lable at ScienceDirect

Cretaceous Research xxx (2010) 1e15

Contents lists avai

Cretaceous Research

journal homepage: www.elsevier .com/locate/CretRes

Redescription of the skull of Saurolophus osborni Brown 1912 (Ornithischia:Hadrosauridae)

Phil R. Bell*

Department of Biological Sciences, University of Alberta, Edmonton, Alberta T6G 2E9, Canada

a r t i c l e i n f o

Article history:Received 4 September 2009Accepted in revised form 1 October 2010Available online xxx

Keywords:AlbertaCanadaCretaceousHadrosauridaeHorseshoe Canyon FormationMaastrichtianSaurolophus osborniTaxonomy

* Fax: þ1 780 492 9234.E-mail address: pbell@ualberta.ca.

0195-6671/$ e see front matter � 2010 Elsevier Ltd.doi:10.1016/j.cretres.2010.10.002

Please cite this article in press as: Bell, P.RCretaceous Research (2010), doi:10.1016/j.cr

a b s t r a c t

A detailed redescription of the skull of the hadrosaurine Saurolophus osborni based on the three knownspecimens of this taxon, elucidates for the first time several important aspects of the cranial morphology.Details of the braincase, palate, and nasal crest that were either missed or disputed by previous authorsare clarified. The nasal crest is buttressed caudally and caudolaterally by elongations of the frontals andprefrontals, respectively. The unique morphology of the frontals and prefrontals permit differentiation ofS. osborni from other hadrosaurs. S. osborni can be differentiated from Saurolophus angustirostris bya relatively shorter frontal contribution to the nasal crest, a relatively straight premaxilla in lateral view,a weakly upturned oral margin of the premaxilla, and the absence of a vertical sulcus on the prooticabove cranial nerve V. A reassessment of the phylogenetic affinities of S. osborni, exclusive of S. angus-tirostris, suggests it is most closely related to the Laurasian taxa Prosaurolophus and Kerberosaurus.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

The Hadrosauridae is divisible into hollow-crested (Lambeo-saurinae) and the solid-crested or flat-headed (Hadrosaurinae)forms. Although typically non-crested, several hadrosaurines haveincipient and/or solid crests formed by outgrowths of the nasalsand in some forms, the frontals. Saurolophus osborni is a large (up to11 m), solid-crested hadrosaurine from the upper HorseshoeCanyon Formation of Alberta, Canada. The type specimen (AMNH5220), a virtually complete skull and skeleton discovered in 1911,was described by Brown (1912, 1913). The specimen was preparedas a panel mount and mounted behind glass with its right side (theside on which the animal was found lying) exposed. A secondskeleton (paratype, AMNH 5221), found in the same year, waslargely eroded and left in situ; however, the well-preserved anddisarticulated skull was collected, the braincase of which wasfigured by Brown (1912) but remains largely undescribed. A thirdskull (CMN 8796, Fig. 1) was found in 1925 by C.M. Sternberg in thesame formation and figured in Russell and Chamney (1967). Thisskull, despite good preservation, was heavily reconstructed duringrestoration rendering anatomical description difficult. Additionalmaterial from the Moreno Formation in California identified as

All rights reserved.

., Redescription of the skulletres.2010.10.002

cf. Saurolophus (Morris, 1973) is both poorly preserved andincomplete. Reanalysis of this material suggests similarities to bothSaurolophus and Edmontosaurus; however, it is best referred to asHadrosaurinae indet (Bell and Evans, 2010). Similarly, reports ofSaurolophus from the Almond Formation of Wyoming (Gates andFarke, 2009) should be considered tentatively in the absence ofmore complete material.

Saurolophus is palaeobiogeographically important as it is theonly Late Cretaceous dinosaur genus to co-occur in North Americaand Asia. Despite a superficial resemblance (at least at the familiallevel) between the Campanian and Maastrichtian faunas of theseregions (Jerzykiewicz and Russell, 1991), the shared presence ofSaurolophus between the Canada and Mongolia is unique. Thisrelationship, however, has been disputed (Norman and Sues, 2000).The only other species in the genus, Saurolophus angustirostrisRozhdestvenskii 1952 from the Nemegt Formation in centralMongolia, closely resembles S. osborni but osteological descriptionsof S. angustirostris (Rozhdestvenskii, 1952, 1957; Marya�nska andOsmólska, 1981, 1984) are limited in their comparisons of the twospecies of Saurolophus. Considerable confusion has since arisenregarding the exact construction of the crest and the relationshipbetween the two species (Ostrom,1961; Horner, 1992; Norman andSues, 2000; Lund and Gates, 2006).

The aim of this study is to 1) provide a comprehensive rede-scription of the skull of S. osborni; 2) clarify important aspects ofthe cranial architecture, in particular the nasal crest; 3) emend

of Saurolophus osborni Brown 1912 (Ornithischia: Hadrosauridae),

Fig. 1. Skull of Saurolophus osborni (CMN 8796) in right lateral view. Photo (above) andinterpretive drawing (below). Grey denotes areas reconstructed and/or obscured bypaint/plaster. Skull length ¼ 945 mm.

Fig. 2. Stratigraphic column of latest Cretaceous deposits in the Central Plains ofAlberta (adapted from Eberth 2004). Large arrow indicates the approximate strati-graphic position of Saurolophus osborni. DMT, Drumheller marine tongue.

P.R. Bell / Cretaceous Research xxx (2010) 1e152

the diagnosis of the type species, and 4) reassess the phylogeneticposition of this species. A more detailed description of S. angustir-ostriswill be necessary to fully resolve the relationship between thetwo species.

2. Materials and methods

This redescription is based on the three known specimens ofS. osborni from Alberta. The paratype (AMNH 5221) and referredspecimen CMN 8796 were central to this study, and supplementedby the holotype. Measurements of the holotype obtained by Brown(1912, 1913) were used to scale digital photos of the specimen inImage J for additional measurements. All observations of theholotype are from the right side except where noted.

Institutional abbreviations. AMNH, American Museum of NaturalHistory, New York; CMN, Canadian Museum of Nature, Ottawa,Ontario; LACM/CIT, Natural History Museum of Los Angeles County(formerly at California Institute of Technology), Los Angeles, Cal-ifornia; MPC, Mongolian Palaeontological Centre, Ulaan Baatar,Mongolia; ZPAL, Institute of Palaeobiology of the Polish Academy ofSciences, Warsaw, Poland.

3. Systematic palaeontology

Ornithopoda Marsh, 1881Iguanodontia Dollo, 1888Hadrosauridae Cope, 1870Hadrosaurinae Cope, 1870

Please cite this article in press as: Bell, P.R., Redescription of the skullCretaceous Research (2010), doi:10.1016/j.cretres.2010.10.002

Saurolophus Brown, 1912Saurolophus osborni Brown, 1912, pl. 1, 2, Figs. 1e4.Holotype. AMNH 5220, almost complete skull and skeletonlacking the most distal caudal vertebrae and the distal endsof both ischia.Paratype. AMNH 5221, nearly complete, disarticulated skullcomprising left and right premaxillae, maxillae, jugals, qua-drates, right quadratojugal, most of the palate, left and rightdentaries, and a nearly complete braincase.Referred Specimen. CMN 8796, partial skull.Horizon and Locality. Only the type and paratype localities ofS. osborni have been so far relocated. Both were collectedfrom the Tolman Bridge area along the banks of the Red DeerRiver in southern Alberta, Canada (Fig. 2). Both specimenscome from unit 4 of the Horseshoe Canyon Formation (sensuEberth, 2004). Eberth and Deino (2005) recently provided40Ar/39Ar dates for a bentonite layer that directly overliescoal seam #10 in Unit 2. Their date of 70.44 � 0.17 Ma iscorrelated with the CampanianeMaastrichtian boundary(Eberth and Deino, 2005). Therefore, S. osborni is of earlyMaastrichtian age (magnetochron 31re31n, Eberth, 2004).Eberth (2004) describes Unit 4 as a non-coaly facies made upof subequal proportions of paleochannel sandstones andoverbank mudstones that represent the driest interval ofHorseshoe Canyon Formation deposition.Emended Diagnosis. Large hadrosaurine (up to 11 m long)with the following combination of characteristics: solid,caudodorsally-directed cranial crest composed of the nasals,prefrontals, and frontals that extends caudal to the squa-mosals in adults; frontals tripartite, each contributinga finger-like caudodorsal ramus that buttresses the under-side of the nasal crest; caudodorsal ramus of frontalapproximately one-fifth of the length of the entire crest inadults; frontal excluded from orbital rim by post-orbitaleprefrontal complex; prefrontal and supraorbital Ipresent and unfused in late ontogeny; rostral process of jugal

of Saurolophus osborni Brown 1912 (Ornithischia: Hadrosauridae),

Fig. 3. Right premaxilla of AMNH 5221 in A. lateral and, B. medial views. Scale ¼ 10 cm. Dashed lines denote approximate outline of missing areas. Shading identifies area obscuredby plaster in the original specimen.

P.R. Bell / Cretaceous Research xxx (2010) 1e15 3

asymmetrical in lateral view; dorsal margin of premaxillastraight or nearly so.

3.1. Description

3.1.1. Cranium3.1.1.1. Premaxilla. The paired premaxillae are elongate U-shapedelements that are in union for their entire length along the midlineto form a strongly vaulted muzzle. Each premaxilla is composed ofa curved body that forms the typical hadrosaur ‘bill’ and a dorsaland a lateral process that form the margins of the slit-like externalnarial foramen (Fig. 3). The dorsal margin of the element is straightfor its entire length in the paratype. It is slightly upturned rostral to

Fig. 4. Left maxilla of AMNH 5221 in A. late

Please cite this article in press as: Bell, P.R., Redescription of the skullCretaceous Research (2010), doi:10.1016/j.cretres.2010.10.002

the naris in the type, although it does not approach the exaggeratedcondition in S. angustirostris. The oral margin is upturned to forma ‘lip’ extending from the rostral tip of the premaxilla to approxi-mately the caudal margin of the naris. This is similar to thecondition of Prosaurolophus but is not as developed as in Edmon-tosaurus. In AMNH 5221, this ‘lip’ is diagenetically exaggerated;in life, these edges were probably not inclined far beyond hori-zontal. Rostral to the external narial foramen is an accessory fossawithin the circumnarial fossa. This accessory fossa is elongate orfinger-like, extending rostrally through approximately 30% of thepremaxillary body length. Its full length is visible only in dis-articulated specimens (AMNH 5221). The dorsal process is rhom-boidal in cross-section at its midpoint and tapers to a slightlyexpanded, spatulate end that is triangular in cross-section. In

ral and, B. medial views. Scale ¼ 10 cm.

of Saurolophus osborni Brown 1912 (Ornithischia: Hadrosauridae),

P.R. Bell / Cretaceous Research xxx (2010) 1e154

AMNH 5220 and CMN 8796, the dorsal processes are united,becoming obscured by the nasals at about the caudal margin of thenaris. In cross-section, the lateral process twists caudally to becomevirtually horizontal where it overlies a lateral extension of thenasals before terminating close to the dorsal tip of the lacrimal.

3.1.1.2. Maxilla. The maxilla (Fig. 4A,B) is typically Hadrosaurinaein design. The rostral process in triangular in lateral view. Dorsally,the rostral one-third of the length of the rostral process is smoothand flatly contacts the premaxilla; the remaining two-thirds rostralto the dorsal process is deeply grooved to receive the lacrimal. InAMNH 5221, the base of the rostrodorsal process (median rostralprocess of Weishampel and Horner, 1990) is preserved dorsal to therostral tip of the maxilla (rostroventral process of Horner, 1992).The outline of the broken surface of the rostrodorsal processsuggests that it was thin and delicate. It is not visible through theexternal narial foramen in AMNH 5220 as it is in Edmontosaurus,Maiasaura, Brachylophosaurus, or S. angustirostris but this maysimply be due to incomplete preparation of the type specimen. Theoral margin is smoothly concave and supported approximately 46alveoli in AMNH 5221. The lateral face of the maxilla is dominatedby a broad, arcuate facet for the jugal. This contact is laterally offsetfrom the tooth row and is inclined dorsomedially where it termi-nates as the triangular dorsal process, which lies at the midpoint ofthe element. Ventrally, the jugal contact is bounded by a prominentlateral shelf that overhangs one or two large foramina. This shelfextends parallel to the tooth row along the dorsal half of the caudalprocess where it eventually forms the caudal margin of the maxilla.Up to five foramina may be present on the lateral surface of themaxilla, and the orientation and sizes of these are variable evenbetween the left and right sides. The dorsal process is medi-olaterally narrow and the caudal edge is nearly vertical. The dorsalprocess contacts the lacrimal rostrodorsally. In AMNH 5221, thetabular caudal process contributes 42% of the length of the maxilla.The base of the dorsocaudal process is preserved in AMNH 5221rostral to the dorsocaudal corner of the caudal process. The dor-socaudal process appears to have been short and triangular, and didnot extend past the caudal margin of the maxilla. Medially, themaxilla is penetrated by a series of small foramina within a longi-tudinal groove that forms a dorsally convex arc (Fig. 4B). This arcterminates rostrally just caudal to the base of the rostrodorsalprocess. A shorter, shallower and almost straight groove alsopersists between the alveolar margin and these foramina nearlyparalleling the occlusal margin of the bone.

Fig. 5. Naso-frontal contact in Saurolophus (CMN 8796). A, photograph of the underside ofoutline of missing areas. Shading identifies area obscured by plaster in the original specimcaudodorsal process of frontal; CdPfr, caudodorsal process of prefrontal; Fr, frontal; Na, nas

Please cite this article in press as: Bell, P.R., Redescription of the skullCretaceous Research (2010), doi:10.1016/j.cretres.2010.10.002

3.1.1.3. Nasal. In none of the specimens are the nasals complete.However, the nasal is still the longest bone in the skull, forming thegreatest extent of the facial angle. It is not invaded by the nasaldiverticula and is nearly straight in lateral aspect. Left and rightnasals are in union for most of their lengths but are separatedrostrally by the dorsal rami of the premaxillae for the entire lengthof the external narial fenestra. The supranarial process forms thecomplete dorsal margin of the naris and terminates abruptly partlyobscuring a deep fossa on the premaxilla rostral to the narialopening. Caudal to the external narial fenestra, the nasals becomestrongly vaulted and broad between the orbits where they arebounded by the premaxilla laterally, the prefrontals caudolaterally,and by the frontals caudally. From this contact, the nasals continuecaudodorsally over the skull roof to form a broad crest that, inAMNH 5220, extends above the squamosals where the crest isbroken. The crest curves slightly dorsally relative to the facial angleand is triangular in cross-section. Based on the curvature in AMNH5220, the crest may have been almost vertical at its terminus. InAMNH 5220, which preserves more of the crest than any otherspecimen, the bone surface has a series of longitudinal canals thatprobably housed blood vessels in life. These differ from the singlelongitudinal septum on the nasal of S. angustirostris (Marya�nskaand Osmólska, 1981). Their presence cannot be confirmed on theleft side of AMNH 5220 or in any other specimen. Although not asdense as in mammalian or ceratopsian horn cores, blood vesselgrooves on the crest of S. osbornimay indicate the possible presenceof a keratinous sheath (Hieronymus et al., 2009). As the tip of thecrest is not preserved, the presence of a bony ‘cap’ at its terminationas in S. angustirostris, cannot be confirmed. In cross-section, eachnasal has an inverted U-shaped groove, which extends for anunknown length along the underside of the crest. The proximalportion of these grooves is occupied by a dorsocaudal process fromeach frontal (Fig. 5). The grooves are twice the width of the frontalprocesses and consequently only the lateral half of each groove isoccupied. The nasal crest is similarly buttressed laterocaudallyby elongate processes of the prefrontals. The broken crest inCMN 8796 shows the nasal diverticula did not invade the nasals assuspected by Ostrom (1961).

3.1.1.4. Lacrimal. The lacrimals are preserved in articulation inAMNH 5220 and CMN 8796. The lacrimal forms a scalene triangle,the shortest edge being the lacrimal contribution to the orbitalmargin. The rostral apex lies between the premaxilla andmaxilla atabout the caudal border of the naris. The premaxilla overlies the

the crest looking rostrally. B. Interpretive drawing. Dashed lines denote approximateen. Parallel lines represent broken surfaces. Cross-hatching represents matrix. CdFr,al; Pfr, prefrontal; Po, postorbital.

of Saurolophus osborni Brown 1912 (Ornithischia: Hadrosauridae),

Fig. 6. Right jugal of AMNH 5221 in A. lateral and, B. medial views. Scale ¼ 5 cm.Parallel lines denote broken surface.

Fig. 7. Left quadratojugal of AMNH 5221 in A. lateral and, B. medial views.Scale ¼ 2 cm.

P.R. Bell / Cretaceous Research xxx (2010) 1e15 5

entire length of the rostrodorsal edge of the lacrimal therebyseparating it from contact with the nasal at least externally. Atits dorsal limit, a short spur from the prefrontal overlaps thelacrimal in AMNH 5220. Based on the disarticulated maxillae(AMNH 5221), the lacrimal and maxilla form an elongate bridlejoint. When the jugal is attached, the caudal half of this contact isobscured; the jugal also meets the lacrimal caudoventrally andlaterally. The lacrimal foramen could not be observed.

3.1.1.5. Prefrontal. The prefrontals of S. osborni are radicallydifferent from all other hadrosaurs except S. angustirostris. InAMNH 5220 and CMN 8796, the prefrontals are preserved incontact with the lacrimal rostrally, nasal rostromedially, frontalcaudomedially and postorbital caudally. In lateral view, theprefrontal forms the rostrodorsal sector of the orbital rim butrather than having a flattened or partly flared lateral edge as inEdmontosaurus, Prosaurolophus, and Kerberosaurus, the prefrontalsare upturned above the orbit so that they partially obscure thenasals in lateral aspect. Caudally, the prefrontal meets the post-orbital to exclude the frontal from the orbital rim (contra Brown,1912) and rostrally, the element tapers to a thin spur of bone thatextends over part of the lateral surface of the lacrimal. A suturevisible on the underside of this element within the orbit dividesthis bone into rostral and caudal halves. Originally (and errone-ously) labelled the frontal and prefrontal by Brown (1912), thesecorrespond to prefrontal and supraorbital 1, respectively, asdescribed for S. angustirostris (Marya�nska and Osmólska, 1981),but unlike that taxon, the suture is perpendicular to the orbitalmargin rather than oblique to it. Brown (1912, p. 135) describesan elongate process of the prefrontal that forms the flank of thecrest, and which apparently fuses with the nasal distally. Thisprocess, not present in other hadrosaurines except S. angustirostris,is difficult to see in the type specimen; however, the base of thisprocess is clearly visible on CMN 8796 (Fig. 5). It appears as analmost cylindrical process that is somewhat offset medially from

Please cite this article in press as: Bell, P.R., Redescription of the skullCretaceous Research (2010), doi:10.1016/j.cretres.2010.10.002

the remainder of that element. It is surrounded in part at its baseby the frontal caudally and medially by the nasal. Althoughincomplete, it was at least as broad as the caudodorsal process ofthe frontal and would have buttressed the underside of the nasalcrest, as suggested by Brown (1912), for an indeterminate length.

3.1.1.6. Jugal. The jugal is roughly W-shaped in lateral view form-ing the ventral borders of the orbit and infratemporal fenestra(Fig. 6A,B). The ventral margin of the W is almost symmetrical. Therostral process is taller than it is long and asymmetrical aboutthe longitudinal axis as in Gryposaurus, Edmontosaurus, Kerber-osaurus, and Prosaurolophus. Rostrally it tapers into a short, narrowspur that is ‘pinched’ between the maxilla and the lacrimal. Themedial surface of the rostral process contacts the maxillary-lacrimal suture, and is delineated caudally by a prominent dorso-ventral ridge. The sutural surface is irregular and undulatory(Fig. 6B), which is mirrored by the maxilla in the extensive contactbetween the two elements. Contact with the palatine would havebeen along the dorsal part of the dorsoventral ridge. The postorbitalprocess is long and narrow, and inclined at 75� to the ventral edgeof the element. The proximal half is mediolaterally compressed;distally it is rostrocaudally compressed where it meets the rostralsurface of the jugal process of the postorbital for the caudal 30% ofits length. The roughly quadrangular caudal process forms a simplelapped joint with the lateral surface of the quadratojugal. Thedorsal corner of the caudal process is drawn out to contact thequadrate, thereby excluding the quadratojugal from the margin ofthe infratemporal fenestra. It lacks the strong jugal flange seen inGryposaurus and Brachylophosaurus (and to a lesser extent,Edmontosaurus) but is smooth and rounded as in Prosaurolophus.Medially, the caudal process is slightly depressed and smoothwhere it contacts the quadratojugal. The caudal edge is almostvertical when in contact with the rest of the skull; the dorsal edge isinclined caudally.

3.1.1.7. Quadratojugal. The quadratojugal is a flat, plate-likeelement that incompletely separates the jugal and quadrate(Fig. 7A,B), but is excluded from the infratemporal fenestra bythese two bones. The caudal edge is C-shaped in lateral view andis relatively robust. In planar cross-section the quadratojugal islenticular. It is met rostrally by the caudal process of the jugal,which overlaps the rather featureless rostrolateral surface of thequadratojugal. The quadrate is received caudomedially within

of Saurolophus osborni Brown 1912 (Ornithischia: Hadrosauridae),

P.R. Bell / Cretaceous Research xxx (2010) 1e156

a sinuous depression of the quadrate that covers the caudal quarter.The caudoventral corner extends to a point.

3.1.1.8. Postorbital. The postorbital is a Y-shaped element thatdefines the dorsal and caudodorsal extent of the orbit, the laterallimit of the supratemporal fenestra, and the rostrodorsal quarter ofthe infratemporal fenestra. Caudally, the cylindrical squamosalprocess strongly interdigitates with the squamosal. The entiremedial and ventral surfaces of the squamosal process are occupiedby an overlapping extension of the squamosal. The jugal processis the longest of the three processes, extending from about themidlength of the postorbital, tapering ventrally and curving slightlyrostrally. The distal half of this process is coarsely striated androstrocaudally compressed distally, forming a simple lapped jointrostrally with the postorbital process of the jugal. The rostralprocess is deflected dorsally, forming an angle of 117� with thesquamosal process in AMNH 5220, and meets with the postorbitalrostrally, frontal medially, and parietal caudomedially. Togetherwith the prefrontal and supraorbital, the rostral process of thepostorbital forms the arched dorsal margin of the orbit. Vieweddorsally, the rostral process is mediolaterally widest where it formsthe rostral margin of the supratemporal fenestra and tapers toa point just caudal to the lateral extent of the nasal crest.

3.1.1.9. Squamosal. The squamosal is a complex element that formsthe caudolateral corner of the skull. Rostrally, the postorbitalprocess of the squamosal has three triangular extensions that

Fig. 8. Left quadrate of holotype 5220 in A. la

Please cite this article in press as: Bell, P.R., Redescription of the skullCretaceous Research (2010), doi:10.1016/j.cretres.2010.10.002

interdigitate with the postorbital. The largest of these extendsalong the medial surface of the postorbital to the base of the jugalprocess of the postorbital. Between the pre- and post-quadraticprocesses lies the deep squamosal cotylus, which is hemisphericaland houses the dorsal head of the quadrate. The finger-like pre-quadratic process extends ventrally from the base of the postorbitalprocess to form a portion of the caudodorsal margin of the infra-temporal fenestra. It is flattened and striated caudally where itcontacts the quadrate. The precotyloid fossa is absent as inEdmontosaurus. The postquadratic process is mediolaterallycompressed and extends caudoventrally along the rostral face ofthe paroccipital process of the exoccipital-opisthotic. Proximally,the paroccipital process rests within a concavity on the post-quadratic process of the squamosal, which becomes shallowerdistally. The medially projecting parietal process is missing in boththe paratype and CMN 8796, and cannot be seen in AMNH 5220.Therefore it cannot be determined if the squamosals meet at themidline or are separated by the parietals.

3.1.1.10. Quadrate. The quadrate is rod-like in lateral view, althoughAMNH 5220 (Fig. 8A,B) is more recurved than in other specimens.The mediolaterally compressed head of the quadrate inserts intoa deep cotyle in the squamosal and is buttressed rostrally by theprequadratic process of the squamosal, which extends ventrallyalong the leading edge of the quadrate. The pterygoid flange origi-nates on the medial surface of the quadrate head and diverts ros-tromedially, which gives the entire element a tick mark-shaped

teral and, B. medial views. Scale ¼ 5 cm.

of Saurolophus osborni Brown 1912 (Ornithischia: Hadrosauridae),

Fig. 9. Braincase of AMNH 5221 in A. right lateral, B. left lateral, C. caudal, D. rostral, and E. dorsal views. Scale ¼ 5 cm.

P.R. Bell / Cretaceous Research xxx (2010) 1e15 7

outline in planar cross-section. This flange is roughly triangularand occupies approximately 55% of the entire height of the quadratein AMNH 5221. It curves rostromedially from a vertical sulcus alongthemedial body of the quadrate, and tapers to a thin edge. In AMNH5220, its dorsal edge is nearly straight whereas in AMNH 5221 it issmoothly concave. Its ventral margin is deeply concave in AMNH5220 but forms an obtuse angle in the paratype with the caudalmargin of the quadrate. Much of the medial surface is ornamentedby a series of fine, radially-oriented ridges that presumablystrengthened the contact with the pterygoid. Ventrally, the quad-rate terminates in a rostrocaudally expanded lateral condyle that isroughly triangular in cross-section and forms most of the jaw joint.Themedial condyle is indistinct from the lateral condyle. The rostralmargin of the quadrate is dominated by the quadratojugal facet,which forms a broad arc almost 45% of the height of the quadrate.The contact surface is bevelled, widest at its dorsal and ventralextremities, and slightly crenulated at its midpoint.

3.1.2. Neurocranium3.1.2.1. Parietals. The fused parietals form the caudal roof of thebraincase. In lateral view, they are saddle-shaped and longer thanhigh. In dorsal view, the parietals are transversely constricted and

Please cite this article in press as: Bell, P.R., Redescription of the skullCretaceous Research (2010), doi:10.1016/j.cretres.2010.10.002

have a ‘length/minimal width’ ratio of >2 as is typical of hadro-saurines (Figs. 9 and 10). The parietals are keeled dorsally as inEdmontosaurus but unlike Corythosaurus, and slope rostroventrallyto form an obtuse angle with the frontals as in Brachylophosaurusand also Parasaurolophus. In each of AMNH 5221 and CMN 8796,the rostral margin is damaged but appears straight and is domi-nated by the suture for the frontals in dorsal view. There is noindication of a rostromedial process that intervenes between thefrontals; however its absence could be due to damage in thatregion. The postorbitals contact the parietals rostrolaterally,although in lateral view the laterosphenoid partially obscures thissuture (Fig. 10B). From the rostrodorsal margin, a triangulardepression tapers caudally terminating near the rostral limit ofthe dorsal keel. Caudally, the parietals expand where they meetthe squamosals. Damage on both specimens makes it unclearwhether or not they were excluded from the caudal margin of theskull by the squamosal. Along its ventral margin, the parietalcontacts the laterosphenoid rostrally and the prootic and opis-thotic-exoccipital caudally. In lateral view, the ventral marginsof the parietals form an angle of 110� and 130� on right and leftsides, respectively in AMNH 5221. The lateral surfaces of theelement are smooth and strongly concave.

of Saurolophus osborni Brown 1912 (Ornithischia: Hadrosauridae),

Fig. 10. Braincase of AMNH 5221 in A. right lateral, B. left lateral, C. caudal, D. rostral, and E. dorsal views. Scale ¼ 5 cm. Shading identifies area obscured by plaster in the originalspecimen. Parallel lines represent broken surfaces. Cross-hatching represents matrix. Boc, basioccipital; Bsp, basisphenoid; Ex, exoccipital-opisthotic complex; FM, foramenmagnum; Fr, frontal; Ifs, interfrontal suture; Lsp, laterosphenoid; Orb, orbitosphenoid; Par, parietal; Pre, presphenoid; Pfr, prefrontal; Po, postorbital; Pro, prootic; Sq, squamosal.

P.R. Bell / Cretaceous Research xxx (2010) 1e158

3.1.2.2. Frontal. Each frontal consists of a body, which forms theskull roof, a rostroventral, and a caudodorsal process. Together,the frontals form the rostral part of the braincase bound caudally bythe parietal and caudolaterally by the laterosphenoid (Figs. 9 and10). Because of damage in this region, it cannot be determinedwhether the fronto-parietal suture, which forms the rostral bordersof the supratemporal fenestrae, is straight or forms an anglebetween the frontals. The rostroventral process (identified asprefrontal by Brown, 1912; Fig. 3) forms an angle of 105� with thelongitudinal axis of the frontal body, forming a broad frontal plat-form over which the nasals and prefrontals lie (Figs. 9D, 10D). Thetwo rostroventral processes apparently did not fuse but abutted oneanother rather bluntly. Medially, each rostroventral process isthickened to form a longitudinal ridge that presumably helpedstrengthen the contact with the nasals ventrally and prefrontalsdorsally. In lateral view, the rostroventral processes taper

Please cite this article in press as: Bell, P.R., Redescription of the skullCretaceous Research (2010), doi:10.1016/j.cretres.2010.10.002

rostroventrally but are uniformly broad in rostral view. As thefrontal platform is broken ventrally, its dorsoventral height isunknown; however, it was at least as tall as the caudodorsal processof the frontal. The frontal platform continues caudodorsally byway of a single finger-like ramus on each frontal that buttressesthe underside of the nasal crest where it overhangs the skull roof.These are best seen on CMN 8796, which shows the buttress of theright frontal nested within a longitudinal groove on the underside ofthe right nasal (Fig. 5). The buttresses are damaged but appear atleast as high as the frontals are long and taper caudodorsally toa point. Left and right buttresses do not contact one another; thefrontals meet at the base of these processes by way of a low,short longitudinal lamina. The frontal bodies are flat lying.The rostrolateral margin of the frontal body partly encloses the baseof the dorsocaudal process of the prefrontals. The frontal is metlaterally by the postorbital along a suture that extends from the

of Saurolophus osborni Brown 1912 (Ornithischia: Hadrosauridae),

Fig. 11. Schematic drawing of the braincase based on AMNH 5221 in right lateralaspect showing foramina for cranial nerves (IeXII) and approximate boundaries(dashed lines) between elements. Scale ¼ 5 cm.

P.R. Bell / Cretaceous Research xxx (2010) 1e15 9

rostral border of the supratemporal fenestrae to the base of theprefrontal buttress thereby excluding the frontal from the orbitalrim (contra Brown, 1912).

3.1.2.3. Presphenoid. Homology of the hadrosaurid presphenoidis dubious (Evans, 2006); however, the name is retained forconsistency within the hadrosaurian literature. With its counter-part, the presphenoid forms most of the interorbital septum,bounded dorsally, caudally, and caudoventrally by the frontal,orbitosphenoid, and basisphenoid, respectively (Figs. 9A,B, 10A,B).It is higher than it is long and rectangular in lateral aspect. Whenviewed rostrally, the presphenoids are Y-shaped, diverging dorsallyto transmit the olfactory nerve. Just ventral to the midpoint onits caudal margin, a semi-circular invagination of the orbitosphe-noid suture forms the rostral half of the foramen for cranial nerveIV. The ventral margin of the bone lies parallel to and directlyabove the presphenoid (cultriform) process of the basisphenoid,although the two do not meet except at the caudoventral edge ofthe presphenoid.

3.1.2.4. Orbitosphenoid. The semi-circular orbitosphenoid issurrounded by the laterophenoid caudally, the frontal dorsally,and the presphenoid rostrally (Fig. 10A,B). The laterosphenoideorbitosphenoid union is straight. Rostroventrally, along with thepresphenoid, the orbitosphenoid forms the caudal half of theforamen for cranial nerve IV. The suturewith the basisphenoid alongthe ventralmargin of the bone is poorly defined, although it appearsthat the orbitosphenoid also contributed either partly or entirely tothe openings for cranial nerves II and III (Figs. 10 and 11). From thedorsal margin of cranial nerve III, a narrow groove for the occulo-motor nerve sweeps rostrodorsally in an arc, fading out dorsal to theforamen for cranial nerve II.

3.1.2.5. Laterosphenoid. Situated between the orbitosphenoid andthe prootic is the dorsoventrally elongate laterosphenoid (Figs.9A,B, 10A,B). It extends dorsally contacting the lateral corner ofthe fronto-parietal suture. The rounded dorsal head forms a syno-vial joint with the postorbital. A median ridge (crista antotica)extends the full height of the lateral surface of the laterosphenoid.The middle third of the caudal edge of the laterosphenoid is inunion with the prootic ventral to which it becomes constricted bythe large foramen for the trigeminal nerve (Fig. 11). A sulcus in thelateral surface of the laterosphenoid for the ophthalmic branch ofcranial nerve V separates the roughly circular preotic pendant fromthe remainder of the laterosphenoid. Ventral to the preotic pendantis a deep, bifurcated sulcus for the maxillary and mandibular

Please cite this article in press as: Bell, P.R., Redescription of the skullCretaceous Research (2010), doi:10.1016/j.cretres.2010.10.002

branches of the trigeminal nerve. Ventral to the foramen forcranial nerve V, the laterosphenoid fuses imperceptibly with thebasisphenoid and the prootic. Similarly, it is unclear to what extentthe laterosphenoid participates in the formation of the foramen forcranial nerve III.

3.1.2.6. Prootic. The large foramen for cranial nerve V divides therostral portion of the prootic (Figs. 9e11): the dorsal tab-shapedprocess abuts the laterosphenoid; the ventral process is less welldefined and fuses imperceptibly with the laterosphenoid andbasisphenoid. As in Kerberosaurus and Prosaurolophus, there is novertical groove extending from the dorsal limit of the openingfor cranial nerve V. This groove is present in S. angustirostris(Bolotsky and Godefroit, 2004). A smooth ridge (crista prootica)extends caudally from the foramen for cranial nerve V, parallel tothe dorsal margin of this element where the prootic abuts theparietal. A weaker ridge arranged dorsoventrally and caudallyconvergent with the crista prootica is continuous with the rostralborder of the oval window (fenestra ovalis). Based on the brain-case of Prosaurolophus, where the sutures are open (Horner, 1992),this ridge probably defines the junction between the prootic andexoccipital-opisthotic. Caudoventral to cranial nerve V, the prooticis perforated by a small opening for the seventh cranial nerve(Fig. 11). The palatine branch of the facial nerve forms a verticalgroove ventral to that opening.

3.1.2.7. Basisphenoid. When viewed ventrally, the unpaired basi-sphenoideparasphenoid complex is a triangular bone that formsthe base of the braincase. Each corner is drawn out to form a longprocess; the caudolateral pterygoid processes and a single,rostrally-directed parasphenoid (cultriform) process. Each finger-like pterygoid process extends ventrally and caudolaterally tomeet with a corresponding fossa on the pterygoid. The ros-troventrolateral margins of the basisphenoid are thickened toform ridges that converge rostrally and are continuous with theparasphenoid process. This process lies parallel to the ventral edgeof the presphenoid and curves slightly ventrally. Only the baseof the parasphenoid process is in contact with the presphenoid.In lateral aspect, the parasphenoid process and the pterygoidprocesses form an S-shaped curve (Fig. 9A). Caudal to the base ofthe pterygoid processes, the basisphenoid expands to form therostral part of the basal tubera. Due to reconstruction, the open-ings for the internal carotid arteries, which should be close to thebase of the pterygoid processes (Ostrom, 1961) are not visible inany specimen. Dorsally, no clear divisions can be discernedbetween the exoccipitals, prootic, laterosphenoid, or orbitosphe-noid with the basisphenoid. Consequently, the extent of thebasisphenoid in the formation of the openings for the secondand sixth cranial nerves is unknown (Fig. 11). A small breakperforates the basisphenoid at the point where cranial nerve VIlikely exited, preventing its identification.

3.1.2.8. Basioccipital. The basioccipital forms the caudoventralquarter of the braincase (Fig. 9A,B). It is divided into subequalhalves by a transverse constriction on its ventral surface into theoccipital condyle caudally and basal tubera rostrally. The occipitalcondyle is formed chiefly by the basioccipital. In caudal view, itsventral outline is rounded, and its dorsal edge concave down for theforamen magnum. The articular surface has a slightly dimpledtexture. The dorsolateral edges are inclined medially where theyabut the exoccipitals.

The basioccipital contributes approximately two-thirds of thevolume of the basal tubera. Together, the basal tubera are roughlyhemispherical and the same size as the occipital condyle. Thesurface of the suture between the two bones is rugose.

of Saurolophus osborni Brown 1912 (Ornithischia: Hadrosauridae),

P.R. Bell / Cretaceous Research xxx (2010) 1e1510

3.1.2.9. Otoccipital. The exoccipitals and opisthotics fuse veryearly in both hadrosaurine and lambeosaurine ontogeny to formthe otoccipital (Horner, 1992; Horner and Currie, 1994). Theyconverge medially and form the caudodorsal margin of theocciput. The otoccipital rises dorsolaterally, abutting a sulcus onthe squamosal at the caudolateral corner of the skull (Fig. 10C).From this sulcus, the paraoccipital process extends ventrolaterally,tapering only slightly before terminating bluntly at about the levelof the dorsal limit of the foramen magnum. The squamosalmaintains contact along most of the length of rostral face of theparaoccipital process, terminating just dorsal to the ventral limitof that process. The otoccipitals diverge ventrally from theirmidline suture sending off elongate basioccipital processes(exoccipital condyloid) that form a butt joint with the dorsolateraledges of the basioccipital. The sutures between these elements areinclined dorsomedially. Where it contacts the basioccipital, thebasioccipital process is slightly expanded laterally but does notcontact its mate. Together, the exoccipitals contribute approxi-mately 90% of the circumference of the laterally compressedforamen magnum (Fig. 10C). In lateral view, a strong ridge of bone(crista tuberalis) extends rostroventrally along the lateral wall ofthe basioccipital process of the otoccipital. On the left side ofAMNH 5221, where the break permits observation of the interiorof the otoccipital, three foramina are present ventral and parallelto this ridge, the rostral two probably corresponding to cranialnerves X and XI, the caudal most opening providing passage forthe twelfth cranial nerve (Fig. 11). The crista tuberalis terminatesrostroventrally at about the boundary between the otoccipital andbasioccipital. A relatively large triangular opening immediately

Fig. 12. Pterygoids of AMNH 5221. AeB, left pterygoid in A. lateral and B. medial views. Ceprocess; ect, ectopteygoid process; pal, palatine process; vq, ventral quadrate process. Dasheobscured by plaster.

Please cite this article in press as: Bell, P.R., Redescription of the skullCretaceous Research (2010), doi:10.1016/j.cretres.2010.10.002

dorsal to the crista tuberalis transmitted cranial nerve IX andconverges with the oval window demarcating the rostral extent ofthe otoccipital with the prootic.

3.1.2.10. Supraoccipital. Although presumably present in bothAMNH 5220 and 5221, the supraoccipital is visible, albeit damaged,only in the latter. It occupies the usual hadrosaur position betweenthe parietals rostrodorsally and the otoccipitals ventrally and cau-dolaterally. The caudal face is entirely missing; however, it is clearlyoccluded from the foramenmagnum by the exoccipital-opisthotics.Based on the inclination of the otoccipitals, the ventral intracranialsurface of the supraoccipital would have been inclined caudally atapproximately 45� as is typical for Hadrosauridae. All other featuresand contacts are obscured by damage.

3.1.3. Palate3.1.3.1. Pterygoid. The pterygoids are incomplete in both AMNH5221 and CMN 8796. Each consists of four triangular processesthat radiate from a central plate (Fig. 12). The two largestprocesses, the palatine and dorsal quadrate processes, diverge at139� along their dorsal edges. The former bifurcates into tworounded prongs for contacts with the vomer and palatine. Thepalatine process can be seen in articulation through the orbit inCMN 8796 where it rises up to meet the palatine extensivelyalong its leading edge. The smooth lateral faces of the dorsal andventral quadrate processes broadly contact the medial surface ofthe pterygoid process of the quadrate. The shorter ventral quad-rate process is broken on both specimens. The stout ectopterygoidprocess projects rostroventrally and medially, and is flattened

D, right pterygoid in C. lateral and B. medial views. Scale ¼ 5 cm. dq, dorsal quadrated line indicates the approximate outline of the missing regions; shading indicates area

of Saurolophus osborni Brown 1912 (Ornithischia: Hadrosauridae),

P.R. Bell / Cretaceous Research xxx (2010) 1e15 11

along its ventral edge where it meets the reciprocal process fromthe ectopterygoid. From the dorsal margin of the element in thesaddle between the dorsal quadrate and palatine processes, threestrong ridges radiate to buttress the medial surfaces of the pala-tine, ectopterygoid, and ventral quadrate process. The dorsalquadrate process is not buttressed.

3.1.3.2. Palatine. Although no complete palatine is known,enough can be observed through the orbits of CMN 8796 to gainan appreciation of the morphology of this bone. The right pala-tine is a subtriangular plate oriented dorsomedially. The rostro-lateral margin is expanded dorsoventrally where it contacts thejugal and lacrimal. The ventrolateral margin is a gentle S-shapedsinuosity where it presumably came into contact with the dorsaledge of the maxilla. The triangular dorsomedial extension ispreserved only on the left side (contra Heaton, 1972) where it isbroken and shifted from life position. The palatine process of thepterygoid contacts the caudal edge of this extension alonga distinct facet.

3.1.3.3. Vomer. The laterally-compressed vomers are triangular inlateral view and meet with each other along their entire medialsurfaces. The ventral edge of the triangle is weakly concave inlateral view. The rostral half of the vomer is of uniform medio-lateral width and tapers to an acute point, the ventral edge of whichis flattened towards the tip (Fig. 13). The relationship between thevomers, premaxillae and maxillae could not be confirmed but it isassumed the three met along the rostral-most portion of thevomers as in other hadrosaurines (Horner, 1992). The caudalprocess is roughly rectangular with a blunt, squared-off terminus.One and two parallel longitudinal ridges occur on both lateral andmedial surfaces, respectively. Ventrally, this process is narrow andblade like, but becomes more robust dorsally. The dorsal apex is themost robust part of the element, expanding transversely to meetthe palatines laterally and the palatine processes of the pterygoidscaudally. The dorsal apex lies at about the midlength of the

Fig. 13. Left vomer of AMNH 5221 in A. lateral and B. medial views. Reconstruc

Please cite this article in press as: Bell, P.R., Redescription of the skullCretaceous Research (2010), doi:10.1016/j.cretres.2010.10.002

element, contrasting with that of Prosaurolophus, where it is situ-ated more caudally.

3.1.4. Mandible3.1.4.1. Predentary. As in all hadrosaurs, the single, horseshoe-shaped predentary wraps around the edentulous portions of thedentaries to form a distinct beak. The rostral margin is smoothlysquared off (contra Horner et al., 2004) as it is in all other Hadro-sauridae. The oral margin of the predentary in AMNH 5220preserves six low, rounded tuberosities, the lateral-most of whichare the smallest. The caudolateral extensions of the predentaryoverlie the dorsal surfaces of the dentaries for approximately halfthe length of the edentulous portion of each of those elements.These processes are smooth and flattened along their dorsal edges,and are dorsoventrally expanded caudally. Because of the curvatureof the premaxillae, the predentary does not occlude with theupper jaw except along the flattened dorsal surfaces of these cau-dolateral extensions. The rostral gap was likely filled by a kerati-nous sheath over one or both of these elements (Versluys, 1923).On either side of the caudal midline, a short, rounded processextends caudally to support the underside of each dentary. As thepredentary is preserved only in articulated specimens, informationabout the dorsomedial surface is not available.

3.1.4.2. Dentary. The dentary is the sole tooth-bearing and largestelement in the lower jaw. Unlike most hadrosaurs, the ventralmargin is straight in lateral view, and medial inflection at thesymphysis is only weakly developed. The rostral edentulousportion of the dentary is about as long as the tooth row and tapersrostrally towards the symphysis where it meets its mate. Severalsmall neurovascular foramina exit rostrolaterally in this region.The alveolar margin is roughly parallel to the ventral margin of thedentary. The prominent coronoid process is offset laterally fromthe tooth row, arising from the caudolateral border of the dentarywhere it projects rostrodorsally into the space between the jugaland maxilla. Dorsally, the coronoid process expands rostrocaudally

ted outline based on the right vomer of the same specimen. Scale ¼ 5 cm.

of Saurolophus osborni Brown 1912 (Ornithischia: Hadrosauridae),

P.R. Bell / Cretaceous Research xxx (2010) 1e1512

for the insertion of several of the major jaw muscles (Rybczynskiet al., 2008). Caudally, the dentary is excavated by a deep Mecke-lian fossa (adductor chamber), which continues rostrally as a deepcleft on the medial surface of the dentary ventral to the dentalbattery. The angular extensively contacts the dentary ventral tothis cleft. The coronoid process of the surangular extends dorso-ventrally on the caudolateral margin of the dentary.

The tooth battery is the most conspicuous feature on the medialaspect of the dentary and is covered (except for the occlusal-mostteeth) by a thin plate of bone. The medial surface of this supra-dentary plate is dorsoventrally grooved for the alveoli. There areapproximately forty alveoli in the adult dentary (Brown, 1912);however, the total number cannot be precisely determined on anyspecimen of S. osborni. At the base of each alveolus, a special dentalforamen (Edmond, 1957) perforates the supradentary plate. Theseforamina form a gently dorsally-concave line in medial view. Thecaudal end of the alveolar process is drawn out into a conicalsplenial process, which medially contacts the splenial.

3.1.4.3. Surangular. The largest of the post-dentary elements isthe surangular. Rostrally, the surangular has a tall, thin coronoidprocess that extends along the medial side of the coronoid processof the dentary. Behind this process, the surangular expandsmediolaterally to form a broad surface for articulation with theventral condyle on the quadrate. The surangular is covered medi-ally by the splenial and ventrally by the angular. The caudalprocess is mediolaterally flattened and curves caudodorsally toform the margin of a V-shaped cleft that separates itself from theangular (and possibly the splenial). The articular rests in this cleft.A ridge that originates on the lateral margin of the quadrate cotyleextends along the length of the lateral surface of the caudalprocess.

3.1.4.4. Articular. The small, laterally-compressed articular residesbetween the surangular laterally and the splenial and angularmedially. In AMNH 5220 and CMN 8796, each articular projectsdorsally from the neighbouring bones to give the caudal terminusof the mandible an almost hook-like appearance. A short ridge onits lateral surface extends rostroventrally from the dorsal tip tomeet a similar ridge on the caudal process of the surangular.

3.1.4.5. Angular. The strap-like angular is bowed to accommodatethe curvature of the post-dentary bones. Rostrally, it extends alongthe ventromedial edge of the dentary below the Meckelian grooveand curves dorsomedially along its caudal length where it meetswith the splenial dorsally and surangular medially. At its caudalend, the surangular and angular form a V-shaped cleft for thearticular.

3.1.4.6. Teeth. Brown (1912) originally described sixty alveoli inthe maxilla and forty-four in the dentary of AMNH 5221. However,plaster reconstruction on these elements has all but obscured theteeth. Where they are visible on the dentaries, the tooth crownsform asymmetrical diamonds; the long-axes are oriented perpen-dicular to the jaw, and are arranged into vertical ‘families’ of threeor possibly four successive teeth, up to three of which may bevisible on the occlusal surface. The point where the two distal edgesmeet is slightly dorsal relative to the equivalent mesial point.This results in the dentary teeth forming longitudinal rows inclinedcaudoventrally at a shallow angle. Ventral to the occlusal plane,each tooth contacts six neighbouring teeth. As in other hadro-saurines, each tooth has a single, straight, non-denticulate mediancarina that extends the length of the enamelled face. The peripheryof each dentary tooth is devoid of papillae as is usual for Hadro-saurinae. The labial margin of the occlusal surface of each tooth is

Please cite this article in press as: Bell, P.R., Redescription of the skullCretaceous Research (2010), doi:10.1016/j.cretres.2010.10.002

roughly U-shaped in occlusal view, whereas the median carinagives the lingual margin an expanded W-shape.

3.1.4.7. Hyoid. The hyoids of AMNH 5220 are disarticulated fromthe rest of the skull, and one lies caudal to the quadrates whereasthe second is resting in the right orbit below the eye. Theypresumably articulated in a similar fashion to those described byOstrom (1961) for Corythosaurus. As preserved, left and right hyoidscannot be distinguished and are observable only inmedial or lateralview. The mediolaterally compressed rostral end is asymmetricalbetween left and right sides, one being rounded and the otherflattened. The rostral end is dorsoventrally expanded and graduallytapers caudally, as in Corythosaurus (Ostrom, 1961). The distal endsare missing but appear to deflect caudodorsally. There is a low,rounded ridge caudal to the rostral end, just ventral to midheight.

4. Phylogenetic analysis

Forty-four cranial and dental characters (with modificationswhere noted in Appendix 1), compiled from Weishampel et al.(1993), Godefroit et al., 1998; Godefroit et al. (2008), Bolotsky andGodefroit (2004), Horner et al. (2004), and Prieto-Marquez(2005), were used to determine the phylogenetic position of Sau-rolophus within the Hadrosauridae. Because the post-cranium of S.osborni could not be adequately reassessed, post-cranial characterswere excluded. Nine ingroup taxa (Hypacrosaurus, Lambeosaurus,Gryposaurus, Brachylophosaurus, Maiasaura, Prosaurolophus,Edmontosaurus, Kerberosaurus, and Saurolophus) and two out-groups, Iguanodon and Bactrosaurus, were selected by specimencompleteness, by the existence of adequate published descriptions,and/or by the availability of specimens to the author. Where generarepresent multispecific groupings, the type species was taken asthe representative of that genus following Godefroit et al. (2008).Kerberosaurus from the Amur region of eastern Russia was includedin the present analysis, as it has been recently suggested to be thesister taxon to a monophyletic clade formed by Prosaurolophus andSaurolophus (Bolotsky and Godefroit, 2004). S. angustirostris isexcluded here pending a thorough redescription (currently inpreparation by the author) and comparison with the North Amer-ican species. All characters were treated as unordered andassigned equal weight. Using the character matrix (Appendix 2),an exhaustive search was performed using PAUP 4.0b10(Swofford, 2002), resulting in three most parsimonious trees witha length of 69 steps. The strict consensus tree had a retention indexof 0.81, a consistency index of 0.84, and a rescaled consistencyindex of 0.68 (Fig. 14). Because of the number of undeterminedcharacter states for Kerberosaurus, a second analysis was performedthat omitted that taxon resulting in a single most parsimonioustree that retained Saurolophus and Prosaurolophus as sister taxawith no change in the length or support values. Bootstrap valuesachieved using the branch-and-bound option show excellentsupport for a HadrosaurinaeeLambeosaurinae split; relationshipswithin the Hadrosaurinae are generally not well supported. Theonly moderately well-supported clade here (Brachylophosaur-useMaiasaura, bootstrap ¼ 74) has been described elsewhere(Prieto-Marquez, 2005; Prieto-Marquez et al., 2006; Godefroitet al., 2008). Moderate support exists for the Edmontosaur-useKerberosauruseProsaurolophuseSaurolophus clade, althoughrelationships within this group are more ambiguous. The phylo-genetic position of Saurolophus has shifted considerably in recentanalyses of the Hadrosaurinae. Norman (2002) and Horner et al.(2004) suggest Saurolophus is a primitive hadrosaurine. Severalmore recent analyses place Saurolophus as the sister taxon toProsaurolophus in a more derived position, but with variablerelationships amongst other hadrosaurines (Bolotsky and

of Saurolophus osborni Brown 1912 (Ornithischia: Hadrosauridae),

Fig. 14. Strict consensus of three most parsimonious trees resulting from a maximumparsimony analysis using the exhaustive search option of nine ingroup and two out-group taxa. Numbers are bootstrap values/decay indices.

P.R. Bell / Cretaceous Research xxx (2010) 1e15 13

Godefroit, 2004; Prieto-Marquez, 2005; Prieto-Marquez et al.,2006; Gates and Sampson, 2007; Godefroit et al., 2008). Resultsof the present analysis agree well with topologies proposed byBolotsky and Godefroit (2004) and Godefroit et al. (2008) despitethe omission of post-cranial characters. Brachylophosaurus andMaiasaura form a well-supported clade that is the sister group toall other hadrosaurines and Edmontosaurus is the sister taxon tothe clade that includes Kerberosaurus, Prosaurolophus, and Sau-rolophus. Although the strict consensus tree shows a polytomybetween Kerberosaurus, Prosaurolophus, and Saurolophus, one ofthe three most parsimonious trees in this study (Kerberosaurus(Prosaurolophus þ Saurolophus)) does match that of Godefroit et al.(2008) thereby adding some support to that phylogeny.However, resolution of these taxa in the analysis by Godefroitet al. (2008) hinged on a single character at each node (Kerber-osaurus þ Prosaurolophus þ Saurolophus united by frontals thatare excluded from the orbital rim by the postorbitalepre-frontal complex [character 5(1), bootstrap ¼ 58]; Prosaurolophus þSaurolophus united by a solid supracranial crest present and exca-vated laterally by the circumnarial fossa [character 20(2),bootstrap ¼ 54]) and was thus weakly supported.

5. Discussion and conclusions

The holotype of S. osborni (AMNH 5220) remains the bestspecimen of this taxon, yet considerable confusion has since arisenregarding the construction of the crest in S. osborni. Brown (1912)posited that the nasals, frontals, and prefrontals all contributed tothe formation of the crest. The nasals, he suggested, werebuttressed laterally and caudally by the prefrontals and frontals butdoes not show the frontals in his caudal rendering of the skull(Brown,1912; Fig.1a). Because of these inconsistencies and becauseexamination of the holotype has been compromised, this view hasbeen challenged by several authors who do not regard the frontalsas part of the crest (Ostrom, 1961; Horner, 1992; Horner et al.,2004). Observations of the crest in CMN 8796 clearly show

Please cite this article in press as: Bell, P.R., Redescription of the skullCretaceous Research (2010), doi:10.1016/j.cretres.2010.10.002

portions of the frontal still in contact with the nasals; each frontalcontributes a single caudodorsal process that is received in itsown recess on the underside of the nasal crest. This condition isvirtually identical to that of S. angustirostris (Marya�nska andOsmólska, 1981, MPC 100/179). However, the caudodorsalprocesses of the frontals in S. osborni are less developed. Assumingsimilar proportions in crest length between the two species, thecaudodorsal process of the frontal would have been only one-fifththe length of the crest in S. osborni as opposed to approximatelyhalf the length in S. angustirostris (Marya�nska and Osmólska, 1981,MPC 100/179).

The frontals of S. osborni appear unique among hadrosaurs intheir configuration. Amongst other crested hadrosaurines (Bra-chylophosaurus, Maiasaura, Prosaurolophus), the frontals contributeonly in Maiasaura to the formation of the crest, rising to form thecaudal and dorsal extent of that structure (Horner, 1983). In Bra-chylophosaurus, the external surfaces of the frontals as viewed fromabove are foreshortened and do not contribute to the formation ofthe crest (Prieto-Marquez, 2005; Fig. 8). In species of Prosaur-olophus, the crest is derived from a short protuberance on thenasals; the frontals flatly abut the caudal margins of the nasals(Horner, 1992). In the architecture of the caudodorsal process,S. osborni most closely resembles the lambeosaurines Para-saurolophus and Charonosaurus, in both of which there is elongationof the frontal platforms to form ‘dorsal promontoria’ (Godefroitet al., 2001; Evans et al., 2007). In Charonosaurus, this promon-torium is the broadest part of the frontal (Godefroit et al., 2001)whereas in Parasaurolophus the finger-like process is more narrowand reminiscent of that in S. osborni. A finger-like caudodorsalprocess was illustrated for S. angustirostris by Marya�nska andOsmólska (1981, Fig. 2); however, in MPC 100/179 this process isdelicate and thin, but is broader than in S. osborni. Similarly, therostroventral process, which supports the nasals and prefrontalsin front of the crest, is unique to Saurolophus. In other crestedhadrosaurines the frontal-nasal union converges on the lambeo-saurine condition in which the frontal has an extended, caudally-inclined sutural surface (rostrallt-inclined in Maiasaura). Therostroventral process in S. osborni is elongate and rather smoothexternally and does not have the deep grooves and ridges seen inlambeosaurines. Rather than forming a strong interdigitatingsuture between these two elements, the nasals appear to simplyoverlie this ramp in a comparatively weak union. This feature isshared by S. angustirostris (e.g. ZPAL MgD-1/159, MPC 100/179) inwhich the frontals can be observed through the orbits extendingrostrally underneath the nasals.

Although Brown (1912, p. 135) remarked on a prefrontal processthat laterally buttressed the crest, this arrangement was neverfigured. Subsequent workers (Ostrom, 1961; Horner, 1992; Horneret al., 2004) dismissed the role of the prefrontals in the crest ofSaurolophus, instead maintaining that the nasals were entirelyresponsible for crest formation. The prefrontals in CMN 8796,although broken, preserve the base of what would have beena caudodorsal extension that would have flanked the nasals asdescribed by Brown (1912). In S. angustirostris, these processes arethin and intervene at the naso-frontal suture on the skull roof. Ifthe identification of the prefrontal on AMNH 5221 is correct, thenthat condition was also present in S. osborni.

Brown (1912) also indicated that the frontals extend laterally toform the dorsal margin of the orbit; however in all three specimens,a suture is apparent between the prefrontals and postorbitals,which excludes the frontals from the orbital margin (Horner, 1992).The prefrontal, which forms the dorsal orbital rim in front of thepostorbital is associated with a supraorbital element as suspectedby Coombs (pers. comm. in Marya�nska and Osmólska, 1979). Theseelements were previously identified as the frontal and prefrontal,

of Saurolophus osborni Brown 1912 (Ornithischia: Hadrosauridae),

P.R. Bell / Cretaceous Research xxx (2010) 1e1514

respectively (Brown, 1912). Although it is only visible in AMNH5220, it appears that the suture between these two elementsdid not close until late in life and perhaps did not close at all.Although supraorbitals have been observed in juvenile hadrosaur-ines (including S. angustirostris Marya�nska and Osmólska, 1979,ZPAL MgD-1/159), the suture between them tends to close rela-tively early in ontogeny such that it is obscured in subadult andadult individuals. A closed suture has also been suspected in anadult specimen of Kerberosaurus (Bolotsky and Godefroit, 2004).

Although it is not the intention of the current study to speculateon the dispersal of this group, the relationship among Kerber-osaurus, Prosaurolophus, and Saurolophus indicates a complexevolutionary history. The solid cranial crest appears to have evolvedat least twice in hadrosaurine phylogeny; once in the group thatled to Brachylophosaurus and Maiasaura and again in the Kerber-osauruseProsaurolophuseSaurolophus clade. Depending on thephylogenetic position of the flat-headed hadrosaurine Kerber-osaurus, then the non-crested condition was either secondarilydeveloped or was the primitive condition for Prosaurolophus andSaurolophus. Crest elongation in crested hadrosaurines (exempli-fied by Saurolophus) appears to have resulted in the convergenceof several characters with some lambeosaurines, in particularwith the ‘parasaurolophs’ (Parasaurolophus and Charonosaurus),whose crests most closely resemble that of Saurolophus. In thesethree genera, crest elongation was accompanied by elongation ofthe frontals into a form of dorsal promontorium and by down-warping of the frontal-parietal suture to form an obtuse anglebetween these two elements in lateral view. In taxa most closelyrelated to Saurolophus (Kerberosaurus and Prosaurolophus) wherethe crest is either absent or incipient, these character states areabsent. This is the case in all other hadrosaurines except Brachylo-phosaurus, in which crest elongation has indeed occurred and isaccompanied by down-warping of the frontal-parietal suture.

Acknowledgements

D. Evans, K. Seymour (ROM), K. Shepherd, M. Feuerstack (CMN),C. Mehling (AMNH), M. Borsuk-Biatynickia (ZPAL), C. Tsogbaatar,D. Badamgarov, and R. Barsbold (MPC) are thanked for theirhospitality and access to specimens in their care. D. Eberthprovided information regarding the stratigraphic context of AMNH5220 and 5221. D. Larson andM. Burns are gratefully acknowledgedfor assistance with PAUP. I am indebted to A. Paulina Carabajal fordiscussions on braincase anatomy. P. Currie is thanked for hisinvaluable support and supervision, and for comments on anearlier version of this manuscript. Helpful comments by twoanonymous reviewers greatly improved this paper. All illustrationsare by the author. This project was made possible by a researchgrant from the Dinosaur Research Institute.

References

Bell, P.R., Evans, D.C., 2010. Revision of the status of Saurolophus (Hadrosauridae)from California. Canadian Journal of Earth Sciences 47, 1417e1426.

Bolotsky, Y.L., Godefroit, P., 2004. A new hadrosaurine dinosaur from the LateCretaceous of far eastern Russia. Journal of Vertebrate Paleontology 24, 351e365.

Brown, B., 1912. A crested dinosaur from the Edmonton Cretaceous. Bulletin of theAmerican Museum of Natural History 31, 131e136.

Brown, B., 1913. The skeleton of Saurolophus, a crested duck-billed dinosaur fromthe Edmonton Cretaceous. Bulletin of the American Museum of Natural History32, 387e393.

Eberth, D.A., 2004. A revised stratigraphy for the Edmonton Group (Upper Creta-ceous) and its potential sandstone reservoirs. CSPG-CHOA-CWLS Joint Confer-ence, Calgary, Alberta, May 31eJune 4, 2004; Pre-Conference Field Trip #7, 44 p.

Eberth, D.A., Deino, A., 2005. New 40Ar/39Ar ages from three bentonites in theBearpaw, Horseshoe Canyon and Scollard formations (Upper Creta-ceousePaleocene) of southern Alberta, Canada. In: D.R.Braman, F.Therrien,E.B.Koppelhus, W.Taylor (Eds.), Dinosaur Park Symposium, Short Papers,

Please cite this article in press as: Bell, P.R., Redescription of the skullCretaceous Research (2010), doi:10.1016/j.cretres.2010.10.002

Abstracts, and Program, Royal Tyrrell Museum, Drumheller, Alberta, September24e25, 2005. Special Publication of the Royal Tyrrell Museum, pp. 23e24.

Edmond, A.G., 1957. On the Special Foramina in the Jaws of Many OrnithischianDinosaurs, vol. 48. Royal Ontario Museum Division of Zoology and Palae-ontology Contributions. 1e14.

Evans, D.C., 2006. Nasal cavity homologies and cranial crest function in lambeo-saurines dinosaurs. Paleobiology 32, 109e125.

Evans, D.C., Reisz, R.R., Dupuis, K., 2007. A juvenile Parasaurolophus (Ornithischia:Hadrosauridae) braincase from Dinosaur Provincial Park, Alberta, withcomments on crest ontogeny in the genus. Journal of Vertebrate Paleontology27, 642e650.

Gates, T.A., Sampson, S.D., 2007. A new species of Gryposaurus (Dinosauria:Hadrosauridae) from the late Campanian Kaiparowits formation, southernUtah, USA. Zoological Journal of the Linnean Society 151, 351e376.

Gates, T.A., Farke, A.A., 2009. Biostratigraphic and biogeographic implications ofa hadrosaurid (Ornithopoda: Dinosauria) from the Upper Cretaceous AlmondFormation of Wyoming, USA. Cretaceous Research 30, 1157e1163.

Godefroit, P., Dong, J.D., Orth, C.J., Pillmore, C.L., Tschudy, R.H., 1998. Bactrosaurus(Dinosauria: Hadrosauroidea) material from Iren Dabasu (Inner Mongolia, P.R.China). Bulletin de l’Institut royal des Sciences naturelles de Belgique, Sciencesde la Terre 68, 3e70.

Godefroit, P., Zan, S., Jin, L., 2001. The Maastrichtian (Late Cretaceous) lambeosaurinedinosaur Charonosaurus jiayinensis from north-eastern China. Bulletin de l’In-stitut royal des Sciences naturelles de Belgique, Sciences de la Terre 71, 119e168.

Godefroit, P., Hai, S., Yu, T., Lauters, P., 2008. New hadrosaurid dinosaurs from theuppermost Cretaceous of northeastern China. Acta Palaeontologica Polonica 53,47e74.

Heaton, M.J., 1972. The palatal structure of some Canadian Hadrosauridae (Reptilia:Ornithischia). Canadian Journal of Earth Sciences 9, 185e205.

Hieronymus, T.L., Witmer, L.M., Tanke, D.H., Currie, P.J., 2009. The facial integumentof centrosaurine ceratopsids: morphological and histological correlates of novelskin structures. Anatomical Record 292, 1370e1396.

Horner, J., 1983. Cranial osteology and morphology of the type specimen of Maia-saura peeblesorum (Ornithischia: Hadrosauridae), with discussion of its phylo-genetic position. Journal of Vertebrate Paleontology 3, 29e38.

Horner, J., 1992. Cranial Morphology of Prosaurolophus (Ornithischia: Hadrosaur-idae) with Descriptions of Two New Hadrosaurid Species and an Evaluation ofHadrosaurid Phylogenetic Relationships. In: Museum of the Rockies OccasionalPaper No. 2 119pp.

Horner, J.R., Currie, P.J., 1994. Embryonic and neonatal morphology and ontogeny ofa new species of Hypacrosaurus (Ornithischia, Lambeosauridae) from Montanaand Alberta. In: Carpenter, K., Hirsch, K.F., Horner, J.R. (Eds.), Dinosaur Eggs andBabies. Cambridge University Press, Cambridge, U.K, pp. 312e336.

Horner, J.R., Weishampel, D.B., Forster, C.A., 2004. Hadrosauridae. In:Weishampel, D.B., Dodson, P., Osmólska, H. (Eds.), The Dinosauria, SecondEdition. University of California Press, Berkeley, California, pp. 438e463.

Jerzykiewicz, T., Russell, D.A., 1991. Late Mesozoic stratigraphy and vertebrates ofthe Gobi Basin. Cretaceous Research 12, 345e377.

Lund, E., Gates, T., 2006. A historical and biogeographical examination of hadro-saurid dinosaurs. Bulletin of the New Mexico Museum of Natural History 35,263e276.

Marya�nska, T., Osmólska, H., 1979. Aspects of hadrosaurian cranial anatomy. Lethaia12, 265e273.

Marya�nska, T., Osmólska, H., 1981. Cranial anatomy of Saurolophus angustirostriswith comments on the Asian Hadrosauridae (Dinosauria). PalaeontologiaPolonica 42, 5e24.

Marya�nska, T., Osmólska, H., 1984. Post-cranial anatomy of Saurolophus angustirostriswith comments on other hadrosaurs. Palaeontologia Polonica 46, 119e141.

Morris, W.J., 1973. A review of Pacific coast hadrosaurs. Journal of Paleontology 47,551e561.

Norman, D., 2002. On Asian ornithopods (Dinosauria: Ornithopoda). 4. Redescrip-tion of Probactrosaurus gobiensis Rozhdesvensky 1966. Zoological Journal ofthe Linnean Society 136, 113e144.

Norman, D., Sues, H.-D., 2000. Ornithopods fromKazakhstan, Mongolia and Siberia. In:Benton,M.J., Shishkin,M.A.,Unwin,D.M., Korochin,E.N. (Eds.), TheAgeofDinosaursin Russia andMongolia. Cambridge University press, Cambridge, UK, pp. 462e479.

Ostrom, J.H., 1961. Cranial morphology of the hadrosaurian dinosaurs of NorthAmerica. Bulletin of the American Museum of Natural History 122, 33e186.

Prieto-Marquez, A., 2005. New information of the cranium of BrachylophosaurusCanadensis (Dinosauria, Hadrosauridae), with a revision of its phylogeneticposition. Journal of Vertebrate Paleontology 25, 144e156.

Prieto-Marquez, A., Gaete, R., Rivas, G., Galobart, A., Boada, M., 2006. Hadrosauroiddinosaurs from the Late Cretaceous of Spain: Pararhabdodon isonensis revisitedand Koutalisaurus kohlerorum gen. et sp. nov. Journal of Vertebrate Paleontology26, 929e943.

Rozhdestvenskii, A.K., 1952. A new representative of duckbilled dinosaurs from theUpper Cretaceous deposits of Mongolia (in Russian). Doklody Akademic SSSR86, 405e408.

Rozhdestvensky, A.K., 1957. The duck-billed dinosaur Saurolophus from the UpperCretaceous of Mongolia (in Russian). Vertebrata Palasiatica 1, 129e149.

Russell, D.A., Chamney, T.P., 1967. Notes on the biostratigraphy of dinosaurian andmicrofossil faunas in the Edmonton Formation (Cretaceous), Alberta. NationalMuseum of Canada Natural History Papers 35, 22.

Rybczynski, N., Tirabasso, A., Cuthbertson, R., Holliday, C., 2008. A three dimen-sional animation of Edmontosaurus for testing chewing hypotheses.

of Saurolophus osborni Brown 1912 (Ornithischia: Hadrosauridae),

P.R. Bell / Cretaceous Research xxx (2010) 1e15 15

Paleontologica Electronica 11 (2), 1e14. http://palaeo-electronica.org/2008_2/132/index.html (accessed 10.10.08).

Swofford, D.L., 2002. Phylogenetic Analysis Using Parsimony (And Other Methods).Version 4.0b10. Sinauer Associates, Sunderland, Massaschusetts, 40 pp.

Versluys, J., 1923. The skull and skeleton of Trachodon annectens in the SenckenbergMuseum (in German). Abhandlungen der Senckenbergischen Nature-forschenden Gesellschaft 28, 1e19.

Weishampel, D.B., Horner, J.R., 1990. Hadrosauridae. In: Weishampel, D., Dodson, P.,Osmólska, H. (Eds.), The Dinosauria. University of California Press, Berkeley,pp. 534e561.

Please cite this article in press as: Bell, P.R., Redescription of the skullCretaceous Research (2010), doi:10.1016/j.cretres.2010.10.002

Weishampel, D.B., Norman, D.B., Grigorescu, D., 1993. Telmatosaurus transsylvanicusfrom the Late Cretaceous of Romania: the most basal hadrosaurid dinosaur.Palaeontology 36, 361e385.

Appendix. Supplementary data

Supplementary data associated with this article can be found, in the onlineversion, at doi:10.1016/j.cretres.2010.10.002.

of Saurolophus osborni Brown 1912 (Ornithischia: Hadrosauridae),