that it allows for investigationof the primary movements only (see text). Reproducedwith permission of the Canadian Museum of Nature,Ottawa, Canada.


5

The Z coordinate is calculated through laser trian-gulation within the camera. At the same time colorinformation at each point is gathered by measuringthe intensity of the reflected laser beams. Colorintensity measurements are accurate, being com-pletely independent of ambient light. The total lightexposure is about 3.5 milliwatts, roughly equivalentto shining a flashlight on the object. The laser lightmoves continuously at about 300 mm/sec acrossthe surface of the object therefore the dosage oflight on the specimen’s surface is extremely small.

For each scan, the laser beam passes overthe surface, one scan line at a time. The laserscans at a resolution as fine as 100 microns,recording 3-D shape and color simultaneously withhigh-resolution and perfect registration. In thisstudy, each cranial component was scanned with

sequential overlapping scans until the entire sur-face was covered. The scanning resolution foreach component ranged from 300 µm to 600 µm,depending on the size of the component beingscanned. Larger components such as the brain-case were scanned at 600 µm. The dentary wasscanned at 400 µm (see Figure 4), whereassmaller elements, such as the palatine, werescanned at 300 µm. Theer’s effective scanningarea (i.e., field of view) is approximately 60 mmwide. Consequently larger objects require multiplescans (see Figure 5). An interactive 3-D model isavailable at: http://www.arius3d.com/Edmontosau-rus/.

Each scan was processed using Pointstream3DImageSuite (2007) to create a point-cloud form.Overlapping scans of the object were aligned inPointstream. This process requires manually pick-ing three to five points that are common betweentwo scans. The alignment algorithm then usesthese points to search for geometric commonalitybetween scans and then shifts the selected scansinto an aligned position.

Once the complete point-cloud model wascreated, it was converted into a triangulated poly-mesh comprising thousands of individual facesusing surfacing software (Paraform v3.1, 2001).Compared to the number of points in the originalpoint-cloud, the number of triangles in the originalconverted polymesh is roughly double. For exam-ple the dentary (scanned at 400µm resolution) con-tained 2.2 million points and when converted topolymesh it yielded 4.4 million triangles. Modelswith such heavy computations require too muchprocessing power and memory to be practically

Figure 4. The Arius 3D laser scanner scanning dentaryof Edmontosaurus, CMN 2289.

Figure 5. Random color display illustrating many scans aligned in Pointstream to make 3-D model. The dentary dataincludes a total of 1,246,121 xyzrgb points.


6

used in animation software. Therefore the facecount in the polymesh models for all the scannedelements was reduced. For example the braincasewas scanned at 500µm and yielded 1.1 millionpoints and 2.2 million polymesh triangles that werethen reduced to 100,000 triangles. The resultingelement was much lighter to work with but stillmaintained its geometry. Other elements that werereduced substantially include the splenial, nasal,ecopterygoid, and palatine. Elements, or parts ofelements, whose surfaces could be involved withcranial kinesis (e.g., upper and lower toothrows,pterygoid, jugal, quadratojugal) were not reducedas much. For example, in its reduced state, themaxillary tooth row comprised 100,000 polymeshtriangles.

3-D animation

Assembling the model. The reduced-sizepolymesh skull-components were imported into theanimation software 3-D Studio Max v. 8 (3ds Max;Autodesk 2006) as object files and digitally assem-bled to form the skull. Right-side elements thatwere not scanned (see above) were constructed inthe animation software by creating a mirroredclone of the left-side elements. Most of the CMN2289 skull components were fit together with littledifficulty. One exception was the pterygoid, whichinitially did not fit well lengthwise within the skull.Given that the pterygoid is a very thin element, itwas likely deformed post-mortem. For this model,the pterygoid was fit to the skull by slightly shorten-ing the quadrate ramus of the pterygoid. For futureversions of the model the question of how best tocorrect for deformation of the pterygoid will beexplored more fully. There is also evidence ofsome deformation within the braincase. Examina-tion of the braincase in anterior view (e.g., seeLambe 1920, figure 7) shows the bottom of thebraincase is sheared slightly to the right. For thisstudy the basipterygoid processes were shiftedslightly back toward the left so that they were moresymmetrical.

Also uncertain in the model is the mediolateralposition of the lower tooth rows relative to theuppers. For the current model the dentaries werepositioned by aligning the upper and lower toothrows so that the Edmontosaurus model is isogna-thous, as per the assumptions of the original pleu-rokinetic-hinge hypothesis (see above). Theresulting configuration shows a small spacebetween the anterior tips of the dentaries. Unfortu-nately, with the CMN 2289 predentary missing, it isdifficult to ascertain whether this configuration of

the dentaries is reasonable. Future studies willexamine the morphology of the predentary regionin hadrosaurs in more detail in order to better con-strain the positioning of the dentaries.

The predentary, premaxilla, and quadratojugalprocesses of the jugal were missing and werereconstructed either with computer modelling, or byscanning and modifying other specimens (Figures6, 7). These reconstructions were for aesthetic pur-poses only. Importantly, these reconstructions donot influence the functioning of the model. The pre-dentary model was created from a scan of an iso-lated CMN uncatalogued specimen that wasmodified to fit the dentaries of the model. The pre-maxilla was reconstructed in two parts. The ante-rior part was reconstructed based on a partial leftpremaxilla from the subadult specimen of Edmon-tosaurus annectens, BHI-6217. The posterior partof the premaxilla was reconstructed using 3-Dsurfacing tools in 3ds Max with reference to otherEdmontosaurus (e.g., type specimen, CMN 2288).Despite the missing premaxilla, the maxilla couldstill be correctly positioned within the skull by con-sidering its articulation with other elements (e.g.,jugal). Also, although the back of the jugal is par-tially missing, part of its articulating surface for thequadratojugal is preserved, so the positional rela-tionship between the quadratojugal and jugal iswell understood. Animation methods. The polymesh maxilla, man-dible, and quadrate were attached to a 3-D frame-work. In animation terminology this framework isreferred to as a “rig.” Figure 8 shows a rig for theCorythosaurus and Edmontosaurus models. TheCorythosaurus rig is derived from the model out-lined in Weishampel (1984). The Edmontosaurusrig differs from the Corythosaurus rig in proportionsof the elements, therefore the location of the articu-lations also differ. Both rigs include six enclosedshapes (i.e., “bones” in animation terminology),representing the paired mandibles, quadrates, andmaxillae. Both rigs include a joint located at themandibular symphysis, as well as at the mandible-quadrate, and the cranium-quadrate joints. In therig the joints are represented by circles. There isalso a hinge-line along the dorsal margin of themaxilla, which represents the maxillary portion ofthe pleurokinetic hinge.Inverse kinematics. This is a tool often used inanimation to recreate accurate limb movements. Itis a method of determining how the linked ele-ments of a system must move in order to produce adesired outcome movement. It is based on a bot-tom-up hierarchal approach where the child ele-


7

ment controls the movement of the parent in thelinked system. For example, the human arm is ahierarchal linked system that can be animated(without IK), by positioning the linked elementsindividually, starting from the parent of the system(e.g., upper arm) down to the last child of the hier-archal linked system (hand). In IK, the human armis animated/moved by using the last child of thehierarchal linked system (hand) to determine theposition of the upper elements of the arm(Autodesk Canada 2006). Figure 8 shows how theIK approach effectively recreates the adduction ofthe mandible and retraction of the quadrateobserved in the pleurokinetic model presented inWeishampel (1984). In the Edmontosaurus 3-Dmodel presented here, the mandible and quadratewere treated as a linked system because theyshared a common pivot point (i.e., at the jaw joint).The child of the system is the mandible, and theparent is the quadrate. Given that the mandibleand quadrate are linked as a hierarchal system, thequadrate was moved by manipulating the positionof the mandible. The 3-D rig with the attached poly-

mesh skull components was mobilized using twotypes of animation methods: Inverse kinematics(IK) and a wiring system. These two approacheswere used to ensure that the primary movements,including their rotational magnitudes presented inthe original hypothesis (see Corythosaurus anima-tion in Figure 3) were replicated in the Edmonto-saurus model. The wiring system was used to linkthe maxilla and quadrate, whereas the mandibleand quadrate were linked by inverse kinematics.

In the Corythosaurus model, the maxilla andquadrate are shown abducting simultaneously(Weishampel 1984, figure 20a). However, thedegree of rotation of the quadrate and maxilla dif-fer: In anterior view the rotation of the maxilla isgreater than that of the quadrate (see Table 1).Weishampel (1984, p. 79) suggested that differ-ences in rotation should be possible because the“palatine pterygoid articulation in hadrosauridsmodifies the degree to which the quadrates areforced laterally and caudally by lateral rotation ofthe maxillae.” Yet, the maxilla and quadrate arealso attached via the jugal and quadratojugal.

Figure 6. Lateral view of model showing premaxilla and predentary reconstruction. Pink regions indicate reconstruc-tion using 3-D surfacing tools in 3dstudio max v. 8. Green portions are scanning from actual elements. Note that pre-dentary scanning was deformed to fit model. Reconstructions are for aesthetic purposes (i.e., for museum exhibit).


8

Thus, the movements of the maxilla and quadrateare linked medially by intervening palatal elements(i.e., palatine and pterygoid) and laterally by lateralfacial elements (i.e., jugal, quadratojugal). For the

maxillary abduction to be greater than the quadratethere has to be movement through both the medialand lateral linkage systems. Lateral facial elementswould have to disarticulate. For example the jugal

Figure 7. Lateral view of model showing jugal reconstruction. Reconstruction was created using 3-D surfacing toolsin 3dstudio max v. 8 and was for aesthetic purposes (i.e., museum exhibit).

Figure 8. Animation of Corythosaurus and Edmontosaurus “rigs.” The animation starts by showing the rig, which cor-responds to the Corythosaurus animation shown in Figure 3. The animation then transform from the Corythosaurusrig to the Edmontosaurus rig. Both rig animations use the same inverse kinematics and wiring system (see text). TheEdmontosaurus rig forms the basis of the animation seen in Figure 8 and 9. Reproduced with permission of theCanadian Museum of Nature, Ottawa, Canada.


9

would be pulled rostrolaterally from the quadratoju-gal. Such a pattern of lateral disarticulation was notdescribed as part of the original model(Weishampel 1984). We, therefore, incorporatedthe simplifying assumption that the lateral facialskeleton should function to ensure that the rotationof the maxilla and the quadrate would be matchedin magnitude.Wiring system. In the Edmontosaurus model, themaxilla and quadrate were linked using an anima-tion technique referred to as a wiring system. A wir-ing system is an animation technique that creates arelationship between two or more elements so thatany change in position or orientation made in theparent element (e.g., the maxilla) is translated tothe child element(s), (e.g., the quadrate) (AutodeskCanada 2006). The wiring system linked a pointnear the posteroventral margin of the maxilla to aposteroventral point on the quadrate, so that whenthe maxilla was moved manually the quadratemoved in response. It is possible to move both themaxilla and quadrate manually; however the wiringsystem was used here to automate the model forthe testing phase.

As mentioned above, only the mandible, max-illa, and quadrate polymesh components wereattached to the rig. The braincase, nasals, and pre-maxilla were not attached to the rig, but were sim-ply held stationary within the animation worldspace. The jugal and palatine were attached rigidlyto the maxilla, so these three elements functionedas a single, solid object. The assumption that theseelements would be a single functional unit is rea-sonable considering their shared attachment sur-face is large and complex. The various elementsforming the mandible were also treated as a singlerigid object.

In order to accommodate the quadrate retrac-tion described in the original hypothesis (Figure 3)it was necessary to allow some mobility in thefacial skeleton of the model. In particular, as thequadrate retracts, it should move away from themaxilla so that one or both of the intervening joints(i.e., the jugal-quadratojugal and/or quadratoju-gal-quadrate contacts) disarticulate. To simplify themodel, we arbitrarily chose to fix the quadratojugalto the quadrate. The jugal and quadratojugal wereleft unconnected so that as the quadrate retractedseparation could occur only between the quadrato-jugal and jugal. It is important to keep in mind thatany observed separation between the quadratoju-gal and jugal should be taken to indicate that thereis separation between elements somewhere alongthe chain of lateral facial elements.

The pterygoid contacts the basipterygoid pro-cess, palatine and quadrate. In this model theseattachments were not rigid. Rather, as the palatineand quadrate moved through the powerstroke, theposition and orientation of the pterygoid wasadjusted “manually” so that, first, connections weremaintained between the pterygoid and its contact-ing elements, and second, there was no interfer-ence between the pterygoid and the surroundingelements.

Assumptions of the model. As in any model,animation models are necessarily simpler thantheir real-world counterparts. For research pur-poses, the simplifying assumptions used in themodel should be matched to the research ques-tion. In this study, the model was developed toinvestigate the intracranial movements that mayhave occurred during chewing. Thus the focus ofthis model is on bony interactions among the cra-nial elements and occlusal surfaces of tooth rows.

Table 1. Comparison of quadrate, maxillary, and mandibular rotational movements in Corythosaurus model(Weishampel 1984, figure 20) and Edmontosaurus model (this study). Rotational movements (in degrees) aredescribed for lateral and anterior views of the skull and occurred at the powerstroke phase of the chewing cycle only.Quadrate movements occur at the squamosal articulation. Maxillary movements occur about the lacrimal joint. Rota-tional movements in anterior view are abduction.

ViewLateral Anterior

ElementCorythosaurus

modelEdmontosaurus

modelCorythosaurus

modelEdmontosaurus

model

Quadrate 5 6 7 3

Maxilla NA NA 10 3

Mandible * 5 5 NA NA

* Value presented in this table represents the change in angle between the mandible and quadrate from thebeginning and to the end of the powerstroke.


10

To investigate intracranial movements, weassumed that the braincase and parts of the facialskeleton were immobile, whereas particular jointsin the palate and face were not. Mobile connec-tions along the face were prevented from being dis-placed mediolaterally relative to one another, whilebeing allowed some rostrocaudal separation. Thisconstraint seems acceptable because most ofthese joints overlap mediolaterally. The remainingcranial elements were allowed to be mobile, allow-ing those movements resulting from mandibularadduction and maxillary abduction to occur.

Some of the assumed mobility, such as thatseen at the otic and basal joints, is similar to thatfound in many birds and squamates (Zusi 1993;Herrel et al. 1999; Metzger 2002). The synovialmorphology of these joints in hadrosaurs suggeststhat these may also have been mobile(Weishampel 1984). On the other hand, the modelalso assumes mobility within the facial and palatalskeleton (e.g., quadrate-quadratojugal, pterygoid-palatine). The occurrence of smooth joint-surfacesin the face of dinosaurs has led many researchersto infer sliding motion between them (e.g.,Weishampel 1984; Bakker 1986; Rayfield 2004).Although facial mobility is present in some birds(e.g., some parrots developed synovial jugal-maxil-lary articulations) and snakes, the presence ofmobility in the facial sutures of non-avian dinosaursand most other diapsids remains uncertain (Holli-day 2006), and the functional link between suturemorphology and intracranial mobility is largelyunexplored. It is beyond the goals of this study toanalyze the morphological basis for cranial kinesis,and we allowed the potential for movement tooccur at all of these joints during the experiment.

RESULTS

The preliminary model presented hereattempted to replicate the primary movements (Fig-ure 9) described in the original pleurokinetic modelof chewing in hadrosaurid dinosaurs (Weishampel1984) (see Figure 3). The original model, based onCorythosaurus, described intracranial movementsduring the powerstroke phase of the chewingcycle. A comparison of the primary movements (i.e.maxillary, quadrate, and mandibular movements) inthe two models is shown in Table 1. In the pre-sented animation model (Figure 8), the mandibleprotracted approximately 4.1 cm. The maxillaabducted 2.1 cm. The jaw joint retracted andabducted approximately 2.1 cm. In lateral view, thedegree of rotation of the mandible (with respect tothe quadrate) in both models is the same, and therotation of the quadrate, leading to mandibularretraction, is also similar. The abduction of thequadrate and maxilla are the same in Edmontosau-rus (3°) because they were linked (see Methods:Animation). The Edmontosaurus and Corythosau-rus models differ dramatically in the degree towhich the maxillae abduct. In Corythosaurus themaxillary rotation (abduction) is 10° and the quad-rate is only 7°. In the Edmontosaurus model themaxillary and quadrate movements were linked,and 3° of movement sufficed to allow the upperand lower teeth to complete the powerstroke.

To accommodate the primary movements,secondary movements were required (Figure 10)at the pterygoid-palatine, pterygoid-ectoptery-goid-maxilla, pterygoid-quadrate, ptery-goid-basisphenoid, and jugal-quadratojugalcontacts. In some cases these movementsinvolved marked separation of the elements. Forexample, during the chewing cycle, the jugal andquadratojugal were separated 1.3 cm, and the sep-aration between the ventral portion of the pterygoidand the quadrate was approximately 1.4 cm.

DISCUSSION AND FUTURE WORK

This model revealed that in order for the pri-mary movements to occur, as described in the orig-inal pleurokinetic hypothesis (Weishampel 1984)(Figure 3), extensive secondary movements mustalso occur. Moreover, some of the secondarymovements observed involved large (> 1 cm) sep-aration of the elements, including movementsbetween the jugal and quadratojugal, pterygoidand palatine, and pterygoid and quadrate (Figure10). Importantly, these movements arose with onlysmall amount of abduction of the maxillae and

Figure 9. Animation showing Edmontosaurus-model"chewing" according to Weishampel's (1984) pleuroki-netic hypothesis, highlighting primary movements.Reproduced with permission of the Canadian Museumof Nature, Ottawa, Canada.


11

quadrate (e.g., 3º, Table 1), which were much lessthan those in the original Corythosaurus model.Presumably, the pleurokinetic model, as originallyproposed by Weishampel (1984) would haveyielded secondary movements much greater thanthose observed in the Edmontosaurus model pre-sented here.

Both the original model and this one did notincorporate soft tissues such as ligaments, syn-ovial capsules, and musculature though their inclu-sion would greatly enhance functional resolutionand allow for more rigorous hypothesis testing. Lig-aments, including those within sutures and largerones, such as the quadrate-articular ligament ifpresent, may have constrained some, if not muchof the movement in the facial skeleton this modelillustrated. In particular, it seems initially doubtfulthat such separation of cranial elements is likely tohave taken place in the living animal consideringthat sutural ligaments are comprised of collagen,which yields at only 4 to 5% strains (measured intendon, Wainwright et al. 1976). Additionally thereis a significant amount of bony overlap betweenmany of these articulations, suggesting that thesecontacts promoted stability rather than mobility.Similarly, synovial capsules were likely present,minimally at the otic (quadrate-squamosal), basal(pterygoid-basisphenoid), and jaw joints. Thesecapsules would have potentially constrained move-ment expressed in the model. The significance ofthese factors would depend on the functional prop-erties assigned to them at different time stages ofthe model, requiring multiple assumptions.

Consideration of jaw musculature may offerfurther insights into mobility and function of cranialelements during feeding. If maxillary and quadrateabduction existed, it would supposedly have beena passive movement because there are no mus-cles in a position that could implement it. Rather,abduction of the tooth rows would be driven by theadduction of the lower teeth against the upper.Resistance to maxillary abduction would havebeen provided by pterygoideus musculature. Theattachments of Mm. protractor pterygoideus, leva-tor pterygoideus, pterygoideus ventralis, and ptery-goideus dorsalis suggest that this powerfulcomplex of muscles could have restrained maxil-lary abduction (Figure 11). If the pleurokinetichypothesis is viable, these antagonistic muscleswould have played an important role modulatingthe cranial stiffness associated with maxillaryabduction. The capacity to modulate stiffness, andtherefore, bite force, would be critical for the sys-tem to be able to mechanically break down foodsof different size and material properties (e.g., hard-ness, toughness).

In addition, many of the aforementioned jawmuscles attach across a number of the elementsand joints that were implicated in secondary move-ments. The lateral surfaces of the palatine, ptery-goid, and quadrate are a likely attachments for M.pterygoideus dorsalis and M. adductor mandibulaeposterior, whereas the medial surfaces of theseelements were probably attachment regions for M.pterygoideus ventralis and M. protractor pterygoi-deus (Figure 11). This pattern of muscle attach-ment suggests that the palatine, pterygoid, andquadrate formed a functional unit, providing a sup-posedly stable, immobile attachment for jaw mus-culature. Yet the model exhibited extensive slidingbetween the quadrate and pterygoid during chew-ing challenging functional hypotheses regardinghow this part of the skull may have simultaneouslyserved to produce and resist chewing forces.Therefore, jaw musculature, if modeled, would notonly provide the power to the modeled feedingapparatus, enabling analyses of acceleration andbite force moments, but would also induce stressesacross sutures and on the bones of attachmentthemselves, enabling analysis of deformation (e.g.,via finite element modeling), and hypotheses ofmobility (versus stability) to be tested.

Full consideration of the soft tissue systemsmay in the end reveal that maxillary abduction isnot a viable mechanism for producing transversechewing movements. With this in mind, additionalinsight into the potential mobility of the hadrosaurid

Figure 10. Animation showing Edmontosaurus-model"chewing" according to Weishampel's (1984) pleuroki-netic hypothesis, highlighting secondary movements.Reproduced with permission of the Canadian Museumof Nature, Ottawa, Canada.


12

Figure 11. Reconstructed jaw muscle anatomy of Edmontosaurus. 1, Left, lateral view of skull with muscle attach-ment surfaces; 2, Medial view of left half of skull with muscle attachment surfaces; 3, Caudal view of skull with mus-cle attachment surfaces; 4, Left, lateral view of skull with reconstructed superficial jaw muscles; 5, Left, lateral viewof skull with reconstructed deep jaw muscles; 6, Medial view of left half of skull with reconstructed jaw muscles; 7,Caudal view of skull with reconstructed jaw muscles. Abbreviations: mAMEM, Adductor mandibulae externus media-lis; mAMEP, Adductor mandibulae externus profundus; mAMES, Adductor mandibulae externus superficialis; mAMP,Adductor mandibulae posterior; mDM, Depressor mandibulae; mLPt, Levator pterygoideus; mPPt, Protractor ptery-goideus; mPSTp, Pseudotemporalis profundus; mPSTs, Pseudotemporalis superficialis; mPTd, Pterygoideus dorsa-lis; mPTv, Pterygoideus ventralis.


13

quadrate may be derived from study of other had-rosaurid taxa. Indeed, a review of the cranial jointsof Brachylophosaurus, another hadrosaurine dino-saur, found that the morphology of the pleuroki-netic hinge regions (e.g., quadrate-squamosaljoint) would have prohibited maxillary abduction(Cuthbertson 2006). Therefore cranial kinesis maybe variable among hadrosaurs, or perhaps thehadrosaurid skull was less mobile than originallyassumed. However, part of the difficulty in analyz-ing ornithopod chewing is that the proposed trans-verse abduction of the maxillae is rare, if notunique among amniotes, so identifying appropriateanalogs is difficult.

In ongoing work we plan to further correctparts of the skull model, which may be stilldeformed, as suggested by the fact that during thepowerstroke the teeth are not in precise occlusion(Figure 9). We will also investigate the role of softtissues in constraining the model (Ostrom 1961,Holliday 2006). The system will be further con-strained by incorporating dental wear data includ-ing the pattern of scratches (i.e., microwearstriations) and orientation of the occlusal surface. Ifthese wear features prove to be well-preservedthey will provide an accurate map of the relativemovements of the upper and lower tooth rows(e.g., Rybczynski and Reisz 2001). The inclusionof the additional functional parameters will allowfuture analyses to constrain the range of possiblechewing movements (e.g., see Hutchinson andGatesy 2006) to gauge how sensitive the model isto different variables. With these additional con-straints in place, alternative hypotheses of feedingbehavior can be more fully explored including sim-ple quadrate streptostyly, akinesis, and mandibularpropaliny or rotation.

In conclusion, preliminary results suggest thatthe integration of three-dimensional scanning tech-nology with animation offers enormous promise forinvestigating the functional morphology of feedingbehavior in Edmontosaurus. Although the goal ofthe methods described here is to resolve the feed-ing kinematics, the approach lays the groundworkfor investigating questions related to biomaterialsand chewing forces across all vertebrates. Ulti-mately, animation and biomechanical methodscould be applied comparatively to yield insight intothe evolution of feeding systems within lineagesand ecosystems.

ACKNOWLEDGEMENTS

We thank M.R. Saccu (Talisman fossil Galleryproject), who provided financial and moral support

for this project. M. Feuerstack, A. MacDonald, andK. Shepherd (Canadian Museum of Nature) pro-vided access to the fossils. We are grateful to R.Holmes (CMN) for helpful discussions, as well asM. Graham (CMN), and two anonymous reviewersof this paper whose comments greatly improvedthe manuscript. We also thank the Black Hills Insti-tute for the loan of the premaxilla, as well as D.Weishampel (Johns Hopkins University) and L.Witmer (Ohio University) for encouragement anddiscussion.

REFERENCES

Autodesk Canada. 2006. 3ds max V.8 User Reference.Autodesk, Inc., Montreal.

Bakker, R. 1986. The Dinosaur Heresies: New TheoriesUnlocking the Mystery of the Dinosaurs and TheirExtinction. William Morrow and Company, New York.

Barrett, P.M. 2000. Prosauropod dinosaurs and iguanas:speculations on the diets of extinct reptiles, p. 42-78.In Sues, H.D. (ed.), Evolution of Herbivory in Terres-trial Vertebrates: Perspectives from the FossilRecord. Cambridge University Press, Massachu-setts.

Crompton, A.W., and Attridge, J. 1986. Masticatoryapparatus of the larger herbivores during Late Trias-sic and Early Jurassic times, p. 223-236. In Padian,K. (ed.), The Beginning of the Age of Dinosaurs. Fau-nal Change across the Triassic-Jurassic Boundary.Cambridge University Press, Cambridge.

Cuthbertson, R. 2006. A redescription of the holotype ofBrachylophosaurus canadensis (Dinosauria: Hadro-sauridae), with a discussion of chewing in hadro-saurs. M.Sc. Thesis, Carleton University, Ottawa,Canada.

Fiorillo, A.R. 1998. Dental microwear patterns of the sau-ropod dinosaurs Camarasaurus and Diplodocus: Evi-dence for resource partitioning in the Late Jurassic ofNorth America. Historical Biology, 13:1-16.

Galton, P.M. 1984. Cranial anatomy of the prosauropoddinosaur Plateosaurus from the Knollenmergel (Mid-dle keuper, Upper Triassic) of Germany, I. Two com-plete skulls from Trossingen/Württemberg withcomments on the diet. Geologica et Palaeontologica,18:139-171.

Haas, G. 1963. A proposed reconstruction of the jawmusculature of Diplodocus. Annals of the CarnegieMuseum, 36:139-157.

Heaton, M.J. 1972. The palatal structure of some Cana-dian Hadrosauridae (Reptilia: Ornithischia). Cana-dian Journal of Earth Sciences, 9(185):185-205.

Herrel, A., De Vree, F., Delheusy, V., and Gans, C. 1999.Cranial kinesis in gekkonid lizards. Journal of Experi-mental Biology, 202(24):3687-3698.

Holliday, C. 2006. Evolution and function of the jaw mus-culature and adductor chamber of archosaurs (croco-dilians, dinosaurs, and birds). Ph.D. Thesis, OhioUniversity, Athens, Ohio, USA.


14

Hutchinson, J.R., and Gatesy, S.M. 2006. Dinosaur loco-motion - Beyond the bones. Nature, 440(7082):292-294.

Lambe, L.M. 1920. The hadrosaur Edmontosaurus fromthe Upper Cretaceous of Alberta. Memoir. GeologicalSurvey of Canada, 120:1-79.

Lauder, G.V. 1995. On the inference of function fromstructure, p. 1-18. In Thomason, J.J. (ed.), FunctionalMorphology in Vertebrate Paleontology. CambridgeUniversity Press, Cambridge, UK.

Metzger, K. 2002. Cranial kinesis in Lepidosaurs: Skullsin motion, p. 15-46. In Aerts, P., D'Août, K., Herrel,A., and Van Damme, R. (eds.), Topics in FunctionalEcological Vertebrate Morphology. Shaker Publish-ing, Maastricht.

Molnar, R.E. 1991. The cranial morphology of Tyranno-saurus rex. Palaeontographica Abteilung, A 217:137-176.

Norman, D.B. 1984. On the cranial morphology and evo-lution of ornithopod dinosaurs. Symposium of theZoological Society of London, 52:521-547.

Norman, D.B., and Weishampel, D.B. 1985. Ornithopodfeeding mechanisms - Their bearing on the evolutionof herbivory. American Naturalist, 126(2):151-164.

Ostrom, J.H. 1961. Cranial morphology of the hadrosau-rian dinosaurs of North America. Bulletin of theAmerican Museum of Natural History, 122:33-186.

Ostrom, J.H. 1964. A functional analysis of jaw mechan-ics in the dinosaur Triceratops. Postilla, PeabodyMuseum of Natural History, Yale University, 88:1-35.

Paraform, Version 3.1, 2001, User Reference. MetrisInternational Holding n.v. (Metris) Leuven, Bel-gium.(www.metris.com)

Pointstream 3DImageSuite for Windows®, Version 2.7,January 2007, User Reference. Arius3D Inc, Missis-sauga, Ontario Canada. (www.pointstream.net)

Rayfield, E.J. 2004. Cranial mechanics and feeding inTyrannosaurus rex. Proceedings of the Royal Societyof London Series B-Biological Sciences,271(1547):1451-1459.

Rayfield, E.J., Norman, D.B., Horner, C.C., Horner, J.R.,Smith, P.M., Thomason, J.J., and Upchurch, P. 2001.Cranial design and function in a large theropod dino-saur. Nature, 409(6823):1033-1037.

Rybczynski, N., and Reisz, R.R. 2001. Earliest evidencefor efficient oral processing in a terrestrial herbivore.Nature, 411(6838):684-687.

Therrien, F., Henderson, D.M., and Ruff, C.B. 2006. Biteme: Biomechanical models of theropod mandiblesand implications for feeding behavior, p. 179-237,The Carnivorous Dinosaurs. Indiana UniversityPress, Bloomington, Indiana.

Upchurch, P., and Barrett, P.M. 2000. The evolution ofsauropod feeding mechanisms, p. 79-122. In Sues,H.D. (ed.), Evolution of Herbivory in Terrestrial Verte-brates: Perspectives from the Fossil Record. Cam-bridge University Press, Massachusetts.

Wainwright, S.A., Biggs, W.D., Currey, J.D., and Gosline,J.M. 1976. Mechanical Design in Organisms. Prince-ton Antiques Bookshop, Atlantic City.

Weijs, W.A. 1994. Evolutionary approach of masticatorymotor patterns in mammals, p. 281-320, Advances inComparative and Environmental Physiology.Springer-Verlag, Berlin.

Weishampel, D.B. 1984. Evolution of Jaw Mechanismsin Ornithopod Dinosaurs. Advances in AnatomyEmbryology and Cell Biology, 87:1-109.

Weishampel, D.B., and Norman, D.B. 1989. Vertebrateherbivory in the Mesozoic: jaws, plants, and evolu-tionary metrics. Geological Society of America Spe-cial Paper, 238:87-100.

Witmer, L.M. 1995. The Extant Phylogenetic Bracket andthe importance of reconstructing soft tissues in fos-sils, p. 19-33. In Thomason, J.J. (ed.), FunctionalMorphology in Vertebrate Paleontology, CambridgeUniversity Press, New York.

You, H.L., Luo, Z.X., Shubin, N.H., Witmer, L.M., Tang,Z.L., and Tang, F. 2003. The earliest-known duck-billed dinosaur from deposits of late Early Creta-ceous age in northwest China and hadrosaur evolu-tion. Cretaceous Research, 24(3):347-355.

Zusi, R.L. 1993. Patterns of diversity in the avian skull, p.391-437. In Hanken, J., and Hall, B.K. (eds.), TheSkull. University of Chicago Press, Chicago.

palaeontologia electronica · multiple sets of elements. in hadrosaurids the pleu-rokinetic hinge...

Documents