Morphometric analysis of evolutionary trends in the ceratopsian postcranial skeleton

BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors nonprofit publishers academic institutions researchlibraries and research funders in the common goal of maximizing access to critical research

MORPHOMETRIC ANALYSIS OF EVOLUTIONARY TRENDS IN THECERATOPSIAN POSTCRANIAL SKELETONAuthor(s) BRENDA CHINNERYSource Journal of Vertebrate Paleontology 24(3)591-609 2004Published By The Society of Vertebrate PaleontologyDOI httpdxdoiorg1016710272-4634(2004)024[0591MAOETI]20CO2URL httpwwwbiooneorgdoifull1016710272-4634282004290245B05913AMAOETI5D20CO3B2

BioOne (wwwbiooneorg) is a nonprofit online aggregation of core research in the biological ecological andenvironmental sciences BioOne provides a sustainable online platform for over 170 journals and books published bynonprofit societies associations museums institutions and presses

Your use of this PDF the BioOne Web site and all posted and associated content indicates your acceptance ofBioOnersquos Terms of Use available at wwwbiooneorgpageterms_of_use

Usage of BioOne content is strictly limited to personal educational and non-commercial use Commercial inquiries orrights and permissions requests should be directed to the individual publisher as copyright holder

MORPHOMETRIC ANALYSIS OF EVOLUTIONARY TRENDS IN THE CERATOPSIANPOSTCRANIAL SKELETON

BRENDA CHINNERYMuseum of the Rockies Montana State University 600 W Kagy Blvd Bozeman MT 59717 bchinnerymontanaedu

ABSTRACTmdashThe cranial anatomy of ceratopsian (ldquohornedrdquo) dinosaurs is well understood but the postcranial skeletonhas been largely ignored in previous studies of phylogeny evolution and function In order to study morphologicaldifferences among ceratopsian postcranial elements and to compare evolutionary patterns multivariate (Principal Com-ponents Analysis) and bivariate methods were used to analyze linear measurement data and the shape methods Resis-tant-Fit Theta-Rho Analysis (RFTRA) Least-Squares Theta-Rho Analysis (LSTRA) and Euclidean Distance MatrixAnalysis (EDMA) were applied to biological landmark data Results of the analyses show that size is the primary changethrough ceratopsian skeleton evolution Elements display positive allometry and increasing structural support is evidentespecially in the radius and fibula

Phylogenetic distribution of ceratopsian postcrania agrees with the skull material included in recent cladistic analysesPsittacosaurus elements are in many ways derived relative to those of non-ceratopsid neoceratopsians and evidencesuggests that Psittacosaurus was bipedal while non-ceratopsid neoceratopsians were not except maybe for Udanocera-tops With increasing body size neoceratopsian limbs bowed laterally

The variety of methods allows for unbiased interpretation of results provides more information than any one methodand provides controls for each other However sample sizes are not ideal and all results should be treated with cautionat this time

INTRODUCTION

Ceratopsia are a clade of ornithischian dinosaurs comprised ofthe family Psittacosauridae and the infraorder Neoceratopsia(Sereno 1984 1986 1990 Dodson and Currie 1990) Althoughthe oldest ceratopsians known may have lived as early as theEarly Cretaceous (Chaoyangsaurus Zhao et al 1999) the cladeis mainly known from the radiations that occurred during themiddle and Late Cretaceous of Asia and the Late Cretaceous ofthe western United States and Canada

Ceratopsian postcranial anatomy varies throughout the cladePsittacosauridae a clade of small herbivores known only fromthe Early Cretaceous of Asia share the diagnostic ceratopsianrostral bone with the other members of Ceratopsia Psittacosaursare characterized by long hind limb elements relatively shortforelimb elements and a reduced manus (Sereno 1990) suggest-ing that members of this clade were facultatively bipedal Psit-tacosaurids are relatively small dinosaurs not more than twometers in length

Neoceratopsia (Ceratopsia excluding PsittacosauridaeSereno 1984 1986) comprise the majority of Ceratopsia Thesedinosaurs first appeared in the Early Cretaceous of Asia as smallpossibly bipedal herbivores and over time evolved into large-bodied graviportal quadrupeds Basal neoceratopsians vary inmorphology continental distribution and fossil age The groupis paraphyletic (Sereno 1986 2000 Chinnery and Weishampel1998) consisting of taxa more closely related to the ceratopsidsof North America Ceratopsidae are a monophyletic clade com-prised of two subfamilies Centrosaurinae and ChasmosaurinaePostcranially these subfamilies have been considered extremelysimilar (Dodson and Currie 1990) All are robust quadrupedsand the functional changes associated with this graviportal life-style were thought to be indistinct at the subfamilial level andeven less distinct among genera within the subfamilies

The present study of ceratopsian postcranial anatomy involvesthe application of morphometric analyses treated in a phyloge-netic context This project was conceived with the realization of

how essential it is that we understand the evolution and growthof the ceratopsian postcranial skeleton because it comprises themajority of the skeletal mass is responsible for stature and lo-comotion (among other functions) and evolved through timealong with the cranial skeleton The postcranial skeleton hasbeen neglected in the past because phylogenetic analyses haverelied primarily on cranial characters (Sereno 1984 1986 2000Chinnery and Weishampel 1998 Makovicky 2002 Dodson etal in press) The postcranial skeleton is not only important forphylogenetic analysis but also for any discussions of functionbehavior or growth within the clade

Linear measurement and landmark data are available in theSupplemental Materials (see Appendix) online at httpwwwvertpaleoorgjvpJVPcontentshtml as are additional re-sults and text not able to be included here Girdle and limbelements of all available ceratopsian taxa were compared at sub-famial and generic levels Only the former is included in thispaper because comparisons of taxa at the generic level are sim-ply too long to include However they are available in theSupplemental Materials Additional objectives of the projectwere to document and describe taxonomic differences in theceratopsian postcranial skeleton and to compare growth changesto evolutionary changes in postcranial skeletal anatomy Theapplication of a variety of morphometric techniques ensuresrelatively non-biased results and provides quantified and testableresults as well as data for inclusion in future analyses

Morphometry consists of techniques used to capture the shapeand size of an object (usually but not always biological) whichcan then be compared with another object The techniques haveevolved through time from the deformation grids of DrsquoArcyThompson (1917) to the myriad methods used today (eg Chap-man 1990 and references therein Richtsmeier et al 1992 Chin-nery and Chapman 1998) As an organism develops stressesincluding gravity movement posture and growth cause changesin shapes and sizes of skeletal (and soft tissue) elements Suchchanges are often subtle and may be missed without the use ofmorphometrics Quantitative approaches have been used to cap-ture and clarify evolutionary changes in both living and extinctorganisms as well as to provide the means to study sexual di-morphism posture and locomotor changes ontogenetically and

Present address 1620 Green Meadow Lane Spring Branch Texas78070

Journal of Vertebrate Paleontology 24(3)591ndash609 September 2004copy 2004 by the Society of Vertebrate Paleontology

591

phylogenetically and to identify heterochronic processes (egreferences in Chapman 1990) These methods provide a power-ful means to study growth skeletal evolution and functionwhich is the reason for applying them to the study of ceratopsianpostcrania

Current phylogenies of neoceratopsians (Makovicky 2002)and ceratopsids (Dodson et al in press) provide the baseline forthe comparisons in this project For placement of postcranialattributes these two trees have been combined into a compositetree (Fig 1) Neither of the original trees is disrupted but theresulting tree is not a true tree in the cladistic sense The basictree topology includes Pachycephalosauria as the outgroup toCeratopsia Ceratopsia include Psittacosauridae and Neocera-topsia Neoceratopsia are comprised of Ceratopsidae and a para-phyletic assemblage of basal forms

Institutional AbbreviationsmdashAMNH American Museum ofNatural History New York ANSP Academy of Natural Sci-ences of Philadelphia BHI Black Hills Institute South DakotaCM Carnegie Museum Pittsburgh DMNH Denver Museum ofNature and Science Colorado GPM Glenrock PaleontologyMuseum Wyoming KUVP Kansas University Vertebrate Pa-leontology MCZ Museum of Comparative Zoology Massachu-setts MNA Museum of Northern Arizona Flagstaff MORMuseum of the Rockies Montana NMC Canadian Museum ofNature Ontario NMMNH New Mexico Museum of NaturalHistory Albuquerque OMNH Oklahoma Museum of NaturalHistory PIN Paleontological Institute Russian Academy of Sci-

ences Moscow ROM Royal Ontario Museum TorontoRTMP Royal Tyrrell Museum of Palaeontology Canada TCMThe Childrenrsquos Museum of Indianapolis TMM Texas MemorialMuseum Austin UCM University of Colorado Museum Boul-der UCMP University of California Museum of PaleontologyBerkeley USNM United States National Museum WashingtonDC UTEP University of Texas at El Paso YPM Yale-PeabodyMuseum Connecticut ZIN Zoological Institute Russian Acad-emy of Sciences St Petersburg ZPAL Palaeozoological Insti-tute of the Polish Academy of Sciences Warsaw

MATERIALS

Both isolated specimens and those from bone-beds are in-cluded in the data set The primary data base consists of elementsfound in direct association with diagnostic skull material as thetaxonomy of Ceratopsia thus far has been based on skull char-acters Ceratopsian anatomy is fairly well understood in terms ofgeneral descriptions particularly in terms of comprehensive cra-nial anatomy studies (eg Hatcher et al 1907 Lull 1933 Brownand Schlaikjer 1940 Maryanska and Osmolska 1975 Sereno1990 Dodson and Currie 1990 Dodson 1996 Forster 1996Sampson et al 1997) Previous postcranial studies include thoseof Adams (1987 1991a b) Lehman (1989) Tereshchenko (19911994 1996) and Johnson and Ostrom (1995) among others Themorphometric methods applied to the study of the ceratopsianpostcranial elements add quantitative data to the qualitative de-scriptions in the literature and allow for comparisons of speci-mens across both North America and Asia without the need totransport specimens

The girdle and limb elements comprise the database for themorphometric methods used in the current study to identifytrends in ceratopsian postcranial evolution (Fig 2) The largersize of girdle and limb bones provided more complete data thanwere available with vertebrae ribs manus and pes elementsand were also more likely to be previously identified to genericlevel due to associated cranial material Accessibility of thespecimens was occasionally limited because they were locatedbehind glass or on a mounted skeleton these could not be mea-sured Occasionally it was not possible to exactly duplicate theorientation of these elements in photographs measurements orscale and these elements were not included in the analyses Thetotal number of elements included in the analyses is 520 somefrom the same individual Taxa represented include Psittacosau-rus nine basal neoceratopsian genera and 12 ceratopsid generaSeveral undescribed specimens were included in the analyses(ie TCM 2001964 and MOR 300) because they are new basalneoceratopsian taxa being described and may be compared tothe other basal taxa (Chinnery pers obs) Anatomical descrip-tions and figures as well as specimen lists can be found in theSupplemental Materials

All measurement data were taken directly from specimens (nophotographs or casts were used) and were included in the analy-ses Several measurements were estimated but most were leftincomplete in order to provide the most accurate data possibleData from opposite sides of a single individual were averaged Ifthe measurement set for a specimen was not complete the speci-men was not included in the Principal Components Analyses(PCAmdasha minimum percentage of data is required for inclusionin the analyses)

Included in the shape analyses were all mostly complete andnon-deformed elements Partial elements were documented inphotographs for later inclusion in the analyses The landmarksare homologous and analogous and were chosen to highlightelement areas of interest Missing landmarks were estimatedonly on nearly complete specimens in order to provide full datasets for these analyses The Resistant-fit Theta Rho Analysis(RFTRA) and Least-Squares Theta Rho Analysis (LSTRA) in-

FIGURE 1 Composite cladogram of Ceratopsia created from modi-fied versions of cladograms from Makovicky (2001 in medium gray) andDodson et al (in press in black) Internode 1 refers to postcranial at-tributes of all neoceratopsians internode 2 refers to postcranial attrib-utes of ceratopsids internode 3 refers to attributes of chasmosaurinepostcrania and internode 4 refers to attributes of Psittacosaurus post-cranial elements These attributes are listed in Table 1

JOURNAL OF VERTEBRATE PALEONTOLOGY VOL 24 NO 3 2004592

clude more specimens than those analyzed with Euclidean Dis-tance Matrix Analysis (EDMA) because the latter program can-not handle extreme variability well (Richtsmeier pers commsee Methods section) In fact most EDMA analyses not onlyinclude fewer specimens but also compare fewer landmarks Ac-cordingly some of the comparisons among the various speci-mens include different data sets although variation in data setswas kept to a minimum

The elements were carefully oriented in identical positions formeasuring and photographing so that the results were compa-rable Please see the Supplemental Materials for discussion ofthe various orientations used and Fig 2 for diagrams of themeasurements taken and landmarks digitized

Potential for error in a study such as this is high and thepossible sources numerous In order to decrease this potentialonly mostly to completely undistorted elements were includedEven so some distortion is inevitable with a data set of this sizeand is most likely responsible for the numerous outliers in thePCAs Some outliers have already been eliminated from the finalanalyses while others have been inspected and judged to showvariation rather than deformation Further analyses will help todecipher the remainder of the outliers All measurements weretaken with the elements at identical orientations usually withone area held in a horizontal position (based on a liquid leveler)Photographs were likewise taken by leveled camera equipmentand from identical orientations (warping of images due to shapeof camera lenses is equal throughout the analyses and thereforedoes not affect the results) and biological landmarks were cho-sen based on an error study conducted prior to this project aswell as previous studies (Chapman and Brett-Surman 1990Weishampel and Chapman 1990) Most landmarks chosen areanalogous including extreme points and inflection points be-cause postcranial elements lack numerous homologous land-marks Results of the shape analyses were analyzed with thechoice of data taken into account

METHODS

For vertebrates in general morphometric methods have beenused to understand variation in postcranial skeletal anatomy(eg Dodson 1975a 1980 Oxnard 1978 1983 1984 Dawson1994 Van Valkenburgh 1987 Weishampel and Chapman 1990Chapman and Brett-Surman 1990 Watabe and Nakaya 1991)However in Neoceratopsia these approaches have focused al-most entirely on skull material (Gray 1946 Lull and Gray 1949Dodson 1976 1990 1993 Forster 1990 1996 Chapman 1990)in large part because much of the taxonomy of this group isskull-based Exceptions to this overwhelming emphasis on skullsare works by Lehman (1989 1990) which focused on variationwithin one ceratopsid genus (Chasmosaurus) and Chinnery andChapman (1998) in which a select group of ceratopsid limb el-ements was compared using several morphometric methods

Linear Measurement Data Collection and Analysis

Because bone adapts to stress through life and through evo-lution by changing shape changes can include increasing or de-creasing length different size or shape of crests and articularareas and different distributions of cortical bone tissue Many ofthese changes can be captured with linear measurements andanalyzed using bivariate and multivariate analyses Some of theearlier morphometric analyses of ceratopsians were based onthese methods (Brown and Schlaikjer 1940 Gray 1946) whendata collection was relatively simple and ratios and bivariateanalyses required little computer dependence Biometric meth-ods both bivariate and multivariate have been used since then(Dodson 1975a 1975b 1976 1980 Chapman et al 1981 Chap-man et al 1997 Chinnery and Chapman 1998) because they are

useful in allometric studies of growth evolutionary change andsexual dimorphism Bivariate analyses supply information com-paring two measurements and multivariate approaches includ-ing Principal Components Analysis (PCA) combine all variablesinto a single analysis

PCA is a multivariate approach used to determine similaritiesand differences in forms in addition to the variation most oftenattributable to absolute size PCA has been successfully appliedto the morphometric study of trilobites (Hughes and Chapman1995) painted turtles (Jolicoeur and Mosimann 1960) Alligator(Dodson 1975b) prosauropod dinosaurs (Weishampel andChapman 1990) pachycephalosaur dinosaurs (Chapman et al1981) Protoceratops (Dodson 1976) and ceratopsids (Forster1996 Chinnery and Chapman 1998) Principal components ordirections of variation are computed from linear correlationsset in a matrix between variables Thus all measurement data ofinterest can be included in a single analysis Size difference isusually contained in the first component especially among rep-tiles which can be evergrowing (Jolicoeur and Mosimann 1960Dodson 1975b) Other components are thought to depict shapedifferences or allometric trends within or among taxonomic orage groups (Jolicoeur and Mosimann 1960 Dodson 1975b)Raw data are typically log-transformed in order to reduce thenumber of residuals and amount of heteroscedasticity (as perReyment et al 1984) and input into a statistics package such asSystat Options within the Systat 8 and 10 programs applied inthis project were a correlation matrix (covariance provided simi-lar results) and pairwise deletion the latter to include as manycases as possible even with missing data The output of a PCAincludes a component loadings matrix which describes theamount of variation contained in each component as well as thevariable loadings within each component A highly positive ornegative loading signifies that a particular variable contributedmore than the others to the variance in that component Bivari-ate plots of the first and second (and other) components depictclustering if it exists among the groups included in the analysisAs clustering along the x-axis (the first principal component)typically indicates size difference the focus is on any clusteringoccurring along the y-axis that would indicate shape differencesapart from size

The more general directions of variation obtained from thePCA can be narrowed to the particular measurements causingthe majority of variation in principal components through theuse of bivariate analysis Thus bivariate analysis is a very pow-erful addition to PCA because overall results can be pinpointedto specific areas of the bone for a more precise understanding ofthe allometric patterns at work and the biological areas that areaffected Slopes can be compared among taxa for determinationof allometric changes through evolution following Huxleyrsquos(1932) equation of allometry (Y bXa) with X as the measure-ment depicting overall size of the element or organism Y as anadditional measurement of a part of this element or organismthat indicates the statistical relationship between X and Y b asintercept on the Y axis and a as the allometric coefficient (Dod-son 1975b) Slopes over 10 depict positive allometric trends andthose under 10 display negative allometry Slopes of differenttaxonomic groups can be compared to find evolutionary differ-ences among them

Linear measurements of ceratopsian elements were first sub-jected to PCA in order to identify those variables that controlledsize and shape variation in the sample of ceratopsian postcraniaThereafter these important variables were regressed againstoverall size of the particular elements (usually element length)and Reduced Major Axis (RMA) slopes were computed for theentire data set using Microsoft Excel and plotted using the pro-gram Slidewrite Individual slopes were tested for significant dif-ference using the post-hoc Tukey test available in the Analysis ofCovariance (ANCOVA) part of the Systat programs This test

CHINNERYmdashCERATOPSIAN MORPHOMETRICS 593

FIGURE 2 Linear measurement and biological landmark data collected on the ceratopsian material visualized on representative elements ofTriceratops (Hatcher et al 1907) Measurement data are indicated by numbered lines on the top or left hand figure in each set while landmarks areshown as numbered dots Elements figured are A scapulocoracoid in lateral (top of each pair) and ventral views B humerus in dorsal (on the left)and lateral views C ulna in lateral (on the left in each pair) and cranial views D radius in cranial view E fibula in cranial view F ilium in ventraland lateral views for the measurement data and dorsal and lateral views for the landmark data G ischium in lateral view H pubis in ventral (top)and lateral views I femur in cranial (on the left in both pairs) and lateral views J tibia in cranial (on the left in both pairs) and lateral view


allows for and includes the error of both the X and Y variablesin determination of significant statistical difference amongslopes RMA slopes are more robust than least-squares slopesbecause variation on both axes is accounted for As ceratopsianbody size increased specific areas of each element changed atdifferent rates The RMA slopes allow an interpretation of ex-actly how the individual elements changed providing possibleinformation on mechanical properties of the bones and changesin muscle size and use

In a bivariate plot clustering of individual element points mayoccur along different slope lines allowing for the discussion ofdiffering growth or evolutionary trends Clustering along a singleslope line is less dynamic but still provides important informa-tion about the groups included

Included with the Supplemental Materials are tables for allelements listing sample sizes correlation coefficients RMA val-ues (allometric coefficients) and intercepts Also listed are theRMA slopes that are statistically significantly different fromisometry (slope 1 p 005 and 001) and the group slopesthat are significantly different from each other (p 005) Bothare based not on RMA slopes but on least-squares values as theresults are essentially equal as long as the correlation coefficientis high (r 098 or higher Ruff pers comm) Coefficientsbetween 095 and 098 are close enough that most differences willremain significant Those significantly different slopes with cor-relation coefficients below 095 need to be accepted with cautionas the slopes may not differ at p 005 if RMA algorithms areapplied

Biological Landmark Techniques

In addition to linear measurements three and two-dimensional biological landmarks and the relative distancesamong them can provide a great deal of information on formchange during ontogeny and phylogeny Although the large size

and inaccessibility of mounted elements precluded the possibilityof collecting three-dimensional landmark data for this analysistwo-dimensional landmark data were obtained from images ofthe ceratopsian material using the program TPSDig version 118(Rohlf 1998) These biological landmarks fall into two catego-ries Biologically homologous landmarks are assumed to be de-velopmentally equivalent on all specimens (Bookstein 1991)and include three-way cranial sutures foramina and other dis-tinct and discrete points Analogous landmarks on the otherhand are not strictly homologous but still provide informationon similarity of shape as long as the results are interpreted inlight of the data chosen Analogous landmarks include points ofinflection on element borders extreme points (ie the most dor-sal point of the element) and distance points (such as the mid-shaft point) Both homologous and analogous biological land-mark data were collected for this project Information from pre-vious morphometric studies (including Adams 1987 Lehman1989 Chapman and Brett-Surman 1990 Chinnery and Chap-man 1998) and other sources (Romer 1956 Dodson 1975b)were used to determine suitable landmark sites

Analyses were conducted using Least Squares Theta-RhoAnalysis (LSTRA) Resistant-Fit Theta-Rho-Analysis (RFTRA)and Euclidean Distance Matrix Analysis (EDMA) The first twomethods compare forms using superimposition fitting one formto another while the third method compares inter-landmark dis-tances without the use of superimposition The output graphicsof the three methods are illustrated in Fig 3

Superimposition of forms is usually accomplished by using ei-ther a least-squares method as in LSTRA or the more resistantrepeated means method of RFTRA (Siegel and Benson 1982Benson et al 1982 Chapman 1990 Rohlf and Slice 1990 Chap-man et al 1997) The former uses the average distances of allpoints to position one form on another thereby providing theopportunity to study form change in the entire specimen It is

FIGURE 2 (Continued)


particularly applicable when the specimens being studied grow atsimilar rates and directions throughout the element Howevergrowth and evolutionary change is rarely average throughout anelement as can be seen in the bivariate plots of ceratopsianpostcranial elements By averaging the inter-landmark distancesgreater change in one area is redistributed throughout the ele-ment and potentially valuable information is lost (Siegel andBenson 1982 Benson et al 1982) In addition a large change inone area of an element will ldquoswamp outrdquo other more subtlechanges as the large change will be redistributed throughout theelement (Fig 3A1 A2)

Resistant-fit methods (for example RFTRA) place the over-lying form on the base form in a different mannermdashthe mostsimilar areas of the two forms are lined up first The placementis made regardless of the importance of the area functionally ordevelopmentally This procedure allows for comparison of subtlechanges that could otherwise be ldquoswamped outrdquo by a large de-formation in one area of the shapes (Siegel and Benson 1982)and for more precise study of localized change (Fig 3B1 B2)Resistant-fit methods are more robust when over half of thepoints fit closely if one-half or fewer points fit closely the su-perimposition of the forms is arbitrary As biological elementsdo not usually grow and evolve similarly throughout the elementRFTRA is probably the more accurate and valuable of the twosuperimposition methods in the study of ceratopsian postcraniaHowever when using RFTRA one must understand that theresults emphasize localized change and may de-emphasize moreimportant but more subtle changes in the element With LSTRAresults for comparison though and especially with EDMA re-sults included as well RFTRA results can be verified and canprovide a great deal of information on element growth and evo-lutionary change

EDMA uses identical landmark data but compares formswithout the use of superimposition (Lele and Richtsmeier 1991Lele 1991 Richtsmeier et al 1992 Lele and Cole 1996) Eachlinear distance between two landmarks on a form is compareddirectly with the same distance on another form This negatesany dependence of the linear distance on surrounding distancesthus solving all of the problems associated with the superimpo-sition aspects of LSTRA and RFTRA Until recently EDMAhas been applied primarily to three-dimensional landmark dataof skull material and prior to this study has not been applied to

postcranial material and in direct comparison with the othermethods included

The results of EDMA differ from those of LSTRA andRFTRA and require different interpretation (Fig 3C1 C2) Inthe latter two methods both forms that are being compared arerepresented by the graphical output of the chosen landmarks andwith vectors signifying the direction and amount of change be-tween each pair of landmarks EDMA provides a data matrix ofthe inter-landmark distances only (a Form Difference Matrix) Aratio is generated of the distance between two landmarks in thedenominator (the second specimen) and the numerator (theequivalent to the base specimen of RFTRA and LSTRA in thisproject) for each pair of landmarks These ratios (all are eitherbelow 10 [the numerator is relatively longer] equal to 10 [theinterlandmark distance is equal] or above 10 [the denominatoris relatively longer]) can be represented by colored lines drawnfrom landmark to landmark on a sample illustration In this proj-ect distance ratios less than 10 are indicated in gray and thosegreater than 10 in black and dotted black lines and all are drawnon illustrations of Triceratops elements Black lines indicate ex-pansion in relative size of the non-base specimen and gray linesindicate relative contraction of distance in the non-base speci-men Directions of change are not known in EDMA as for ex-ample a longer interlandmark distance may be due to either orboth of the landmarks moving relative to the other Howeversome changes such as rotation of a part of the element arerepresented by longer inter-landmark distances in one area ofthe element and shorter ones in another area

The data collected were analyzed using all of the above mor-phometric methods with the goal of combining various types ofoutput for the determination of evolutionary changes in the post-cranial anatomy of ceratopsian taxa All data were placed intoone of four groups Psittacosaurus basal neoceratopsians Cen-trosaurinae or Chasmosaurinae The basal taxa do not form amonophyletic group but are grouped together for the purposesof determining the differences in element shape between thesesmall taxa and the much larger ceratopsids

Discussed here are comparisons of Psittacosaurus to basal neo-ceratopsians basal neoceratopsians to ceratopsids and centro-saurines to chasmosaurines reflecting the phylogenetic contextof these taxa When more than one example of an element wasavailable for inclusion an average specimen was generated byeach method In effect the average specimen of the basal neo-ceratopsian scapula (for example) is a composite including allbasal neoceratopsian scapulae This provides a more represen-tative example of the element for that group and helps to lessenthe effect of preservational deformities Use of average forms isnecessary when doing form comparisons as using each separatespecimen would take an exorbitant amount of calculation timeand the selection of a single representative can provide biasedresults Admittedly averaging basal neoceratopsians is awkwarddue to paraphyly of the taxa but the purpose of this part of theproject was to study general trends in postcranial anatomy due toevolution (at least some of the basal neoceratopsians lived priorto the ceratopsid dispersal [Chinnery and Weishampel 1998] andin general they exhibit a suite of more basal characters) sizechange (most of the basal forms are much smaller than the cera-topsids) and function (many of the basal forms have similaranatomical characteristics)

Generic-level comparisons were also made for each elementand all landmark methods were applied to these as well Due tolength constraints these results are not included in the currentpaper but can be found in the Supplementary Materials Gen-erally these individual comparisons were conducted for one ofseveral purposes Enigmatic taxa including GraciliceratopsAvaceratops and Brachyceratops were compared with others todetermine how close in shape the elements are to those of (sup-posed) closely related taxa Basal neoceratopsian genera from

FIGURE 3 Comparison of RFTRA LSTRA and EDMA outputgraphics on a simple shape modified from Benson et al (1982) TheEDMA graphics were produced by drawing lines between landmarksbased on the Difference Matrix (DMmdashsee text for explanation) Theshapes change differently according to RFTRA (A) and LSTRA (B)and EDMA (C) shows only relative distance changes SuperficiallyEDMA in these cases appears to resemble the LSTRA graphics How-ever values in the DM can be used to determine relative magnitude ofchange in the interlandmark distances and the potential errors of super-imposition are avoided


the same continent (either Asia or North America) or from eachcontinent were compared for the purpose of seeing how close therelationships or functional attributes are between the two conti-nents Finally members within Centrosaurinae and Chasmosau-rinae were compared to see if postcranial attributes could befound at the generic level in these two subfamilies

The results are of a qualitative nature whereas the methodsused are all quantitative in nature Results from all five methodsemployed were compared prior to the final interpretations andconsequently many of the evolutionary and functional trendsdiscussed here are qualitative in nature Output parameters ofthe shape analyses (difference coefficients in RFTRA andLSTRA and a distance ratio number in EDMA) provide thequantitative output for these methods and can be found with theSupplemental Material These parameters were combined withresults of the multivariate and bivariate analyses prior informa-tion of element differences and knowledge of functional areas ofthe elements For example LSTRA may superimpose a speci-men differently onto a base specimen than does RFTRA due tothe averaging of the landmark distances RFTRA lines up thelandmarks with the least amount of apparent change first but thesuperimposition can still be faulty EDMA graphics tell whichdistances are longer and shorter but not which landmarks havechanged position to cause the distance changes The combinationof all results is necessary to gain the most accurate conclusionsand therefore the discussions are more qualitative than quanti-tative

RESULTS AND DISCUSSION

Evolutionary Patterns

Evolutionary patterns among ceratopsian dinosaurs can be in-dicated in several ways Bivariate plots of the components in aPCA show clustering if the groups are different in shape Thefour groups included in this project Psittacosaurus basal neo-ceratopsians centrosaurines and chasmosaurines are by defaultalready clustered in each graph due to size (indicated in Com-ponent 1 the x-axis variable [C1]) No Psittacosaurus or basalneoceratopsian elements are as large as those of the adult cera-topsids The y-axis in a C1C2 or C1C3 plot is the axis of interestand any clustering along this axis indicates form differences andpossibly phylogenetic differences If the only distinct clustering isseen along the x-axis the conclusion is that the element increasesin shape solely in response to size increase

The RMA slopes in the bivariate plots depict general evolu-tionary trends RMA slopes above 10 indicate that the y-variable increased in size at a greater rate than the generalizedlsquosizersquo variable (usually length of the element) and slopes below10 indicate the opposite Although Psittacosaurus is the out-group to Neoceratopsia in the majority of measurements all datapoints are located close to the RMA line suggesting much simi-larity in the postcranial elements within Ceratopsia This simi-larity is forced in part by the inclusion of all specimens in theregression equations However relative positions of the groupsaround the regression line can still be compared Different slopesfor each group in the analyses show allometric patternsmdashas sizeincreases traits change at different rates Not many group slopesdiffer from each other at a statistically significant level andtherefore the ones that do differ are important for phylogeneticdistinction

General Evolutionary Trends Throughout Ceratopsia

The groups included in the analyses cluster differently in thevarious PCA bivariate plots of C1 and C2 The first componentof nearly all analyses contains a very large percentage of thevariance verifying that size is the primary change within Cera-topsia in these elements Subsequent components contain such

small percentages of the variance that the following interpreta-tions must be viewed with caution However size changethrough evolution is understood in this clade we are interestedin other shape changes and therefore the small differences areimportant Psittacosaurus elements generally cluster together (itis only one genus while the other groups contain more than onegenus so this clustering is expected) and the more derived con-dition of some of the elements is apparent in the clustering ofthese small elements along C2 with the much larger chasmosau-rine elements especially in the scapula coracoid femur andtibia (Figs 3A 3B and 3C) Closer relationships in these analy-ses to the basal neoceratopsians might be expected if the neo-ceratopsian common ancestor was psittacosaur-like and espe-cially considering the greater similarity in size of the elements inpsittacosaurs and basal neoceratopsians compared with those ofthe ceratopsids From these analyses it appears however thatevolution within Psittacosauridae along with element differencesdue to functional changes (see Functional Morphology section)have reduced similarities in many postcranial elements withthose of basal neoceratopsians

Basal neoceratopsians plot similarly to centrosaurines alongthe C2 axis in the ilium (first PCA) ischium (Fig 4G) and tibia(second PCA) and are located with the chasmosaurines in theradius (first PCA Fig 4F) femur (first PCA Fig 4B) tibia (firstPCA Fig 4C) and fibula Basal neoceratopsians cluster lessthan the other groups reflecting the paraphyly of the groupespecially in the scapula (Fig 4A) coracoid humerus and fe-mur However all plot apart from the ceratopsids in the radius(second PCA Fig 4D) ilium (Fig 4I) and pubis (Fig 4E) Thismixture of relationships shows the more basal nature of the basalneoceratopsians and that various areas of various elementsevolved differently some remaining similar to the basal neocera-topsian condition and others changing to a greater degree

Centrosaurines routinely cluster apart from chasmosaurines inthe PCA plots In the scapula (first PCA Fig 4A) coracoidradius (first PCA Fig 4F) ischium (Fig 4G) pubis (Fig 4E)and tibia (Fig 4C) the centrosaurines cluster primarily on thepositive side of the C2 axis while the chasmosaurines are typi-cally located on the negative side (apart from outliers) Overlapon the C2 axis of these two groups occurs in some of these(coracoid and tibia) but the rest are more discrete clusters Di-visions at the subfamily level are still apparent but positions arereversed in the second scapula PCA as well as the radius iliumand fibula plots in which the centrosaurines are primarily lo-cated below the zero axis of C2 and chasmosaurines above Thusthese two monophyletic groups quite definitely exhibit postcra-nial differences at the subfamily level The variables contributingto the location of these taxa along C2 seem fairly arbitrary as faras the above groupings are concerned (no similarity of variablesoccurs in the clustering of centrosaurines positively in specificelements and negatively in others)

Taxonomic variation within the subfamilies of Ceratopsidaecan be found in the humerus in which two distinct groups areapparent in each subfamily on the C2 axis of the first PCA(greater dorsoventral width of the shaft on the positive side ofthe zero axis and wider lateral side of the proximal end below theaxis Fig 4H) and in the second humerus PCA in which bothgroups are spread across C2 Centrosaurines also do not clusterin the tibia PCA suggesting interspecific variation in this ele-ment (the main contributing variables being the proximal widthand the calcaneal facet width Fig 4C) Although division of thesubfamilies is most likely due to intrataxonomic variation evenwhen genera are themselves divided sexual dimorphism is aplausible alternative hypothesis to be considered Further analy-ses will help to resolve this issue

Within Centrosaurinae Pachyrhinosaurus elements cluster to-gether in the coracoid PCA (with large glenoid fossae) and Cen-trosaurus humeri cluster along with Styracosaurus ones (with


FIGURE 4 Selected bivariate plots of the first and second components of Principal Components Analyses (PCA) of ceratopsian postcranialelements Component loadings tables are available with the Supplemental Materials (Appendix)


wide dorsoventral shafts) Centrosaurus and Styracosaurus hu-meri are separated from those of Einiosaurus MonocloniusPachyrhinosaurus and Avaceratops (heads positioned more me-dially) Therefore this analysis supports the close relationship ofCentrosaurus and Styracosaurus (Fig 1) but does not support thepossibility of the Monoclonius material as juvenile Centrosaurus(although this does not preclude the nomen dubium materialfrom belonging to other centrosaurine taxa Sampson et al1997)

Several PCA analyses appear to show negative correlationsbetween C1 and C2 including the radius (second PCA Fig 4D)the pubis (through ceratopsids only Fig 4E) and the sinuouspattern seen in the ulna (Fig 4J) These plots show clear patternsof change in the elements as size increases In the radius as sizeincreases the mediolateral width of the proximal end increases ata relatively higher rate (length is decreasing relative to thebreadth of the proximal end as size increases) In the pubis thearticular area becomes taller in the larger elements (relativelength of the element decreases with size) In the ulna the ratesof decrease are higher among the smaller-sized individuals ofeach group As the specimens become larger in size the ratelessens and then increases again with the largest individuals Al-though a hypothesis of size as the primary influence on the C2values is possible evolutionary mechanisms are most likely par-tially responsible due to a difference in the pattern formed by thegrowth analysis (Chinnery pers obs) According to the maxi-mum loadings in the analyses as well as the bivariate plots theC2 values are primarily the relative lengths of the proximal anddistal portions of the ulna As the individuals become larger thedistal length decreases relative to the proximal length at firstabruptly and then at a slower rate before decreasing at a higherrate again Functionally this means that the triceps in-lever isincreasing and the out-lever decreasing therefore increasing thepower of the muscle

The scapula RMA slopes show a greater increase in medio-lateral width than in height of the element as length increasedwith a slope of 1209 versus the lowest height slope of 104 (high-est height slope is 1192 Figs 4A1 A2) This pattern shows agreater relationship between increasing structural support(weight bearing) and increasing size rather than increasing sur-face area for muscle attachments Slopes are significantly differ-ent from isometry except for the articular area height and themaximum height of the scapula both at the proximal end of theelement Thus the more distal part of the element (and thelateral width) increased more through evolution

Coracoid glenoid fossa length increases more with maximumcoracoid length than do the other measurements Only this mea-surement and maximum width of the element differ from isom-etry (at the p 01 level) The groups are not significantly dif-ferent from each other in any measurement suggesting verysimilar allometric trends among them

As humeral length increases mediolateral widths increasewith positive allometry 1215 proximally and 1238 distally Asthe deltopectoral length increases at a low slope with length ofthe element among these taxa (slope 109) the distal portionextends farther laterally with length (slope 1403 Figs 4B1B2) The relationship of the humeral head to the proximal end ofthe element changes with length as the proximolateral cornergrows more than the proximomedial one The remaining areas ofthe element tend to increase with slopes ranging from 12 to 13All slopes are significantly different from isometry even thedeltopectoral crest length slope of 109 the latter due to thecorrelation and intercept values Significant differences betweenthe group slopes occur in nearly all mediolateral width measure-ments the exceptions being the distance from the head to themedial border of the element and minimum shaft width In al-most all graphs the slopes for the Psittacosaurus elements arelowest and those for Chasmosaurinae are highest The low re-

gression slope of the deltopectoral crest length does not accu-rately reflect change in this feature as the scatter pattern iscurvilinear concave to the slope line and reflecting the largecrest lengths of Psittacosaurus the relatively smaller crests of thebasal neoceratopsians and then the increase in crest size withlength of the ceratopsid elements (Fig 5B1)

Proximal width of the ulna increases at a higher rate thandistal width and maximum width of the element across the sig-moid fossa rim increases at a rate greater than those of theremaining measurements (a result of a necessary increase instrength of the articulation with the humerus as the animals be-came larger) All regression slopes differ significantly from isom-etry except for the length of the proximal and distal parts of theulna and none of the group slopes vary significantly The slopepattern is again one of the chasmosaurines having higher slopesthan the other taxa with one exception Length of the ulna fromthe olecranon rim to the distal end of the bone is lowest inchasmosaurines higher in basal neoceratopsians and highest incentrosaurines (Figs 4C1 C2)

The size or robusticity of the radius increased much morethrough the evolution of this group than the other forelimb el-ements with regression slopes between 1285 (craniocaudal shaftwidth) and 151 (mediolateral distal width) and all vary signifi-cantly from isometry Chasmosaurine slopes are again higherthan those of the centrosaurines and the basal neoceratopsians

The ceratopsian pelvic girdle shows the most distinct evolu-tionary changes of the postcranial skeleton especially in terms ofthe eversion of the dorsal ilium the increasing curvature of theischium and the expansion of the prepubis in ceratopsids Al-though these patterns in the pelvic element analyses are appar-ent the mechanisms are less clear Pelvic elements were rare forthe growth analyses and therefore some possibly important in-formation is lost from the evolutionary analyses Ilium widthsincrease at higher slopes than length and height measurementswith slopes of 14ndash17 versus 10ndash12 Ischial and pubic pedunclelengths change the least (slopes 08) Length slopes are similaramong the groups but width and height measurements are char-acterized by the highest slopes occurring with the chasmosau-rines and in nearly all cases the lowest slopes are exhibited bythe centrosaurines The ilium expanded mediolaterally as thelength increased most likely to provide expanded muscle attach-ment sites for the larger limb and tail muscles but this expansionis more pronounced in the chasmosaurines The intermediateposition of the basal neoceratopsian slopes suggests that withlength increase width increase was just as important as in theceratopsids However the small numbers of specimens includedmake this a tentative and preliminary observation

The ischium changed primarily as previously mentioned withgreater curvature of the ischial shaft As articular length in-creased the length of the element (measured in a straight line)increased with the highest slope in chasmosaurines followed bybasal neoceratopsians and centrosaurines Increase in curvatureoccurred with nearly identical slopes in the two ceratopsidgroups and height of the articular area increased at a slightlyhigher rate in basal neoceratopsians than in ceratopsids Againsmall sample sizes and also a lack of a good size variable makethese conclusions very tentative

The bivariate analyses of pubes include only two basal neo-ceratopsian elements and any discussion of the slope patterns inthis group are suspect However in all plots the longer basalneoceratopsian element is noticeably larger suggesting that in-crease in dimensions apart from length occurred at a higher rateamong these more basal forms (Fig 5D)

In the bivariate plots of femora the groups are all located onor very close to the RMA slope suggesting little difference in theelement structure among taxa All measurements are positivelyallometric compared with length of the element the mediolateralwidth from the head to the greater trochanter and the width at


FIGURE 5 Selected bivariate plots of ceratopsian postcranial elements including Reduced Major Axis (RMA) slope lines calculated on allspecimens and RMA slope equations See the Supplemental Materials and text for additional information on the x- and y-axis variables Data aremeasured in millimeters and are log-transformed


one-quarter length having the highest slopes whereas width atthree-quarter length and distal mediolateral width have the low-est (Figs 4E1 E2) Thus throughout the clade the proximal endof the element expands more as length increases than does thedistal end presumably owing to a need to increase strength at thepelvic girdle with increasing weight With phyletic size increasethe lateral condyle increases at a greater rate than the medialcondyle suggesting a shift of the femur laterally and a longerlateral condyle for continued articulation with the tibia or achange in knee biomechanics

The tibia is characterized in the bivariate analyses as beingsimilar in the ceratopsid subfamilies but the plots indicate muchmore variation among basal neoceratopsians In all the chasmo-saurine slope is the highest followed by the slopes of the cen-trosaurines and Psittacosaurus and finally the lower basal neo-ceratopsians Again all width measurements are positively allo-metric with the highest slope indicating that the calcaneal facetexpands mediolaterally more than other areas of the elementwith length (Fig 5F) This suggests a strengthening of the lateralaspect of the ankle and may correspond to the suggested shift inthe femur discussed above

The dimensions of the fibula all increase at high rates com-pared to length from 12 up to 16 and more variability is evi-dent within the groups possibly due to preservational differ-ences (Fig 5G) The highly positively allometric slopes indicatethat the element increased more in robustness as length in-creased than did other hind limb elements a trend similar to thatseen in the forelimb where the radius exhibited the largest re-gression coefficients Again this may be due to the increase instructural support needed to support the larger animals

Shape Differences Between Psittacosaurus and BasalNeoceratopsian Postcrania

The following observations are again drawn on a combinationof all three shape analyses applied Not all figures could be in-cluded but they are available with the Supplementary Materials

The basal neoceratopsian scapula is shorter both proximally(owing to its smaller acromion process) and distally (reflectingexpansion of the scapular blade in Psittacosaurus [Fig 6A1ndashA4]) This group of figures demonstrates the graphical output ofthe various shape analyses The distal expansion of the Psittaco-saurus scapula seems to be concentrated at the ventral cornercausing the distal end to appear shifted dorsally as comparedwith the proximal end The RFTRA output shows this trendespecially well Both the scapular and coracoid contributions tothe glenoid fossa are smaller in basal neoceratopsians than inPsittacosaurus and the humeral head is also larger in Psittaco-saurus The Psittacosaurus coracoid is very diagnostic resem-bling the neoceratopsian coracoid only in lateral view (Fig 6C1C2) In ventral view the prominent division of the Psittacosaurusdistal coracoid is clear separated by a deep groove The basalneoceratopsian humerus is less robust than that of Psittacosau-rus with a smaller deltopectoral crest and a narrower head (Fig6E) No Psittacosaurus ulnae were available for inclusion in thisanalysis The basal neoceratopsian radius is more rounded bothproximally and distally than in Psittacosaurus and the most dis-tal point is located more laterally

The basal neoceratopsian ilium is taller than that of Psittaco-saurus and both the pubic and ischiadic processes are larger(Fig 6K) The iliac contribution to the acetabulum is greatestcompared to length in the basal neoceratopsian element This isalso the case with the ischium the pubic and iliac peduncles ofthe Psittacosaurus ischium are close enough to each other tosuggest little or no contribution of the element to the acetabu-lum

Contrary to the expected reduction of the femoral head owingto the acetabular differences noted above the femoral head of

Psittacosaurus is only slightly smaller in relative size than thoseof basal neoceratopsians (Fig 6P) The distal end of the basalneoceratopsian femur as well as the shaft is less robust the distalarticular areas are smaller and the greater trochanter is betterdeveloped than those of Psittacosaurus This last observation isnotable in light of the much smaller mediolateral proximal tibiain Psittacosaurus The proximal tibia is instead more expanded inthis group in a craniocaudal direction than in basal neoceratop-sians suggesting a difference in orientation of the femur-tibiaarticulation (Fig 6R) Distally the basal neoceratopsian tibia isless robust than that of Psittacosaurus and the astragalar facet islarger in both height and width The same trends occur in thefibula of the two groups

Postcranially Psittacosaurus differs more from basal neocera-topsians than from ceratopsids and the differences discussedabove suggest a combination of plesiomorphic and apomorphiccharacters in the Psittacosaurus postcrania Some plesiomorphic(ancestral) characters are found in the forelimb and pelvis andmay be associated with bipedality Apomorphic (derived) char-acters include the greater robustness of all of the Psittacosauruselements

Shape Differences Between Basal Neoceratopsian andCeratopsid Postcrania

The increase in robustness of the ceratopsid shoulder girdle isexhibited by a larger articular area of the scapula and coracoid(Fig 6B1 B2) This articulation is wider in ceratopsids in ventralview thus mediolaterally strengthening the narrower dimensionsof the elements In particular the medial side of the glenoid fossaarticulation is expanded in the ceratopsid scapulocoracoid

Other examples of increased robustness of the ceratopsidshoulder girdle include a more caudal position of the ventralcurvature of the dorsal border of the scapula in ceratopsids (Fig6B1) In ceratopsids the distal scapular blade is taller and shiftedmore ventrally than in basal neoceratopsians the glenoid fossa isincreased in size and less mediolateral curvature is evidentthroughout the scapula The coracoid body is shifted from asquared shape with a large craniomedial corner in basal forms toa wider shorter shape in ceratopsids with a longer caudal pro-jection (coracoid process)

The basal neoceratopsian scapulocoracoid is more mediolat-erally curved than that of ceratopsids indicative possibly of adifference in position of the element on the rib cage or of adifference in curvature of the rib cage A caudal shift of thescapulocoracoid would help to support a more horizontally po-sitioned humerus (discussed below) as well as possibly increas-ing the stability in the front of the body and in the support oflarger skulls in ceratopsids Skulls and frills of basal neoceratop-sians are relatively smaller than the more elaborate and devel-oped ceratopsid skulls and musculature extending from thescapula to the back of the skull (mm trapezius and levator scapu-lae) may have been more developed in the latter Certainly theoverall mass of the body that the forelimbs needed to supportwas higher in ceratopsids Position of the element may havediffered through either a shift of the element caudally andordorsally in ceratopsids which would position it over more of therib cage and is suggested by a flattening of the rib cage seen insome specimens The shift in position of the scapulocoracoidmay also be explained by an increase in the thorax dimensions inceratopsids due to relatively larger heart and lungs in theselarger quadrupeds (Sampson pers comm)

The humeral head is positioned more medially in ceratopsidsand the distal humerus is shifted slightly in a ventral direction(Fig 6F) The deltopectoral crest of the ceratopsid humerus islarger longer and extends farther laterally The olecranon pro-cess of the ulna is longer in ceratopsids than in basal neocera-topsians and the caudodistal corner is more pronounced (Fig


FIGURE 6 Selected shape comparisons of ceratopsian postcranial elements The element analyses are shown in the following order scapula (A1A3 A4 B1 in lateral position A2 B2 ventral) coracoid (C1 D [lateral] C2 [ventral]) humerus (E F G [all dorsal view]) ulna (H1-2 [lateral view])radius (I J [cranial view]) ilium (K L M1 [lateral view] M2 [dorsal view]) ischium (N lateral view) pubis (O1-3 [lateral view]) femur (P Q [cranialview]) tibia (R S T [cranial view]) and fibula (U cranial view) Resistant-Fit Theta-Rho Analysis (RFTRA) and Least-Squares Theta-Rho Analysis(LSTRA) analyses are labeled and Euclidean Distance Matrix Analyses (EDMA) are indicated by gray and dotted black lines on illustrations ofTriceratops elements modified from Hatcher et al 1907 Please see the Methods section for additional information




6H1 H2) In the ceratopsid radius the proximomedial corner ismore expanded than the proximolateral corner compared tothose of basal neoceratopsians and the most distal point of thedistal end of the element is located more medially (Fig 6I)

The pelvic girdle of the ceratopsid skeleton is both more ro-bust and different in shape from that of basal neoceratopsiansFunction of the hind limb does not appear to have differed muchbetween the two groups as the hind limb elements and the dorsaland caudal vertebrae are very similar The girdle elements areexpanded in order to accommodate the greater musculatureneeded to support the more massive body of the ceratopsids

The major change in the ilium is the lateral eversion of the iliacblade in ceratopsids (Fig 6L) Thus the height of the element isgreater in basal neoceratopsians and the mediolateral width isgreater in ceratopsids The pubic peduncle is shorter in ceratop-sids suggesting a decrease in the iliac contribution to the ac-etabulum However in the basal forms the ischium and pubis areboth taller through the area of articulation with the pubic pe-duncle than in ceratopsids and so a relatively smaller bony ac-etabulum is present in the ceratopsid pelvis

The ischial shaft is more curved in ceratopsids than in basalneoceratopsians (see Adams 1987 for one explanation of theincreased curvature Fig 6N) The pubic peduncle is longer than

that of the basal neoceratopsian ischium in contrast to the iliacpeduncle and the consequence is a caudodorsal rotation of theelement and greater articulation between the ischium and thepubis in ceratopsids The rotation is an alternate explanation forthe increase in curvature of the shaft or is a response to thecurvature

The pubis changes dramatically from the basal forms to theceratopsids as the prepubic process is greatly expanded both inlength and in height (distal expansion) accompanied by a cor-responding decrease in size of the pubic body (Fig 6O1ndashO3)The larger articulation of the ceratopsid ischium and pubis dis-cussed above could explain the decrease in size of the pubicbody as the extension of the pubic body down the ischial shaft ofbasal neoceratopsians may have provided additional support tothe articulation of the elements

As with the rest of the postcranial skeleton the hind limbelements increase in robustness through this part of the phylog-eny The changes are more conservative than those observed inthe fore limb as the hind limb is columnar and the stance issimilar among basal neoceratopsians and ceratopsids Bipedal-ism has been suggested for basal neoceratopsian taxa by someauthors (mainly Graciliceratops [Maryanska and Osmolska1975 Sereno 2000] and Archaeoceratops [Dong and Azuma



1997]) based primarily on limb ratios and immature specimens(Maryanska and Osmolska 1975 Sereno 2000) The forelimband pelvic differences found in Psittacosaurus have not yet beenconfirmed in any basal neoceratopsian taxon As with the radiusthe greatest increases in robustness of the hind limb are seen inthe smallest element the fibula

Femoral differences due to the increase in robusticity includean increase in the height of the greater trochanter and an exten-sion of the cranial trochanter laterally both owing to increases inmoment arms acting around the hip joint (Fig 6Q) EDMAshows that the ceratopsid femoral head is taller than those ofbasal neoceratopsians Because the bony acetabulum is relativelysmaller in ceratopsids (see section on pelvis) the femur eitherfits more tightly into the bony acetabulum in ceratopsids (egthe joint capsule is tighter) or it articulates obliquely relative tothe vertical plane The latter condition would cause a lateral shiftof the distal end of the femur As the distal condyles do not seemto differ between the two groups aside from robusticity theformer is the more parsimonious interpretation

Increase in robustness of the tibia is concentrated on the me-dial side of the element according to both RFTRA and LSTRAresults (Fig 6S) Distally the astragalar facet is similar in widthin the two groups but the medial side of the facet is taller inceratopsids Greater width of the ceratopsid calcaneal facet in-dicates an increase in weight bearing of the lateral aspect of theankle The only shape difference in the fibula apart from theoverall increase in robustness is a slight shift of the distal end ofthe element medially in ceratopsids (Fig 6U)

Shape Differences Between Chasmosaurine andCentrosaurine Postcrania

The chasmosaurine scapula is characterized by greater robust-ness than the centrosaurine element with the distal end extend-ing farther dorsally than in the latter subfamily and the proximalend extending farther ventrally due to the larger glenoid fossarim The coracoid contribution to the glenoid fossa is in contrastlarger in the centrosaurine element (Fig 6D)

The humerus radius and ulna are also more robust in thechasmosaurine skeleton whereas length measurements are rela-tively longer in the centrosaurine elements (Fig 6G J) In theulna the distance from the sigmoid fossa rim proximally is longerin chasmosaurines and the distance from the rim distally islonger among centrosaurines

Shape differences between the pelves of the two subfamiliesinclude greater dorsal and ventral curvature of the iliac blades aswell as a larger acetabulum contribution and greater mediolat-eral width of the chasmosaurine ilium (Fig 6M1 M2) The chas-mosaurine ischium is more curved than that of centrosaurines anobservation made previously by several authors (Adams 1987Dodson and Currie 1990) The pubis is also more robust inchasmosaurines with a more vertically extended distal end andlonger articular area

The chasmosaurine femoral head is taller than that of centro-saurines the medial condyle extends farther distally and thecranial trochanter diverges from the shaft farther distally (Fig6R) The proximal end of the tibia is larger in centrosaurines inlateral view and larger craniocaudally in chasmosaurines (Fig6T) The distal end of the chasmosaurine tibia is also wider me-diolaterally Finally the fibula is wider distally among chasmo-saurines

Phylogenetic Positions of Postcranial Differencesin Ceratopsia

The changes in the postcranial skeleton discussed above wereapplied to a composite cladogram for the purpose of determiningif and how the postcranial characters will support current cla-

distic analyses The cladograms of Makovicky (2002) focusing onbasal neoceratopsians and Dodson et al (in press) with the focuson Ceratopsidae are the most current and comprehensive analy-ses available at this time The composite cladogram (Fig 1) issimply the addition of one to the other no nodes are displaced

Characters supporting Neoceratopsia Ceratopsidae andChasmosaurinae are listed in Table 1 for all elements (whenavailable) Psittacosaurus elements were compared with those ofpachycephalosaurs in an attempt to determine if derived states ofcharacters exist (Maryanska and Osmolska 1974 Maryanska1990) Characters at the generic level were also inserted into thecomposite cladogram for each element (see Appendix) As theshape changes are comparisons of only two (usually composite)groups at a time only those that include taxonomic groups in-cluded in the composite cladogram are included Comparisons oftaxa not recognized by Makovicky (2001 TCM 2001964) andDodson et al (in press Monoclonius Brachyceratops and Ava-ceratops) were not included Thus basal neoceratopsians andgenera within the ceratopsid subfamilies were variously com-pared depending on the specimens available

The results listed in Table 1 and discussed throughout thispaper are qualitative in nature due to the combination and com-parison of the five quantitative methods employed The quanti-tative results individually are not necessarily accurate represen-tations of phylogenetic differences and should be applied withcaution Further quantitative analyses at the generic and specificlevels will allow for more accurate descriptions of phylogeneticcharacters and character states

Functional Morphology

Some of the functional attributes of the elements and the dif-ferences among groups have already been discussed but thestance and locomotor patterns within Ceratopsia remain to beaddressed

Psittacosaurids have previously been determined to be bipedalbased on limb proportions and the loss of manus digits (Osborn1923 Maryanska and Osmolska 1975 Sereno 1990) Severalcharacteristics of the forelimbs of this group support this hypoth-esis or at least suggest a different functional complex than thatseen in neoceratopsians The Psittacosaurus humerus is morerobust than those of the basal neoceratopsians and the delto-pectoral crest extends farther laterally The lateral position of thedeltopectoral crest is associated with the partially sprawling po-sition (and weight-bearing) of the element in neoceratopsians Insome of the basal neoceratopsians (ie TCM 2001964) this crestextends much farther ventrally (caudally it wraps around fromthe side of the element with the head to the back) than in Psit-tacosaurus suggesting a more upright position of the forelimbsof TCM 2001964 (Chinnery pers obs) However the morerobust deltopectoral crests of Psittacosaurus indicate relativelylarge deltoid andor pectoralis muscles tentatively indicating analternate functional use of the forelimb in this taxon Supportingthis interpretation in addition to the crest size is the observationthat the Psittacosaurus humeral head and corresponding glenoidfossa on the scapulocoracoid are both larger than those of basalneoceratopsians This suggests that the shoulder experienced agreater load than in the basal neoceratopsians with their smallerless congruent shoulder joints The Psittacosaurus radius is morerobust than in basal neoceratopsians and has a more roundedproximal end suggesting the possibility of increased mobility(full rotation of the forelimb is not being suggested here butmaybe a more movable manus and forelimb than seen in basalneoceratopsians) and thus some form of manipulation of theenvironment different from what the neoceratopsians were do-ing

The pelvis of Psittacosaurus exhibits traits that suggest either adifference in stance or in locomotion from that of neoceratop-


sians The pubic peduncle of the ilium is located farther craniallyboth articular processes of the ischium are more cranially ori-ented and the acetabulum is smaller If psittacosaurs were fac-ultatively bipedal one would expect the pelvis to be rotatedcaudoventrally the acetabulum to be located farther craniallyand more bone mass farther caudally in the pelvis all to accom-modate the difference in center of gravity These differencessuggest that this was the case and argue for the bipedal stance ofthe group

The pelvic traits of Psittacosaurus support the hypothesis ofbipedality in this group and the forelimb traits discussed abovemay therefore be interpreted as a functional complex for someform of manipulation of the environment such as reaching forplants

Bipedality in basal neoceratopsians has been suggested by themore ventrocaudally projecting deltopectoral crest of the hu-merus as well as limb proportions (hind limb elements relativelylonger than the forelimb elements Maryanska and Osmolska1975) However the deltopectoral crest does not project caudallyin Psittacosaurus suggesting that this character is insufficient todetermine bipedality Other interpretations for the orientationof the deltopectoral crest follow The radius of basal neocera-

topsians is less mobile and none of the manus and pelvic traitsdiscussed above for Psittacosaurus is present In contrast thebroad robust manus and straight femur of basal neoceratopsiansargue against bipedality (caudodistally recurved femora arefound in many archosaurian bipeds Russell 1970) The onlybasal form that exhibits postcranial traits similar to those of Psit-tacosaurus is Udanoceratops which has a radius with a broad flathead and a tall ilium Udanoceratops also exhibits many otherpostcranial characters that are unique among basal neoceratop-sians and because the skeleton is incompletely known the lo-comotor patterns of this genus are unclear

A large shift in body size accompanies the characters separat-ing basal neoceratopsians and ceratopsids Stance may have dif-fered among neoceratopsians with the forelimbs of basal neo-ceratopsians more upright in position than those of ceratopsidsas is suggested by the ventrally projecting deltopectoral crests inthe former In addition the more lateral position of the ceratop-sid humeral head and the ventral shift of the distal end of theelement suggest that it was held more horizontally and laterallythan in the basal neoceratopsians A more horizontal humeruswould explain the taller olecranon process on the ceratopsidulna as well because the triceps muscles would need to be

TABLE 1 Evolutionary trends in ceratopsian postcranial elements as they occur phylogenetically on the composite cladogram in Figure 1

Iternode 1 2 3 4

Scapula Less robust acromion less robustand shifted distally distal shiftof the ventral curve on thedorsal blade

More robust greater height proxand distally shorter distal shiftof ventral curve larger glenoidless laterally curved

Shorter shaft laterally wider lesslaterally curved deeper ventralcurve

Coracoid Longer caudal process lesspronounced rim larger mainbody larger and flatter inventral view less distancebetween rim and caudalprocess

Less distance between rim andcaudal process longer caudalprocess larger foramen

Fossa is sharper and deeper andslightly smaller smallerforamen

Humerus Less robust taller and morerounded head larger and morerobust distal end

More robust longer and largerDP crest head larger and morelateral ventral shift of distalend

More robust DP crest welldeveloped

Ulna More robust larger caudodistalcorner longer olecranon

More robust larger proximalolecranon expanded distalend laterally wider across notch

Radius Less robust proximal end shiftedlaterally most ventral pointmore medial

More robust more expansion ofends most ventral point ismore medial expansion ofproximomedial corner

Slightly more robust the mostventral point is more medialproximomedial and distolateralcorners expanded

Flattened proximalend

Ilium Taller through peduncles morenarrow pubic peduncle pubicpeduncle less cranial

Shorter throughout shorter pubicpeduncle ventral dip of thedorsal border is more cranialhighest point is more cranial

More ventrally curved shorterpre-and postacet blades morecranial peduncles largeracetabulum more lateral widthshorter lateral overhang

Preacetabular blademore ventrallylocated

Ischium Greater curvature shorterthrough iliac peduncles largerarticular area

More robust more curvedproximal shaft curves dorsally

Peduncles pinchedtogether peduncleswider shorter shaft

Pubis Reduction of pubic body andextension of prepubis shorteriliac peduncle

Taller prepubis most ventralpoint of dorsal border is moreproximal

Femur Wider proximal end shorter grtrochanter more narrow distalend shorter distal articulararea

Taller head taller greatertrochanter more distal cranialtrochanter longer articulararea

More robust proximally widertaller head medial condyleextends further distally

Short femoral head

Tibia Proximal end wider and tallerastragalar facet wider andtaller calcaneal facet morenarrow

More robust especially mediallyastragalar facet is tallermedially wider calcaneal facet

More robust more narrowproximally in cranial view

distally wider calcaneal facetwider

Fibula Wider proximally and atmidshaft more narrow distally

More robust taller articular endsmedial shift of proximal end

More robust but more narrowdistal end taller articular ends


strong to keep the elbow in extension and to bear the weight ofthe animal with possibly a corresponding increase in size of themuscle attachment site on the olecranon Results of the PCA inthis study (Fig 4J) support at least a correspondence of olecra-non process size and overall size in ceratopsians but it should benoted that size of the olecranon process does not correspond tosprawling or non-sprawling stance in other groups (ie crocodil-ians have small olecranon processes and stegosaurs [non-sprawlers] have tall ones Carpenter pers comm) Ceratopsidforelimb position has been debated for decades with interpre-tations of posture ranging from sprawling with the humerus al-most horizontal (Gilmore 1905 Sternberg 1927 Lull 1933) toalmost fully erect (Bakker 1987 Paul 1991) A sprawling pos-ture seems to be inherent in the articulations of the pectoralgirdle to the humerus (Osborn 1933 Johnson and Ostrom 1995Dodson and Farlow 1997) whereas footprints suggest a moreupright stance (Lockley and Hunt 1995) Dodson and Farlow(1997) suggested an inclined ulna and radius along with a par-tially sprawling humerus accounting for the discrepancy Myresults indicate a partially sprawling humerus a result interme-diate between the above studies apart from Dodson and Farlow(1997) consistent with the results obtained from the one-eighth-size model of Triceratops in the Virtual Triceratops Project at theUSNM (Chapman et al 1999) Most likely the humeruschanged positions during the gait cycle with the element moreupright at the ends of the cycle and more sprawling as the leg wasbrought around from back to front during the cycle (Chapman etal 1999)

The neoceratopsian hindlimb exhibits changes primarily asso-ciated with increase in size and weight but the limbs appear tobow out laterally more as size increases an unexpected trend(mechanically limbs would be expected to become more colum-nar with increasing weight as they do in extant large mammals)Characters supporting this observation include the greater in-crease in lateral condyle size versus that of the medial condylelateral bowing of the femur itself and an increase in femoralhead height with size In the tibia the lateral side of the distalend of the element expands at a much greater rate than themedial side perhaps to accommodate the greater amount ofweight placed here due to the lateral bowing of the limb Al-though counterintuitive this trend has been noted previously(Paul 1991) and is most likely due to a relative increase in ab-domen size with overall increase in size A comparable trend isseen in titanosaurid limb posture and is similarly interpreted(Wilson and Carrano 1999)

SUMMARY

Ceratopsian girdle and limb elements increased in robustnessthrough Neoceratopsia and within Psittacosauridae The major-ity of the measurements increased with positive allometry com-pared to length of the elements The ventral scapulocoracoiddistal humerus proximal ulna proximal and distal ends of theradius prepubic process distal femur distal tibia and proximaland distal ends of the fibula reflect this the most Some mea-surements increased with negative allometry while still provid-ing evidence for increasing robustness These include the lengthfrom the rim of the sigmoid fossa of the ulna distally length ofthe postacetabular process of the ilium all height measurementsof the ilium and height of the articular portion of the pubisSome measurements that would be assumed to show allometrictrends change nearly isometrically with increasing length includ-ing maximum height of the scapula deltopectoral crest length ofthe humerus and width of the femoral shaft at 75 length

The plesiomorphic condition of bipedality in Psittacosaurus issupported by the shape comparisons Indicators of bipedalityand possible alternate forelimb use in Psittacosaurus as con-trasted to neoceratopsians include a more mobile radial head

reduced manus digits and modifications of the pelvis that sup-port more weight cranially Forelimb elements of Psittacosaurusare more robust than those of basal neoceratopsians and thistrend together with increased mobility suggests that their fore-limbs were used for manipulation of the environment in someway The shift to quadrupedalism in basal neoceratopsians isverified by the change in orientation of the deltopectoral crest onthe humerus the more compact manus and the shift in the pelvicelements to support more weight centrally

The increase in size from the basal neoceratopsians to cera-topsids is reflected in a change of stance from the more uprightone of the basal neoceratopsians to a more sprawling postureespecially in the forelimbs The deltopectoral crest is orientedmore horizontally in ceratopsids the olecranon process of theulna is larger and the lateral condyle of the femur and lateralside of the distal tibia are more developed

The morphometric methods applied to this material comple-ment and allow for control of each other Each method providescertain information and allows for certain conclusions Observa-tions and results interpreted from a combination of all methodsare more robust closer to being unbiased and are more infor-mative than those formed from any one method

ACKNOWLEDGMENTS

I thank the faculty and students of the Functional Anatomyand Evolution Program at the Johns Hopkins University Schoolof Medicine for their help encouragement and support throughmy graduate school years I also thank the many people whohelped me with this project especially my thesis committeemembers Dave Weishampel Ken Rose Ralph Chapman PeterDodson and Cathy Forster Curators and collections managersat the museums where I collected data are too numerous to listplease accept my apologies and my thanks for all of your helpThis project is part of the doctoral thesis of the author andwould not have been possible without grant money provided bythe Society of Vertebrate Paleontology the Jurassic Foundationthe Geological Society and the Paleontological Society

LITERATURE CITED

Adams D 1987 The bigger they are the harder they fall implications ofischial curvature in ceratopsian dinosaurs pp 1ndash6 in P J Currie andE H Koster (eds) Fourth Symposium on Mesozoic TerrestrialEcosystems Occasional Paper of the Tyrrell Museum of Palaeon-tology 3

Adams D 1991a The significance of sternal position and orientation toreconstruction of ceratopsid stance and appearance Journal of Ver-tebrate Paleontology 11(3 Supplement)14A

Adams D 1991b A scaling equation for humeral diameter in ceratop-sian dinosaurs The Texas Journal of Science 433ndash11

Bakker R T 1987 The return of the dancing dinosaurs pp 39ndash69 inS J Czerkas and E C Olson (eds) Dinosaurs Past and PresentVolume I Natural History Museum of Los Angeles County andUniversity of Washington Press Seattle

Benson R H R E Chapman and A F Siegel 1982 On the measure-ment of morphology and its change Paleobiology 8328ndash339

Bookstein F L 1991 Morphometric Tools for Landmark Data Geom-etry and Biology Cambridge Cambridge University Press 435 pp

Brown B and E M Schlaikjer 1940 The structure and relationship ofProtoceratops Annals of the New York Academy of Sciences 40135ndash266

Chapman R E 1990 Shape analysis in the study of dinosaur morphol-ogy pp 21ndash42 in K Carpenter and P J Currie (eds) DinosaurSystematics - Approaches and Perspectives Cambridge UniversityPress New York

Chapman R E P M Galton J J Sepkoski Jr and W P Wall 1981A morphometric study of the cranium of the cranium of the pachy-cephalosaurid dinosaur Stegoceras Journal of Paleontology 55608ndash618

Chapman R E and M K Brett-Surman 1990 Morphometric observa-tions on hadrosaurid ornithopods pp 163ndash177 in K Carpenter and


P J Currie (eds) Dinosaur Systematics Cambridge UniversityPress Cambridge

Chapman R E D B Weishampel G Hunt and D Rasskin-Gutman1997 On using the shapes of dinosaurs Dinofest International 199731ndash37

Chapman R E A F Andersen and S J Jabo 1999 Construction ofthe virtual Triceratops procedures results and potentials Journalof Vertebrate Paleontology 19(3 Supplement)37A

Chinnery B J 2002 Morphometric analysis of evolution and growthin the ceratopsian postcranial skeleton Unpublished PhD disser-tation Johns Hopkins University School of Medicine Baltimore360 pp

Chinnery B J and R E Chapman 1998 A morphometric study of theceratopsid postcranial skeleton Journal of Vertebrate Paleontology18(3 Supplement)33A

Chinnery B J and D B Weishampel 1998 Montanoceratops cerorhyn-chus (Dinosauria Ceratopsia) and relationships among basal neo-ceratopsians Journal of Vertebrate Paleontology 18569ndash585

Dawson S D 1994 Allometry of cetacean forelimb bones Journal ofMorphology 222215ndash221

Dodson P 1975a Taxonomic implications of relative growth in lambeo-saurine hadrosaurs Systematic Zoology 2437ndash54

Dodson P 1975b Functional and ecological significance of relativegrowth in Alligator Journal of Zoology London 175315ndash355

Dodson P 1976 Quantitative aspects of relative growth and sexual di-morphism in Protoceratops Journal of Paleontology 50929ndash940

Dodson P 1980 Comparative osteology of the american ornithopodsCamptosaurus and Tenontosaurus Memoirs of the Society of Geol-ogy France 13981ndash85

Dodson P 1990 On the status of the ceratopsids Monoclonius andCentrosaurus pp 231ndash244 in K Carpenter and P Currie (eds)Dinosaur Systematics Perspectives and Approaches CambridgeUniversity Press Cambridge

Dodson P 1993 Comparative craniology of the Ceratopsia pp 200ndash234in P Dodson and P Gingerich (eds) Functional Morphology andEvolution American Journal of Science Special Volume 293-A

Dodson P 1996 The Horned Dinosaurs A Natural History New JerseyPrinceton University Press 360 pp

Dodson P and P J Currie 1990 Neoceratopsia pp 393ndash398 in D BWeishampel P Dodson and H Osmolska (eds) The DinosauriaUniversity of California Press Berkeley

Dodson P and J O Farlow 1997 The forelimb carriage of ceratopsiddinosaurs Dinofest International Proceedings 1997 pp 393ndash398

Dodson P C A Forster and S D Sampson In press Ceratopsidae inD B Weishampel P Dodson and H Osmolska (eds) The Dino-sauria University of California Press Berkeley

Dong Z and Y Azuma 1997 On a primitive neoceratopsian from theEarly Cretaceous of China pp 68ndash89 in Z Dong (ed) Sino-Japanese Silk Road Dinosaur Expedition China Ocean Press Bei-jing

Forster C A 1990 The cranial morphology and systematics of Tricera-tops with a preliminary analysis of ceratopsian phylogeny Unpub-lished PhD dissertation University of Pennsylvania Philadelphia227 pp

Forster C A 1996 New information on the skull of Triceratops Journalof Vertebrate Paleontology 16246ndash258

Gilmore C W 1905 The mounted skeleton of Triceratops prorsus Pro-ceedings of the US National Museum 29433ndash435

Gilmore C W 1917 Brachyceratops a ceratopsian dinosaur from theTwo Medicine Formation of Montana with notes on associated rep-tiles United States Geological Survey Professional Paper 1031ndash45

Gray S W 1946 Relative growth in a phylogenetic series and in anontogenetic series of one of its members American Journal of Sci-ence 244792ndash807

Hatcher J B O C Marsh and R S Lull 1907 The Ceratopsia U SGeological Survey Monograph 491-XXIX1ndash300

Hughes N C and R E Chapman 1995 Growth and variation in theSilurian proetide trilobite Aulacopleura konincki and its implica-tions for trilobite palaeobiology Lethaia 28333ndash353

Huxley J S 1932 Problems of Relative Growth New York DoverPublications 302 pp

Johnson R E and J H Ostrom 1995 The forelimb of Torosaurus andan analysis of the posture and gait of ceratopsians pp 205ndash218 in JThomason (ed) Functional Morphology in Vertebrate Paleontol-ogy Cambridge University Press Cambridge

Jolicoeur P and J E Mosimann 1960 Size and shape variation in thepainted turtle A principal components analysis Growth 24339ndash354

Lehman T M 1989 Chasmosaurus mariscalensis n sp a new ceratop-sian dinosaur from Texas Journal of Vertebrate Paleontology 9137ndash162

Lehman T M 1990 The ceratopsian subfamily Chasmosaurinae sexualdimorphism and systematics pp 211ndash230 in K Carpenter and PCurrie (eds) Dinosaur Systematics Perspectives and ApproachesCambridge University Press Cambridge

Lele S 1991 Some comments on coordinate-free and scale-invariantmethods in morphometrics American Journal of Physical Anthro-pology 85407ndash417

Lele S and T M Cole 1996 A new test for shape differences whenvariance-covariance matrices are unequal Journal of Human Evo-lution 31193ndash212

Lele S and J T Richtsmeier 1991 Euclidean Distance Matrix Analy-sis A coordinate-free approach for comparing biological shapes us-ing landmark data American Journal of Physical Anthropology 86415ndash427

Lockley M G and A P Hunt 1995 Ceratopsid tracks and associatedichnofauna from the Laramie Formation (Upper Cretaceous Maas-trichtian) of Colorado Journal of Vertebrate Paleontology 15592ndash614

Lull R S 1933 A revision of the Ceratopsia or horned dinosaurs Pea-body Museum of Natural History Memoirs 31ndash175

Lull R S and S W Gray 1949 Growth patterns in the CeratopsiaAmerican Journal of Science 247492ndash503

Makovicky P J 2001 A Montanoceratops cerorhynchus (DinosauriaCeratopsia) braincase from the Horseshoe Canyon Formation ofAlberta pp 243ndash262 in D H Tanke and K Carpenter (eds) Me-sozoic Vertebrate Life Indiana University Press Bloomington

Makovicky P J 2002 Taxonomic revisions and phylogenetic relation-ships of basal Neoceratopsia (Dinosauria Ornithischia) Unpub-lished PhD dissertation Columbia University 288 pp

Maryanska T 1990 Pachycephalosauria pp 564ndash577 in D B Weisham-pel P Dodson and H Osmolska (eds) The Dinosauria Universityof California Press Berkeley

Maryanska T and H Osmolska 1974 Pachycephalosauria a new sub-order of ornithischian dinosaurs Palaeontologica Polonica 3045ndash102

Maryanska T and H Osmolska 1975 Protoceratopsidae (Dinosauria)of Asia Palaeontologica Polonica 33133ndash182

Osborn H F 1923 Two Lower Cretaceous dinosaurs of MongoliaAmerican Museum Novitates 951ndash3

Osborn H F 1933 Mounted skeleton of Triceratops elatus AmericanMuseum Novitates 6541ndash14

Ostrom J H and P Wellnhofer 1986 The Munich specimen of Tric-eratops with a revision of the genus Zitteliana 14111ndash158

Oxnard C E 1978 Primate quadrupedalism some subtle structural cor-relates Yearbook of Physical Anthropology 20538ndash554

Oxnard C E 1983 Anatomical biomolecular and morphometric viewsof the primates Progress in Anatomy 3113ndash142

Oxnard C E 1984 The place of Tarsius as revealed by multivariatestatistical morphometrics pp 17ndash32 in C Niemitz (ed) Biology ofTarsiers Fischer Verlag Tubingen

Paul G S 1991 Giant horned dinosaurs did have fully erect forelimbsJournal of Vertebrate Paleontology 11(3 Supplement)50A

Reyment R A R E Blackith and N A Campbell 1984 MultivariateMorphometrics 2nd edition Academic Press London 233 pp

Richtsmeier J T J M Cheverud and S Lele 1992 Advances in an-thropological morphometrics Annual Review of Anthropology 21283ndash305

Rohlf F J 1998 TPS Dig computer program Stony Brook New YorkRohlf F J and D Slice 1990 Extensions of the Procrustes method for

the optimal superimposition of landmarks Systematic Zoology 3940ndash59

Romer A S 1956 Osteology of the Reptiles Chicago University ofChicago Press 772 pp

Russell D A 1970 A skeletal reconstruction of Leptoceratops gracilisfrom the upper Edmonton Formation (Cretaceous) of Alberta Ca-nadian Journal of Earth Sciences 7181ndash184

Sampson S D M J Ryan and D H Tanke 1997 Craniofacial ontog-eny in centrosaurine dinosaurs (OrnithischiaCeratopsidae) taxo-nomic and behavioral implications Zoological Journal of LinneanSociety 121293ndash337


Sereno P C 1984 The phylogeny of the Ornithischia a reappraisal pp219ndash226 in W Reif and F Westfal (eds) Third Symposium onMesozoic Terrestrial Ecosystems Short Papers Attempto VerlagTubingen University Press Tubingen

Sereno P C 1986 Phylogeny of the bird-hipped dinosaurs (Order Or-nithischia) National Geographic Research 2234ndash256

Sereno P C 1990 Psittacosauridae pp 579ndash592 in D B Weishampel PDodson and H Osmolska (eds) The Dinosauria University ofCalifornia Press Berkeley

Sereno P C 2000 The fossil record systematics and evolution of pacy-cephalosaurs and ceratopsians from Asia pp 480ndash516 in M J Ben-ton M A Shishkin D M Unwin and E N Kurochkin (eds) TheAge of Dinosaurs in Russia and Mongolia Cambridge UniversityPress Cambridge

Siegel A F and R H Benson 1982 A robust comparison of biologicalshapes Biometrics 38341ndash350

Sternberg C M 1927 Horned dinosaur group in the National Museumof Canada Canadian Field-Naturalist 4167ndash73

Tereshchenko V S 1991 The reconstruction of the vertebral column ofProtoceratops Paleontological Journal 286ndash96 (English 106ndash116)

Tereshchenko V S 1994 A reconstruction of the erect posture of Pro-toceratops Paleontological Journal 28104ndash119

Tereshchenko V S 1996 A reconstruction of the locomotion of Proto-ceratops Paleontological Journal 30232ndash245

Thompson D W 1917 On Growth and Form Cambridge CambridgeUniversity Press 1116 pp

Van Valkenburgh B 1987 Skeletal indicators of locomotor behavior inliving and extinct carnivores Journal of Vertebrate Paleontology7162ndash182

Watabe M and H Nakaya 1991 Phylogenetic significance of the post-cranial skeletons of the hipparions from Maragheh (Late Miocene)Northwest Iran Memoirs of the Faculty of Science Kyoto Univer-sity Series of Geology and Minerology LVI11ndash53

Weishampel D B and R E Chapman 1990 Morphometric study ofPlateosaurus from Trossingen (Baden-Wurttemberg Federal Re-public of Germany) pp 43ndash51 in K Carpenter and P J Currie(eds) Dinosaur Systematics Perspectives and Approaches Cam-bridge University Press Cambridge

Wilson J A and M T Carrano 1999 Titanosaurs and the origin ofldquowide-gaugerdquo trackways a biomechanical and systematic perspec-tive on sauropod locomotion Paleobiology 25252ndash267

Zhao X C Zhengwu and X Xing 1999 The earliest ceratopsian fromthe Tuchengzi Formation of Liaoning China Journal of VertebratePaleontology 19681ndash691

Received 2 December 2002 accepted 6 October 2003

APPENDIX

Supplemental data available from SVP website httpwwwvertpaleoorgjvpJVPcontentshtml


MORPHOMETRIC ANALYSIS OF EVOLUTIONARY TRENDS IN THE CERATOPSIANPOSTCRANIAL SKELETON

BRENDA CHINNERYMuseum of the Rockies Montana State University 600 W Kagy Blvd Bozeman MT 59717 bchinnerymontanaedu

ABSTRACTmdashThe cranial anatomy of ceratopsian (ldquohornedrdquo) dinosaurs is well understood but the postcranial skeletonhas been largely ignored in previous studies of phylogeny evolution and function In order to study morphologicaldifferences among ceratopsian postcranial elements and to compare evolutionary patterns multivariate (Principal Com-ponents Analysis) and bivariate methods were used to analyze linear measurement data and the shape methods Resis-tant-Fit Theta-Rho Analysis (RFTRA) Least-Squares Theta-Rho Analysis (LSTRA) and Euclidean Distance MatrixAnalysis (EDMA) were applied to biological landmark data Results of the analyses show that size is the primary changethrough ceratopsian skeleton evolution Elements display positive allometry and increasing structural support is evidentespecially in the radius and fibula

Phylogenetic distribution of ceratopsian postcrania agrees with the skull material included in recent cladistic analysesPsittacosaurus elements are in many ways derived relative to those of non-ceratopsid neoceratopsians and evidencesuggests that Psittacosaurus was bipedal while non-ceratopsid neoceratopsians were not except maybe for Udanocera-tops With increasing body size neoceratopsian limbs bowed laterally

The variety of methods allows for unbiased interpretation of results provides more information than any one methodand provides controls for each other However sample sizes are not ideal and all results should be treated with cautionat this time

INTRODUCTION

Ceratopsia are a clade of ornithischian dinosaurs comprised ofthe family Psittacosauridae and the infraorder Neoceratopsia(Sereno 1984 1986 1990 Dodson and Currie 1990) Althoughthe oldest ceratopsians known may have lived as early as theEarly Cretaceous (Chaoyangsaurus Zhao et al 1999) the cladeis mainly known from the radiations that occurred during themiddle and Late Cretaceous of Asia and the Late Cretaceous ofthe western United States and Canada

Ceratopsian postcranial anatomy varies throughout the cladePsittacosauridae a clade of small herbivores known only fromthe Early Cretaceous of Asia share the diagnostic ceratopsianrostral bone with the other members of Ceratopsia Psittacosaursare characterized by long hind limb elements relatively shortforelimb elements and a reduced manus (Sereno 1990) suggest-ing that members of this clade were facultatively bipedal Psit-tacosaurids are relatively small dinosaurs not more than twometers in length

Neoceratopsia (Ceratopsia excluding PsittacosauridaeSereno 1984 1986) comprise the majority of Ceratopsia Thesedinosaurs first appeared in the Early Cretaceous of Asia as smallpossibly bipedal herbivores and over time evolved into large-bodied graviportal quadrupeds Basal neoceratopsians vary inmorphology continental distribution and fossil age The groupis paraphyletic (Sereno 1986 2000 Chinnery and Weishampel1998) consisting of taxa more closely related to the ceratopsidsof North America Ceratopsidae are a monophyletic clade com-prised of two subfamilies Centrosaurinae and ChasmosaurinaePostcranially these subfamilies have been considered extremelysimilar (Dodson and Currie 1990) All are robust quadrupedsand the functional changes associated with this graviportal life-style were thought to be indistinct at the subfamilial level andeven less distinct among genera within the subfamilies

The present study of ceratopsian postcranial anatomy involvesthe application of morphometric analyses treated in a phyloge-netic context This project was conceived with the realization of

how essential it is that we understand the evolution and growthof the ceratopsian postcranial skeleton because it comprises themajority of the skeletal mass is responsible for stature and lo-comotion (among other functions) and evolved through timealong with the cranial skeleton The postcranial skeleton hasbeen neglected in the past because phylogenetic analyses haverelied primarily on cranial characters (Sereno 1984 1986 2000Chinnery and Weishampel 1998 Makovicky 2002 Dodson etal in press) The postcranial skeleton is not only important forphylogenetic analysis but also for any discussions of functionbehavior or growth within the clade

Linear measurement and landmark data are available in theSupplemental Materials (see Appendix) online at httpwwwvertpaleoorgjvpJVPcontentshtml as are additional re-sults and text not able to be included here Girdle and limbelements of all available ceratopsian taxa were compared at sub-famial and generic levels Only the former is included in thispaper because comparisons of taxa at the generic level are sim-ply too long to include However they are available in theSupplemental Materials Additional objectives of the projectwere to document and describe taxonomic differences in theceratopsian postcranial skeleton and to compare growth changesto evolutionary changes in postcranial skeletal anatomy Theapplication of a variety of morphometric techniques ensuresrelatively non-biased results and provides quantified and testableresults as well as data for inclusion in future analyses

Morphometry consists of techniques used to capture the shapeand size of an object (usually but not always biological) whichcan then be compared with another object The techniques haveevolved through time from the deformation grids of DrsquoArcyThompson (1917) to the myriad methods used today (eg Chap-man 1990 and references therein Richtsmeier et al 1992 Chin-nery and Chapman 1998) As an organism develops stressesincluding gravity movement posture and growth cause changesin shapes and sizes of skeletal (and soft tissue) elements Suchchanges are often subtle and may be missed without the use ofmorphometrics Quantitative approaches have been used to cap-ture and clarify evolutionary changes in both living and extinctorganisms as well as to provide the means to study sexual di-morphism posture and locomotor changes ontogenetically and

Present address 1620 Green Meadow Lane Spring Branch Texas78070

Journal of Vertebrate Paleontology 24(3)591ndash609 September 2004copy 2004 by the Society of Vertebrate Paleontology

591





MATERIALS










METHODS



















































































































Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX







MATERIALS










METHODS



















































































































Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX






METHODS



















































































































Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX













































































































Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX











































































































Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX

































































































Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX























































































Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX









































































Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX







































































Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX


























































Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX
























































Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX








































Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX






































Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX




































Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX



























Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX









Iternode 1 2 3 4




























Short femoral head











SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX





SUMMARY






ACKNOWLEDGMENTS


LITERATURE CITED




















































































APPENDIX











































































APPENDIX



















APPENDIX



Documents

Morphometric analysis of evolutionary trends in the ceratopsian postcranial skeleton