A new ceratopsid dinosaur (Ornithischia) fr om
the uppermost Horseshoe Canyon Formation
(upper Maastrichtian), Alberta, Canada
Xiao-chun Wu, Donald B. Brinkman, David A. Eberth, and Dennis R. Braman
Abstract: A skeleton of a new ceratopsid dinosaur, Eotriceratops xerinsularis gen. et sp. nov., is described in this paper.
It is the first associated vertebrate skeleton found within the upper 20 m of the Horseshoe Canyon Formation.
Eotriceratops xerinsularis is a large chasmosaurine that differs from other chasmosaurines in a unique set of features in
the premaxilla, nasal horn core, squamosal frill, and epijugal. The most striking of those features includes an extremely
tall, non-recessed narial process of the premaxilla; the presence of greatly elongate, spindle-shaped epoccipitals on the
squamosal frill; a deep, well-demarcated fossa on the anteroventral surface of the squamosal frill; a sharply conical
epijugal with a pronounced proximoposterior process and separate fossa-like facets for the jugal and quadratojugal; and
the presence of an obliquely extending vascular trace meeting a transverse vascular trace ventrally on the anterior surface
of the nasal horn core. Our phylogenetic analysis suggests that E. xerinsularis is nested within a clade including
Triceratops, Diceratops, and Torosaurus, which are all from late Maastrichtian deposits. The upper 20 m of the Horseshoe
Canyon Formation comprises a coal-rich interval (Carbon–Thompson coal zone, unit 5), which previously has been as
signed to upper Maastrichtian magnetochrons 31n and 30r, and the Mancicorpus gibbus miospore subzone. The ceratopsid
specimen was collected from between the Carbon and Thompson coal seams, and thus, is inferred to (1) occur near the
top of magnetochron 31n and (2) have an age of 67.6–68.0 Ma. Large chasmosaurine ceratopsids, such as Triceratops and
Torosaurus, have not previously been described from the Horseshoe Canyon Formation or from magnetochron 31n or the
M. gibbus miospore subzone. Thus, Eotriceratops is distinctly older than any other ceratopsid in the Triceratops group,
and the discovery of E. xerinsularis helps fill a biostratigraphic gap between early and late Maastrichtian chasmosaurines.
Wu et al. 1265Résumé : Le squelette d’un nouveau dinosaure cératopsidé, l’Eotriceratops xerinsularis gen. et sp. nov., est décrit
dans cet article. Il s’agit du premier squelette vertébré associé observé dans les 20 derniers mètres de la Formation de
Horseshoe Canyon. L’Eotriceratops xerinsularis est un grand chasmosauriné qu’un ensemble unique de caractères des
prémaxillaires, du centre de la corne nasale, de la frange squamosale et de l’épijugal distinguent d’autres chasmosauri-
nés. Parmi les caractères les plus saillants figurent un processus narial des prémaxillaires non renfoncé et extrêmement
long, la présence d’époccipitaux en forme de fuseaux très allongés sur la frange squamosale, une fosse profonde et
bien démarquée sur la surface antéroventrale de la frange squamosale, un épijugal fortement conique présentant un pro
cessus proximopostérieur prononcé et des facettes séparées en forme de fosse pour le jugal et le quadratojugal, et la
présence d’une trace vasculaire oblique qui rejoint ventralement une trace vasculaire transversale sur la surface anté
rieure du centre de la corne nasale. L’analyse phylogénétique laisse croire que l’E. xerinsularis s’insère dans un clade
qui comprend le Triceratops,leDiceratops et le Torosaurus, tous présents dans des dépôts du Maastrichtien tardif. Les
20 derniers mètres de la Formation de Horseshoe Canyon comprennent un intervalle riche en charbon (zone à charbon
Carbon–Thompson; unité 5), précédemment affectée aux magnétochrons 31n et 30r du Maastrichtien supérieur, ainsi
qu’à la sous-zone à miospores du Mancicorpus gibbus. Le spécimen de cératopsidé a été prélevé entre les filons de
charbon Carbon et Thompson et, par conséquent, il en est déduit (1) qu’il proviendrait de la partie supérieure du ma
gnétochron 31n et (2) que son âge serait de 67,6–68,0 Ma. De grands cératopsidés chasmosaurinés tels que le Tricera
tops et le Torosaurus n’ont pas été décrits auparavant dans la Formation de Horseshoe Canyon, le magnétochron 31n
ou la sous-zone à miospores du M. gibbus. Ainsi, l’Eotriceratops est nettement plus vieux que tout autre cératopsidé
du groupe des Triceratops et la découverte d’E. xerinsularis contribue à combler une lacune biostratigraphique entre les
chasmosaurinés du Maastrichtien précoce et du Maastrichtien tardif.
[Traduit par la Rédaction]
Can. J. Earth Sci. 44: 1243–1265 (2007) doi:10.1139/E07-011 © 2007 NRC Canada
Received 13 November 2006. Accepted 20 March 2007. Published on the NRC Research Press Web site at http://cjes.nrc.ca on
10 October 2007.
Paper handled by Associate Editor H.-D. Sues.
Canadian Museum of Nature, P.O. Box 3443, STN “D”, Ottawa, ON K1P 6P4, Canada.
D.B. Brinkman, D.A. Eberth, and D.R. Braman. Royal Tyrrell Museum of Palaeontology, P.O. Box 7500, Drumheller,
AB T0J 0Y0, Canada.
Corresponding author (e-mail: firstname.lastname@example.org).
Southern Alberta is famous for its rich record of upper
Cretaceous dinosaurian and other fossil vertebrates (Ryan
and Russell 2001; Currie and Koppelhus 2005). Ceratopsids,
or horned dinosaurs, are one of the best represented dinosaur
groups in the region—and one of the last non-avian dinosaur
groups to go extinct (Russell 1967; Weishampel et al. 2004).
In western Canada, ceratopsid remains are abundant and
well documented in the middle to upper Campanian Belly
River Group and Wapiti Formation, the upper Campanian to
lower Maastrichtian lower Horseshoe Canyon Formation,
and the upper Maastrichtian Scollard and Frenchman forma
tions (equivalent to the Hell Creek and Lance formations of
the USA). However, until now, no ceratopsid remains have
been documented in the upper one-quarter of the Horseshoe
Canyon Formation and its equivalent deposits in neighboring
areas, a peculiar pattern that contrasts greatly with the rich
occurrence of other dinosaurs in this interval (Eberth et al.
2001). This gap in ceratopsid biostratigraphic data from
western Canada is unfortunate in that it apparently coincides
with a time of major faunal transition between the so-called
“Edmontonian” and “Lancian” dinosaur assemblages (sensu
Russell 1975; Sullivan and Lucas 2003, 2006).
In the 2001 field season, a team from the Royal Tyrrell
Museum of Palaeontology and Canadian Museum of Nature
began a systematic exploration for vertebrate fossils in the up-
per half of the Horseshoe Canyon Formation, with a particular
focus on the vicinity of Dry Island Buffalo Jump Provincial
Park, about 70 km northwest of Drumheller , Alberta (Fig. 1).
During that first season, Mr. Glen Guthrie, the camp cook,
found an incomplete skeleton of a v ery large ceratopsid near
the western bank of the Red Deer River within the park.
ments of the skeleton were not fully articulated, but were as-
sociated across an area of 3 m
(Fig. 2). Because no duplicate
skeletal elements were found, we interpret these remains as
pertaining to a single individual. This specimen represents the
first discovery of associated-to-articulated and identifiable (to
the species level) dinosaur skeletal remains from the top 20–
25mofthe-290 m thick Horseshoe Canyon F ormation
(Fig. 3); it is the first occurrence of identifiable ceratopsid re
mains in the top quarter of the formation. The large size and
initial morphological assessment suggested that the specimen
was a chasmosaurine that might be closely related to
Triceratops, but its stratigraphic position, well below the
Triceratops-producing beds of the Scollard Formation, indi
cated that the specimen was significantly older than any pre
viously identified Triceratops remains from Alberta or from
formations in Montana or Wyoming that are stratigraphically
equivalent to the Scollard Formation. In this paper, we de
scribe this specimen and discuss its systematic status, phylo
genetic relationships, and stratigraphic significance.
Geological setting, stratigraphy, and age
The Horseshoe Canyon Formation is the lowest formation
of the Edmonton Group, and is overlain successively by the
Whitemud, Battle, and Scollard formations (Fig. 3). Five in
formal subdivisions (units) have been described within the
Horseshoe Canyon Formation based on the presence or ab
sence of coal and stratigraphic architecture (Eberth and
O’Connell 1995; Eberth 2004). In the vicinity of Dry Island
Buffalo Jump Provincial Park, outcrops of the Horseshoe
Canyon Formation include, in ascending order, unit 3 (an in
terval of stacked sandstones), unit 4 (a mudstone dominated,
non-coaly interval), and unit 5 (a coal-rich and organic
shale-rich interval). Combined units 3 and 4, and unit 5, are
broadly similar to Hamblin’s (2004) Tolman and Carbon
tongues of the Horseshoe Canyon Formation, respectively,
but differ importantly in their contacts and inferred origins.
Unit 5 is largely equivalent to the Carbon–Thompson coal
zone of McCabe et al. (1989).
The new ceratopsid was collected from the middle of unit
5, 13.5 m above the base of the unit (Fig. 4). Unit 5 forms
the uppermost 20–25 m of the formation and is characterized
by the presence of coal seams that have been assigned to two
thin, but laterally extensive, coal-producing intervals or
zones: the lower, Carbon coal zone (seam 11 of Gibson
(1977)) and the upper, Thompson coal zone (seam 12 of
Gibson (1977)). The specimen occurs 9 m above the base of
the Carbon coal zone and 6 m below the lowest coal in the
Thompson zone (Fig. 4). Accordingly, the specimen occurs
between both coal zones and is not assignable with any con-
fidence to either.
Based on Srivastava (1970) and results from ongoing
palynological study of the Horseshoe Canyon Formation
(D.R. Braman, unpublished data, 2005), most of unit 5 (in-
cluding the specimen locality) can be assigned to the
Mancicorpus gibbus miospore subzone of the Scollardia
trapaformis miospore Zone. This biozone is distinctive, can
be traced across the Canadian portion of the Western Interior
Basin (Braman and Sweet 1999), and correlates with
magnetochrons 31n and 30r (Lerbekmo and Braman 2002).
The lithostratigraphic interval, including and bounded by the
Carbon and Thompson coal zones (including the specimen
locality), has been assigned to the upper portion of
magnetochron 31n by Lerbekmo (1985, fig. 4) and
Lerbekmo and Coulter (1985, figs. 3, 17); we accept that
chronostratigraphic assignment here. Ogg et al. (2004,
fig. 19.1) place the 31n magnetochron within the upper
Maastrichtian substage and assign it an age of 67.6–68.0 Ma.
We tentatively accept these assignments until the lower and
upper Maastrichtian substage boundary is formalized and
more radioisotopic dates are available from unit 5.
Large chasmosaurine ceratopsids, such as Triceratops and
Torosaurus, have not previously been described from the
Horseshoe Canyon Formation or from magnetochron 31n or
the M. gibbus miospore subzone, but are known from above
these intervals. For example, the Frenchman Formation of
southern Saskatchewan yields the late Maastrichtian
ceratopsids Triceratops and Torosaurus, but is younger than
unit 5 in having been deposited during magnetochron 29r
(Lerbekmo 1985) and containing a palynological assemblage
© 2007 NRC Canada
1244 Can. J. Earth Sci. Vol. 44, 2007
Barnum Brown, the famous fossil collector working for the American Museum of Natural History, apparently reported this specimen in his
1910 field notes as one of the team’s first discoveries during their first field season in Alberta. However, he chose not to collect the speci
men because it was considered low quality relative to other specimens in the area, including an assemblage of large theropod skeletal ele
ments in what is now known as the Albertosaurus bonebed.
that is assignable to the Wodehouseia spinata miospore
Zone (Braman and Sweet 1999). Similarly, the Lance For-
mation of Wyoming, and the Hell Creek Formation of
Montana and North Dakota also yield late Maastrichtian
ceratopsids (Triceratops, Torosaurus, Diceratops) but are
regarded as younger than the Horseshoe Canyon Formation
because they yield Aquilapollenites bertillonites and
Wodehouseia spinata (Norton and Hall 1969; Leffingwell
1971; Farabee and Canright 1986; Nichols 2002;
D.R. Braman, unpublished data, 2005). Paleomagnetic data
also support a younger age assignment for the Lance and
Hell Creek formations, and suggest that their oldest non-
marine portions are largely limited to magnetochron 30n
(Hicks et al. 2002; Lund et al. 2002), but may be as old as
30r (Hicks et al. 1999).
In summary, all available stratigraphic data indicate that
the new chasmosaurine ceratopsid is (1) younger than any
ceratopsid previously collected from the Horseshoe Canyon
Formation and (2) older than any previously known large
chasmosaurine collected from the Triceratops-producing
beds of the Scollard, Frenchman, Hell Creek, and Lance for
mations. Our review of magnetostratigraphic and palynological
evidence indicates that this specimen fills a significant bio
stratigraphic gap between chasmosaurines of the so-called
Edmontonian and Lancian dinosaur assemblages.
AMNH, American Museum of Natural History, New
York, N.Y.; ANSP, Academy of Natural Sciences, Philadel
phia, Pa.; CM, Carnegie Museum of Natural History, Pitts-
burgh, Pa.; DMNH, Denver Museum of Natural History,
Denver, Colo.; LACM, Natural History Museum of Los An-
geles County, Los Angeles, Calif.; MOR, Museum of the
Rockies, Bozeman, Mont.; NMC, Canadian Museum of Na
ture (formerly National Museum of Canada), Ottawa, Ont.;
RTMP, Royal Tyrrell Museum of Palaeontology,
Drumheller, Alta.; SDSM, South Dakota State School of
Mines, Rapid City, S.D.; USNM, National Museum of Natu
ral History (Smithsonian Institution), Washington, D.C. (for
merly the United States National Museum); YPM, Yale
Peabody Museum of Natural History, New Haven, Conn.
Ornithischia Seeley, 1888
Ceratopsia Marsh, 1890
Ceratopsidae Marsh, 1888
Eotriceratops gen. nov.
DIAGNOSIS: As for the type and only known species.
TYPE SPECIES: Eotriceratops xerinsularis gen. et sp. nov.
HOLOTYPE: RTMP 2002.57.7, an incomplete and disarticulated
skeleton, including right rostral, both premaxillae, both
maxillae, left supraorbital horn core with lacrimal, prefrontal,
frontal, postorbital and jugal, left epijugal, right
quadratojugal, left quadrate, partial parietal, left squamosal
frill, braincase, syncervical, cervicals 4 and 5, a string of
© 2007 NRC Canada
Wu et al. 1245
Fig. 1. Locality map for Eotriceratops insularis. The specimen locality is indicated by an asterisk in the northeast portion of Dry Island
Buffalo Jump Provincial Park. More detailed locality information is available from the Royal Tyrrell Museum, Collections Section. Upper
right inset shows field area in relation to the Province of Alberta.
vertebrae including posterior cervicals and anterior dorsals,
some ribs and fragments of ossified ligaments.
LOCALITY AND HORIZON: Within Dry Island Buffalo Jump Pro
vincial Park (northeastern quarter), southern Alberta; be
tween coal seams 11 and 12, in the upper 20 m of the
Horseshoe Canyon Formation; late Maastrichtian age.
ETYMOLOGY: Generic name Eotriceratops implies that it is an
early member of the “Triceratops” group; specific name
xerinsularis refers to Dry Island Buffalo Jump Provincial
Park, where the specimen was collected.
DIAGNOSIS: A large chasmosaurine ceratopsid with a skull
length of about 3 m, measured from the tip of the snout to
the posterior edge of frill. It differs from other
chasmosaurines in having the following combination of de
rived features: (1) tall lamina-like narial process of
premaxilla lacking any fossa or recess and having its dorsal
margin well above ventral border of interpremaxillary
fenestra; (2) epoccipitals of squamosal frill extremely elon
gate, spindle-shaped, and contacting each other; (3) a well-
demarcated elliptical depression or fossa on anteroventral
surface of squamosal frill; (4) a deep, slightly oblique trace
for blood vessels meeting a transverse trace ventrally on an
terior surface of nasal horn; and (5) sharply conical epijugal
with a pronounced proximoposterior process and separate
fossa-like articular facets for jugal and quadratojugal.
The specimen is preserved in carbonaceous shale. This
structurally weak matrix allowed for crushing and distortion
of bones during burial, resulting in a substantially flattened
skeleton. The most striking evidence for distortion is visible
in the occipital condyle and supraorbital horn core, which
show a significant amount of compression.
The rostral is well preserved, although the posterior tip of
the left ventral ramus and the posterior half of the right ven
tral ramus are missing (Fig. 5). Its left lateral surface is
© 2007 NRC Canada
1246 Can. J. Earth Sci. Vol. 44, 2007
Fig. 2. Spatial arrangement of preserved elements of Eotriceratops xerinsularis showing the distorted preservational state of the specimen.
c4, c5, and c6, cervical vertebrae 4, 5, and 6, respectively; dp, diapophysis of vertebrae; ej; epijugal; j, jugal; lm, left maxilla; lpm, left
premaxilla; lro, left rostrum; nsv, neural spine of vertebrae; oh, orbital horn core; ol, ossified ligament; p, parietal; po, postorbital;
q, quadrate; rb, rib; rm, right maxilla; rpm, right premaxilla; sq, squamosal; syc, syncervical vertebrae.
clearly convex, but its right lateral surface is much less so
because of preservational distortion. Both lateral surfaces
bear elongate pits or irregular grooves of different lengths.
The rostral is large and roughly boomerang-shaped in lateral
view, with elongate dorsal and ventral rami as in many
chasmosaurines. The dorsal ramus is dorsally convex and ta
pers to a point posteriorly. It extends posteriorly along the
dorsal edge of the premaxilla and sheathes the anterior three
quarters of that bone, as suggested by an articular facet on
the premaxilla (Figs. 6A–6D). The ventral ramus is shorter
than the dorsal ramus and slightly concave ventrally. The
ventral ramus may have tapered to a point in life. It differs
from the dorsal ramus in that the articular facet for the
premaxilla is dorsoventrally much broader on the lateral side
than on the medial side (Figs. 5C, 5D). The anterior tip of
the rostral bends slightly downward to form a short pointed
beak. The palatal surface of the rostral is not visible because
its two rami were strongly compressed toward the midline.
The left premaxilla is much more complete than the right,
but the posterior end of its posteroventral process (prong) is
missing. In lateral view (Figs. 6A, 6C), the premaxilla has a
triangular anterior portion with dorsal and ventral margins
that form an angle of about 75°. Both dorsal and ventral
margins are convexo-concave, but much more so in the case
of the ventral margin. The premaxillary septum anterior to
© 2007 NRC Canada
Wu et al. 1247
Fig. 3. Stratigraphy of Eotriceratops insularis. Specimen occurs between the Carbon and Thompson coal zones (seams 11 and 12, Gibson
1977) in the middle of unit 5 (Eberth 2004), which occurs in the uppermost 20 m of the Horseshoe Canyon Formation. Biostratigraphic,
magnetostratigraphic, and absolute age assignments are drawn from Srivista va (1970), Lerbekmo and Braman (2002), Lerbekmo (1985),
Lerbekmo and Coulter (1985), and Ogg et al. (2004). Assignment to the upper Maastrichtian (see text) is compatible with Ogg et al.
(2004); however , because the upper and lower Maastrichtian boundary is not formally fixed, we regard this assignment as tentative and
not yet clearly defined for the Alberta Basin (cf. Braman and Sweet 1999). Numbered horizontal bars indicate the stratigraphic occurrence
of coal seams recognized by Gibson (1977) and other authors. Fm., Formation; B. Fm., Battle Formation; W. Fm., Whitemud Formation.
the external naris is moderately elongate, and bears a fossa
(premaxillary fossa) that is perforated by a large and oval
fenestra (interpremaxillary fenestra). The premaxillary fossa
appears shallow, although this may have been a result of
mediolateral compression during fossilization. A depression
or fossa is present anteroventral to the interpremaxillary
fenestra. This fossa is also shallow as a result of the flatten
ing during fossilization. An opening within the fossa has an
irregular shape but may be an artifact of preservation (its mar
gins are broken). A semicircular process (interpremaxillary
process) projects from the anteroventral border of the
premaxillary fossa into the interpremaxillary fenestra. In
life, this process was flat and contacted its opposite. A com
parable process is present in many specimens of Triceratops
(USNM 4928, CM 1221, YPM 1822, YPM 1883, LACM
7207), Diceratops (USNM 2412), and Pentaceratops
(AMNH 1624, AMNH 6325; Lehman 1998), although it is
often triangular in shape, weakly developed, and sometimes
more posteriorly positioned in Pentaceratops. The narial
strut along the posterior margin of the premaxillary septum
is thickened but narrow, with a straight posterior edge form
ing the anterior border of the external naris. A tall lamina-
© 2007 NRC Canada
1248 Can. J. Earth Sci. Vol. 44, 2007
Fig. 4. Location, stratigraphic position, and facies association of Eotriceratops xerinsularis. (A) Specimen occurs between the Carbon
and Thompson coal zones, 9 m above the base of the Carbon zone (a double seam) and 13.5 m above the base of unit 5 (here marked
by the base of a fine-to-medium grained sandstone body). (B) Specimen occurs near the base of a broadly lenticular (tens of metres
wide), dark red-brown, silty, carbonaceous shale that is up to 50 cm thick. Carbonaceous shale is sharp based to erosional, and exhibits
patchy, irregular ironstone nodules, and local sulfur staining.
like narial process is present posteroventral to the strut. This
process is nearly complete except for its posterior edge. Its
lateral surface is slightly concave but does not have a deep
invaginated fossa or recess as seen in Triceratops (Forster
1990), Torosaurus (MOR 1122), and Diceratops (USNM
2412). The narial process is 120% taller than the narial strut
(measured along arrows in Fig. 6C). The dorsal margin of
the narial process is positioned well above the ventral border
of the interpremaxillary fenestra. The narial process is simi
larly tall in some specimens of Triceratops (USNM 4928,
USNM 1201, LACM 7207, AMNH 5116, YPM 1821) but,
as discussed later in the text, never reaches the proportions
seen in Eotriceratops. Ventral to the narial process, the
premaxilla is thickened laterally and bears an articular facet
for the maxilla along its ventrolateral surface. Dorsal to the
premaxillary fossa, the premaxilla is very narrow but thick.
Posterodorsal to the narial strut, the premaxilla forms a short
but sharp triangular nasal process. As indicated by a facet,
the lateral surface of the nasal process was entirely over
lapped by the nasal in life, except for a narrow area along
the ventral margin. There is no evidence as to whether or not
the nasal process extended up into the anterior base of the
nasal horn core.
In medial view (Figs. 6B, 6D), the palatal portion of the
premaxilla appears very narrow transversely because of
mediolateral compression during preservation. As a whole,
the medial surface is flat, with facets for the opposite
premaxilla, the rostral, and the maxilla. The facet for the op
posite premaxilla is extensive, occupying most of the medial
surface including the narial strut. The facet for the ventral
ramus of the rostral is narrow, running along the ventral
edge, from the anterior tip to a midpoint along the preserved
portion of the bone. The facet for the maxilla is restricted to
the area ventromedial to the lamina-shaped narial process. A
deep groove in the ventral portion of the maxillary facet may
represent a trace for blood vessels and nerves. Relationships
with other bones along the posterior border of the external
naris are unknown because of the damage to the posterior
© 2007 NRC Canada
Wu et al. 1249
Fig. 5. Left rostrum of Eotriceratops xerinsularis in lateral (A, B) and medial (C, D) views. fpm, facet for premaxilla; lv, lateral por
tion of ventral ramus; mv, medial portion of ventral ramus.
© 2007 NRC Canada
1250 Can. J. Earth Sci. Vol. 44, 2007
Fig. 6. Some skull elements of Eotriceratops xerinsularis. (A, C) Left premaxilla in lateral views. (B, D) Left premaxilla in medial
views. (E, F) Nasal horn core in slightly different anterior views. (G, H) Right quadratojugal in lateral and medial views, respectively.
avf, fossa or recess anteroventral to the interpremaxillary fenestra; bnp, dorsoventral breadth of narial process; bnst, dorsoventral
breadth of narial strut; fbv, foramen for blood vessels and nerves; fj, facet for jugal; fm, facet for maxilla; fn, facet for nasal; fq, facet
for quadrate; fro, facet for rostrum; ipf, interpremaxillary fenestra; ipp, interpremaxillary process; np, narial process; nst, narial strut;
otbv, oblique trace for blood vessels; pmf, premaxillary fossa; ppm, palatal portion of maxilla; pvp, posteroventral process;
qjg, quadratojugal groove; sr, swollen ridge; tbv, trace for blood vessels and nerves; ttbv, transverse trace for blood vessels.
portion of the posteroventral prong. Two foramina open into
canals for blood vessels and nerves on the dorsolateral mar
gin of the palatal portion. The foramina are close to each
other and positioned just posteroventral to the inter
premaxillary process. The canals exit through foramina on
the ventral surface of the palatal portion, but are farther
apart from one another. The posteriormost foramen opens
into a groove that continues ventrally and slightly anteriorly
to the edge of the bone.
Nasal horn core
The nasal horn core is nearly complete (Figs. 6E, 6F). It is
longer than wide, and broader anteriorly than posteriorly. Its
upper portion is laterally compressed. An obliquely oriented
groove and a slightly arched, transversely oriented groove
are present on the anterior surface of the horn core and are
probably vascular traces. Each groove is incompletely pre
served. A transverse groove, which extends along the base of
the horn core, is also present in Triceratops (Forster 1990).
The oblique groove extends dorsally and laterally, and then
turns onto the left lateral surface where it gradually fades
away. It was probably connected ventrally to the transverse
groove, although the area of inferred contact is not pre
served. As described earlier in the text, the nasal process of
the premaxilla is short, and its lateral surface is entirely
overlapped by the nasal, except for the ventral margin of the
process. This is comparable to the situation seen in
Triceratops (CM 1221, YPM 1822, USNM 4928, DMNH
48617), Torosaurus (ANSP 15192), and Diceratops (sensu
Forster 1990, 1996a (USNM 2412)), indicating that, as in
those genera, the nasal horn core is anteriorly positioned just
dorsal to the anterior edge of the external naris.
The left maxilla is almost complete with only its
posteroventral tip missing (Fig. 7). It is a thick and roughly
triangular bone, as in other chasmosaurines. It has a straight
anterodorsal edge, a slightly incurved ventral (dental) edge,
and a strongly convexo-concave posterodorsal edge. In lat
eral view (Figs. 7A, 7C), its surface is uneven. Articular fac
ets suggest that its sloping anterodorsal edge overlaps the
premaxilla but underlies the nasal. The suture between the
maxilla and premaxilla is slightly longer than the maxillary–
nasal suture. The convexo-concave posterodorsal edge of the
maxilla can be divided into three portions: a greatly incurved
dorsal portion, a vertical middle portion, and a strongly
oblique ventral portion. Articular facets on the dorsal portion
indicate that the maxilla overlaps the lacrimal and widely
underlies the jugal. The deep notch between the lacrimal
facet and the jugal facet is the part of the antorbital fenestra.
This fenestra is interpreted to have been oval in shape, with
an anteroventrally directed long axis. The articular facet for
the jugal indicates that the latter may have been excluded
from the fenestra. The dorsal portion of the posterior edge of
the maxilla overhangs the top of the middle portion. The
middle portion bears a thin free edge, forming a deep
embayment between the dorsal and ventral portions. The
ventral portion of the posterior edge of the maxilla runs
along the posterodorsal margin of the alveolar process and
bears an extensive facet for the ectopterygoid and a smaller
facet for the pterygoid. The lateral surface of the maxilla,
anterior to the posterior embayment, is concave and contains
a series of foramina. The concave area becomes narrow and
shallow anteriorly. Ventral to the concave area, the surface
of the maxilla is smooth.
In medial view (Figs 7B, 7D), the surface of the maxilla is
concave, especially dorsal to the level of the secondary pal
ate. A deep triangular fossa on the upper part extends anteri
orly into the bone. This fossa is referred to here as the
maxillary sinus, and may be equivalent to the vacuity be
tween the temporal fossa and mouth (Hatcher et al. 1907).
The maxilla is thicker anterior to the maxillary sinus. Sutural
facets for the palatine are located dorsal and posteroventral
to the maxillary sinus. The palatine facet extends
posteroventrally onto the alveolar process, meeting the facet
for the ectopterygoid laterally and the facet for the pterygoid
posteriorly. Anteriorly, the articular facet for the
posteroventral prong of the premaxilla is deep and groove-
like. The facets for the palatal portion of the premaxilla and
the vomer are not preserved because much of the palatal part
of the maxilla anterior to the choana is missing. The lower
part of the maxilla, ventral to the level of the secondary pal
ate and dorsal to the tooth row, is divided into a dorsal and a
ventral portion by a longitudinally elongate, step-like struc
ture. The dorsal portion slightly overhangs the ventral and
bears a series of foramina along the ventral side of the over-
The maxilla has 35 tooth rows preserved, and there may
have been one or two more positioned posteriorly. The pre-
served dental margin of the maxilla is 60 cm long, and the
anteriormost 5 cm of the margin are toothless.
Supraorbital horn core and circumorbital elements
The left supraorbital horn core is nearly complete, miss-
ing only its distal tip. Adjacent bones, especially those
around the orbit, are present, although not all are in articula-
tion. The lacrimal is indistinguishably fused with the
prefrontal. It is rough and thickened where it forms the rim
of the orbit (Figs. 8A, 8B). The lacrimal–nasal suture and
the articular facet for the maxilla are not preserved. An artic
ular facet on the jugal indicates that the jugal was over
lapped by the lacrimal.
The prefrontal is fused with the postorbital posterodorsally
and the frontal anterodorsally. It is strongly thickened and
very rough, forming a rugose swollen rim around the
anterodorsal border of the orbit (Fig. 8B). The prefrontal rim
of the orbit is much broader than the lacrimal rim.
Much of the left jugal is preserved, and is slightly dis
placed beneath the postorbital (Fig. 8). The jugal is T-shaped
as in other chasmosaurines. In lateral view, the jugal–
postorbital suture is probably V-shaped as in some speci
mens of Triceratops (e.g., YPM 1821). Anteriorly, the jugal
overlaps the postorbital, but much of the posterior portion of
the jugal may have been overlapped by the postorbital. The
jugal forms the ventral rim of the orbit, which is rough but
not very swollen. The anteroventral portion of the jugal,
which typically overlaps the maxilla, and the posterodorsal
portion, which typically meets the squamosal, are incom
plete. The anterior margin of the jugal flange is complete al
though cracked. The posterior edge of the jugal flange
ventral to the jugal–squamosal suture and dorsal to the
jugal–quadratojugal facet is well-preserved. This edge forms
© 2007 NRC Canada
Wu et al. 1251
the anterior border of the infratemporal fenestra. The distal
end of the jugal become narrow but does not taper to a
point. In medial view, the surface of the jugal is smooth. Its
articular surface for the postorbital is obscured, but the facet
for the quadratojugal is clear along the posteroventral mar
gin of the flared flange (Fig. 8B).
The left epijugal, which is preserved together with the
left quadrate, is almost complete. In lateral outline it is trian
gular and taller than wide, with a straight proximal margin
that extends posteriorly onto a proximoposterior process that
overhangs the main body of the triangle (Figs. 9A–9C). In
life, it may have been roughly cone shaped and oval in
cross-section. The lateral surface of the epijugal is orna
mented by vertically oriented grooves and ridges. In medial
view, the lower half of the epijugal is flat, but its upper half
is occupied by a triangular, cup-like fossa that extends in
wards and downwards into the bone (Fig. 9B). In proximal
view, the surface of the epijugal is entirely occupied by a
trough-like fossa that expands onto the proximoposterior
process (Fig. 9C). These two fossae are well separated. It is
likely that the fossa on the proximal surface capped the dis
tal end of the jugal flange, whereas the fossa on the
proximomedial surface received the anterodistal end of the
quadratojugal. These two fossa-like articular facets are not
present in other chasmosaurines.
Much of the postorbital is preserved, although the areas
where it meets the squamosal and parietal are missing
(Fig. 8). As described earlier, its suture with the jugal is
clear but its suture with the prefrontal is obscured by fusion
of the two bones. The supraorbital horn core was laterally
compressed during preservation. The preserved portion is
about 77 cm long from the dorsal edge of the orbit to its tip,
and originally may have been 80 cm long. The body of the
horn core curves anteriorly. The external surface of the horn
core exhibits many vertically oriented, fine grooves and
ridges. Posterior to the orbit, the grooves—some of which
are deep and broad—extend anteroposteriorly. The
postorbital forms the smooth posterior border of the orbit.
Three shallow pits dorsal to the orbit are interpreted as bite
In medial view, breakage near the base of the supraorbital
horn core has exposed internal sinuses (Fig. 8C). The frontal
sinus is surrounded by the postorbital and internal portion of
the frontal. As in Triceratops (Lull 1933; Forster 1996b), the
sinus is divided into the anterior and posterior portions by a
septum (transverse buttress). The anterior portion is more
complete than the posterior portion. It is longer than high,
and is enclosed dorsally by the postorbital and ventrally by
the frontal. Bony struts or lamina from the postorbital par
tially subdivided the anterior portion of the frontal sinus. A
canal just anterior to the base of the transverse buttress ap
pears to perforate the floor of the anterior portion of the
frontal sinus to enter the brain cavity. The shape and size of
the posterior portion of the frontal sinus are unknown be
cause only the lateral wall and part of the floor are pre
served. No struts or lamina are present. The cornual sinus is
© 2007 NRC Canada
1252 Can. J. Earth Sci. Vol. 44, 2007
Fig. 7. Left maxilla of Eotriceratops xerinsularis in lateral (A, C) and medial (B, D) views. af, antorbital fenestra; fec, facet for
ectopterygoid; fj, facet for jugal; fl, facet for lacrimal; fn, facet for nasal; fpm, facet for premaxilla; fpt; facet for pterygoid;
ms, maxillary sinus; pmx, palatal portion of maxilla.
visible within the posterior half of the frontal sinus. It is
small but enters deeply into the supraorbital horn core
anterodorsally, as in Triceratops (Forster 1996b). The ven-
tral surface of the preserved part of the frontal, which roofs
the brain cavity and olfactory tube, is smooth but concave.
Much of the right quadratojugal is preserved. It is broader
ventrally than dorsally, and it has a thickened and strongly
convex ventral edge and a dorsal edge that slopes anteriorly
(Figs. 6G, 6H). The lateral surface of the bone is arched
along a swollen ridge (here termed the lateral quadratojugal
ridge) that runs from a midpoint on the ventral margin of the
bone to its posterodorsal tip. The lateral quadratojugal ridge
forms the posterior edge of the facet that receives the over
lapping jugal as in other ceratopsids. In Eotriceratops, the
dorsal most part of the quadratojugal is very narrowly ex
posed and may have not bordered the infratemporal fenestra
dorsally, a situation opposite that of Triceratops. Posterior to
the ridge, the bone surface is smooth and concave. Ventrally,
the thickened edge of the quadratojugal overlaps the lateral
side of the quadrate. A convex ventral prominence of the
quadratojugal inserts into the fossa on the proximomedial
surface of the epijugal. The medial surface of the
quadratojugal is flat. The full extent of the articular facet for
the quadrate is not clear. An area marked by many fine
ridges on the anteroventral portion of the inner surface is in
terpreted as one portion of this facet, and a narrow rough
surface along the posterior edge of the bone may represent
another portion. These two surfaces meet ventrally. A
groove or trough, presumably for blood vessels and nerves,
is present within the quadrate facet.
The left quadrate is well preserved although its pterygoid
ramus and dorsal edge are missing (Figs. 9D–9F). This bone
is similar in shape to that of Triceratops. The preserved part
of the bone is 42 cm tall and 15 cm wide (measured along
the long axis of the mandibular condyle). It is broad but very
thin dorsally and narrow but thickened ventrally, where it
forms the jaw joint. This joint is subdivided into medial and
lateral condyles by a shallow trough. The lateral condyle
is slightly larger than the medial, and because the
quadratojugal also contributes to the lateral condyle, the lat
eral portion of the jaw joint would have been more massive
than the medial. In anterior view, the surface of the quadrate
is smooth and concave (Fig. 9D). A depression or fossa that
may be exaggerated by preservation is present on the ventral
quarter, close to the mandibular condyles. Although the
pterygoid ramus is not preserved, the base of the ramus, as
seen medially, suggests a fan shape, as in other ceratopsids.
In posterior view, the surface of the quadrate is smooth and
convex. A depressed area on the ventral part of the posterior
surface (Fig. 9E) is an artifact of preservation. The lateral
margin of the quadrate is thickened to form a ridge, and a
low round process is present midway along this margin. The
articular facet for the quadratojugal, seen in lateral view, is
very rough. The quadrate thins dorsal to this facet (Fig. 9F).
© 2007 NRC Canada
Wu et al. 1253
Fig. 8. Supraorbital horn core and bones around orbit of Eotriceratops xerinsularis in lateral (A, B) and medial views. asf, anterior
frontal sinus; bift, border of infratemporal fenestra; bm, bite marks; cs, to conual sinus; f, frontal; fj, facet for jugal; fl, facet for
lacrimal; fmj, free margin of jugal flange; fqj, facet for quadratojugal; j, jugal; l, lacrimal; ob, orbit; pfs, posterior frontal sinus;
po, postorbital; prf, prefrontal; sfs, septum of frontal sinus; tc, trace of canal. Two arrows below the postorbital–jugal suture indicate
that the jugal was pushed upward underneath the postorbital during preservation.
Only the anterior part of the parietal is preserved (Figs. 10A,
10B). It is conv ex dorsally and shows no midline suture. Its
dorsal surface is rough, with fine traces for blood vessels. A
portion of the edge of the supratemporal fossa is preserv ed on
the anterolateral corner of the bone. The remaining edges of
the bone are broken. The ventral surface is not exposed. The
preserved portion of the parietal is thickest (2.3 cm) at the pos
terior midline. Laterally, it thins to about 1.3–1.5 cm, and is
much thinner than in T riceratops. A fragmentary bone pre
served close the parietal is identified as a fragment of the pari
etal frill (Figs. 10C, 10D) and is 3.1 cm thick.
The left squamosal is represented by the portion of the
bone that contributed to the frill. Much of the anteromedial
part and anterolateral corner of the squamosal is missing
(Figs. 10E–10G). The preserved portion of the squamosal is
113.8 cm long and 43.3 cm wide. Although no trace of a
fenestra is present, the element is incomplete, and the possi
bility of fenestration in its more anterior portion cannot be
excluded. Except for the anterior portion, the lateral margin
of the squamosal is nearly complete. This edge is convex
and not obviously scalloped as in most specimens of
Triceratops (e.g., YPM 1822). Four and one-half
epoccipitals are present, with the partial epoccipital being
the anteriormost. The posteriormost epoccipital is missing.
The sutural pattern, seen most clearly on the ventral surface,
indicates that the more posteriorly positioned epoccipitals
are more loosely attached to the frill and would have fused
to the frill later in life (Lehman 1990). The lengths of the
© 2007 NRC Canada
1254 Can. J. Earth Sci. Vol. 44, 2007
Fig. 9. Left epijugal and left quadrate of Eotriceratops xerinsularis. Left epijugal in lateral (A), medial (B), and dorsomedial
(C) views. Left quadrate in posterior (D), anterolateral (F, lower half), and anterior (E) views. epp, proximoposterior process of
epijugal; fj, facet for jugal; fo, fossa; fqj, facet for quadratojugal; prq, pterygoid ramus of quadrate.
© 2007 NRC Canada
Wu et al. 1255
Fig. 10. Frill fragments of Eotriceratops xerinsularis . (A, B) Anterior part of parietal frill in dorsal views. (C, D) A piece of parietal
frill in dorsal and cross-section view, respectively. (E–G) Partial squamosal frill in dorsal (E) and ventral (F, G (a reconstruction))
views. (H) lateral edge of squamosal. epo.1–epo.5, epoccipitals 1–5, respectively; fp, sutural facet for parietal frill; fsq, fossa on ventral
surface of squamosal frill; pstf, parietal portion of supratemporal fenestra; 1.1 cm, 1.3 cm, 1.5 cm... indicate, respectively, number of
centimetre thicknesses of bone in areas shown by arrows.
epoccipitals suggest that the posteriormost epoccipital would
have extended across the suture between the squamosal and
parietal. Thus, if the reconstruction of the anterior portion of
the squamosal is correct, the squamosal would have had five
and one-half epoccipitals (Fig. 10G). All epoccipitals are
slender and differ little in length. The fourth epoccipital,
which is the most complete, is 18.5 cm long. This is an ex
tremely elongated, crescent- or spindle-like element, with a
length to width ratio of about 10:1 when viewed ventrally.
As shown by the third and fourth epoccipitals, these ele
ments contacted one another in life, a feature that differs
from the arrangement of the epoccipitals in most
chasmosaurines. The squamosal epoccipitals most resemble
those of Torosaurus utahensis in shape (Sullivan et al. 2005)
but, as discussed later in the text, differ in shape and propor
tions. In dorsal view, the surface of the squamosal is rela
tively smooth, with some fine traces for blood vessels. In
ventral view, the surface is smooth and characterized by an
elliptical fossa that is incomplete anteriorly and located in
the anterior half of the bone. With its well-defined margin,
this fossa differs from the depression or concave area seen
in other chasmosaurines. Posterior to the fossa, the surface
of the squamosal is concave. The squamosal–parietal suture
is preserved along the posteromedial side of the squamosal.
The suture is not thickened to form a distinct ridge like that
of Torosaurus (ANSP 15192). As preserved, the frill differs
from the evenly thickened frill of Triceratops in being rela-
tively thin, having a thickness of 1.1 cm along the
posteromedial edge, 2.2 cm at the posterior tip, 2.3 cm along
anterolateral edge, 2.0 cm along the anteromedial edge, and
about 3 cm along the medial edge at the midpoint. However,
we cannot exclude the possibility that this variable thickness
was accentuated during preservation.
The braincase is badly crushed and the foramen magnum
has collapsed (Fig. 11). Some elements are missing, and the
preserved bones are difficult to distinguish from one another.
In dorsal view, much of the left side and the anterior portion
of the braincase are damaged. The right wall of the
braincase is displace towards the left side and has shifted
posteriorly to cover most of the occipital condyle (Fig. 11A).
The preserved part of the left wall may include the prootic
(anterior portion), the supraoccipital (mid-portion), and
large exoccipital (posterior portion). No foramen or fenestra
can be identified in dorsal view. Anteriorly, the keel-like
ventral ridge of the basisphenoid is visible because of the
missing anterior portion of the braincase. This ridge is in
complete anteriorly and dorsally. On the left side, a bone
fragment is also visible in ventral view, where it is separated
from the braincase by the matrix. It is likely that this frag
ment is not part of the braincase.
The occipital condyle was strongly compressed during
preservation and is preserved as a round plate-like structure
in lateral view (Figs. 11C, 11D). It is not known how much
of the exoccipital would have contributed to the formation of
the occipital condyle. The fan-like ventral process of the
basioccipital is partially visible in lateral view, and it has a
thickened and rough ventral edge. A foramen anterolateral to
the neck region of the basioccipital may be an exit for cra
nial nerve XII. Anterodorsally, the lateral surface of the
exoccipital–prootic part of the braincase bears a series of de
pressions. Posteroventrally, a pair of large processes (basal
tubera of the basisphenoid) are directed posteroventrally as
in other ceratopsids. These tuberosities, which are thickened
posteroventrally and become flattened anteriorly, extend fur
ther posteriorly than the fan-shaped ventral process of the
basioccipital. As seen in ventral view (Figs. 11E, 11F), the
basal tuberosities of the basisphenoid, together with the fan-
shaped ventral process of the basioccipital, were displaced
toward the left side. The posterior end of the right basal
tuberosity is missing. The two basal tuberosities converge
anteriorly toward the midline and merge with the keel-like
The syncervical is incomplete and distorted, with much of
its centrum damaged and the dorsal edge of its neural spine
missing (Figs. 12A, 12B). Although none of the sutures be
tween the individual vertebrae can be recognized, as a
whole, the syncervical resembles that of Triceratops.Two
intervertebral foramina are present, although they are filled
with matrix and were laterally flattened during preservation.
We identify the two foramina as those that occur between
the atlas and axis, and between the axis and cervical 3. Thus,
the syncervical would have been formed by the first three
cervical vertebrae plus a ring-like hypocentrum, as is typi-
cally the case in ceratopsians (Lull 1933). The anterior sur-
face of the centrum is deeply concave. Anterodorsal to the
anterior intervertebral foramen, the neural arch of the atlas is
broad and posterodorsally merges into the neural spine of
the syncervical. The neural spine is low anteriorly and be-
comes tall posteriorly. A thickened ridge posterodorsal to the
posterior intervertebral foramen extends dorsally and poste-
riorly. It is bordered ventrally by a shallow groove that most
probably represents the suture between the axis and the third
cervical. A slightly swollen area just anteroventral to the
intervertebral foramen on the right side may represent the
transverse process of the axis. The postzygapophyses face
ventrolaterally and extend only a little posterior to the
centrum. The posterior surface of the centrum is slightly
Cervical vertebrae 4 and 5
Cervical vertebrae 4 and 5 are preserved in articulation
and are closely associated with the syncervical (Fig. 2). Cer
vical 4 is almost complete, and both cervical vertebrae 4 and
5 are better preserved on the right side. Each cervical bears
the proximal portion of the corresponding rib on the right
side, although these are displaced anteriorly. The centra are
shallowly amphicoelous, laterally concave, and short, with
the length to height ratio of the fourth being about 1:2
(Figs. 12C, 12D). The parapophysis is located at the
anterodorsal edge of the centrum in cervical 4, but slightly
posterior and dorsal to the edge in cervical 5. The
diapophysis is well developed and is longer in cervical 5
than in cervical 4. This is also the case with the neural spine.
The complete prezygapophysis of cervical 4 faces medially
and dorsally, and extends anteriorly well beyond the
centrum. The ventral surfaces of both cervicals are incom
© 2007 NRC Canada
1256 Can. J. Earth Sci. Vol. 44, 2007
Posterior cervical and anterior dorsal vertebrae
An articulated series of vertebrae including posterior
cervicals and anterior dorsals is preserved close to cervicals
4 and 5 (Fig. 2). Based on the number of the diapophyses
and attached ribs, seven vertebrae, identified as the eighth to
fourteenth, are present in the series. Thus, if the cervical se
ries consists of nine vertebrae, these likely represent cervical
8 to dorsal 5 (Figs. 13A, 13B). Both the first and last verte
brae in this series are missing their centra and, except for the
third dorsal, they are largely covered by the attached ribs.
The centrum of the third dorsal is laterally concave. Its
parapophysis is located on the neural arch at the level of the
prezygapophysis, and its diapophyses is very pronounced
and widely separated from the parapophysis, corresponding
to the relative positions of the two heads of the attached rib.
Parapophyses of the first and second dorsals are hidden by
the attached ribs but, as indicated by the tall position and
massiveness of the diapophyses, are probably on the neural
arch as well. The relative position of the parapophyses and
diapophyses on the eighth and ninth cervical vertebrae is un
clear because of poor preservation, and those on the fourth
and fifth dorsal vertebrae are covered by ribs. The rib facet
of the diapophyses becomes anteroposteriorly elongate and
dorsoventrally narrow in the posterior vertebrae in this se
ries. The neural spine is anteroposteriorly broader in the dor
sal vertebrae than in the cervicals.
© 2007 NRC Canada
Wu et al. 1257
Fig. 11. Braincase of Eotriceratops xerinsularis in dorsal (A, B), right lateral (C, D) and ventral (E, F) views. As the occipital condyle
shows, the braincase was laterally compressed. bo, basioccipital; eo, exoccipital; btb, basal tuber of basisphenoind; kbs, ventral keel of
basisphenoid; pr, prootic; so, supraoccipital; vfbo, ventral fan of basioccipital; ?, a bone fragment not related to braincase; XII, cranial
Ribs and ossified ligaments
Except for two isolated and nearly complete ribs, all of
the preserved ribs are represented only by their proximal
ends and are preserved in articulation with the vertebrae.
Two rib heads on the right side of cervicals 4 and 5 are typi
cal for archosaurs in bearing three heads: a central one that
is free and extends anteriorly, a short but stout dorsal pro
cess that articulates with the diapophysis, and an elongate
ventral process that articulates with the parapophysis of the
vertebra (Figs. 12C, 12D). The central head of the rib asso
ciated with cervical 4 and the anterior tip of the central head
of the rib associated with cervical 5 are missing. When
complete, it is likely that these were as long as the dorsal
process. The series, including vertebrae 8 to 14 (cervicals 8
and 9, and dorsals 1 to 5), has at least eight rib heads at
tached on the right side (Figs. 13A, 13B). The rib heads as
sociated with cervicals 8 and 9 are fragmentary. Of the other
six rib heads, only that associated with dorsal 3 is nearly
complete. It has only two processes: a very short but broad
dorsal process, and a long and slender ventral process. The
two more posterior rib heads are similar in outline. The three
more anterior ribs have a similar dorsal process but their
© 2007 NRC Canada
1258 Can. J. Earth Sci. Vol. 44, 2007
Fig. 12. Anterior cervical vertebrae of Eotriceratops xerinsularis in right lateral views. (A, B) Syncervical. (C, D) cervical vertebrae 4
and 5. cr, cervical rib; dp, diapophysis of vertebrae; ivf, intervertebral foramen; nsv, neural spine of vertebrae; poz, postzygapophysis;
pp, parapophysis; prg, prezygapophysis; stv, sutural trace between vertebrae.
ventral processes are missing. A bar-like bone fragment
overlapping the rib heads 13 and 14 may represent the slen
der ventral process of rib 15. The two isolated ribs are
nearly complete although their distal ends and the posterior
margins of their proximal heads are damaged (Fig. 13C).
The better preserved of the two is 118.5 cm long, measured
from the knob-like dorsal process to the distal end. The
other is 116.5 cm long. The knob-like dorsal process for the
diapophysis is preserved in the longer rib, and the length of
the ventral process, which articulated with the parapophysis
of the vertebra, is similar to that of the third dorsal rib. On
the basis of the morphology of their heads and relatively
straight shafts, these two ribs are likely from a relatively an
terior position along the dorsal series. Mediolaterally com
pressed ossified ligaments run between the transverse
processes and neural spines in the articulated string of cervi
cal and dorsal vertebrae (Fig. 13A, 13B).
Comparison and discussion
Eotriceratops is a chasmosaurine because it shares with
other members of the group the synapomorphies of a large
rostral with deeply concave posterior margin and hypertro
phied dorsal and ventral rami, an anteriorly elongate
premaxillary septum, the presence of the thickened narial
strut along the posterior margin of the premaxillary septum,
the perforation of the premaxillary septum, the presence of a
narial process projecting into the external naris from the
posteroventral margin of the premaxillary septum, and the
presence of an anteriorly curved supraorbital horn core
(Dodson et al. 2004; also see later in the text). Among the
chasmosaurines, Eotriceratops has a closer relationship with
more crownward members than Chasmosaurus and
Pentaceratops on the basis of two synapomorphies: the
supraorbital ornamentation centered posterodorsally or pos
terior to the orbit; and the broad supracranial cavity (sinus),
underlying the supraorbital ornamentation and confluent
with the cornual sinus (Dodson et al. 2004; also see later in
the text). Therefore, in the following comparison, those
crown chasmosaurines that occur in horizons either lower
(Anchiceratops, Arrhinoceratops) or higher (Triceratops,
Diceratops, Torosaurus) than Eotriceratops, are empha
Although Eotriceratops is represented by an incomplete
specimen, it can be defined by a combination of derived fea
tures. Among these, the large, non-recessed narial process of
the premaxilla (Figs. 6A–6D) is unique for Eotriceratops.As
indicated by the arrows in Fig. 6C, its dorsoventral height is
greater than the height of the narial strut. Also, its dorsal
margin is much higher than the ventral border of the
interpremaxillary fenestra. In most other chasmosaurines, the
narial process is anteroposteriorly short and dorsoventrally
© 2007 NRC Canada
Wu et al. 1259
Fig. 13. Some vertebrae and ribs of Eotriceratops xerinsularis. cv, centrum of vertebra; dp, diapophysis of vertebrae; nsv, neural spine of
vertebrae; ol, ossified ligament; pp, parapophysis; r.1, r.2, r.5, and r .6, ribs 1, 2, 5, and 6, respectively, reserved in the vertebral series.
narrow, and in more derived taxa, such as Triceratops (CM
1221, YPM 1822, YMP 1823, DMNH 48617), Torosaurus
(ANSP 15192), and Diceratops (USNM 2412), it bears a
deep recess laterally. In some specimens of Triceratops
(e.g., USNM 4928, USNM1201, LACM 7207, AMNH 5116,
YPM 1821), the laterally recessed narial process also ap
pears to be dorsoventrally broad. However, the process is
still relatively dorsoventrally narrower than in Eotriceratops
when measured in the same way (Fig. 6C), and the dorsal
margin of the process is lower than the ventral border of the
interpremaxillary fenestra because of the small size of the
fenestra (e.g., YPM 1821; Hatcher et al. 1907, fig. 18). The
condition of this feature is unknown in Arrhinoceratops
(ROM 5135) because of poor preservation. In Anchiceratops
(TMP 83.01.01), the narial process is short compared with
the narial strut and the dorsal margin of the narial process is
at the level of the ventral margin of the interpremaxillary
Eotriceratops also differs from other chasmosaurines in
having extremely elongate, crescent- or spindle-shaped
epoccipitals on the squamosal (Fig. 10H). As mentioned ear
lier in the text, the epoccipitals are most similar to those of
Torosaurus utahensis (this taxon may represent a new
chasmosaurine (R.M. Sullivan, personal communication,
2005). However, they are still distinguishable in shape, size,
and relationship with the squamosal from the epoccipitals in
that taxon. In Eotriceratops, the squamosal epoccipitals have
a similar shape and size, and contact one another along the
margin of the frill. The slenderness of the epoccipitals is in-
dicated by the length/width ratio, which is approximately 10.
In Torosaurus utahensis (USNM 15583 = USNM 494473 of
Sullivan et al. (2005)), 13 disarticulated epoccipitals are pre-
served. Nine of the 13 are more triangular than elongate
crescent- or spindle-shaped, have a pronounced medial apex
(sensu Sullivan et al. 2005), and are relatively larger. The re-
maining four are small, narrow, and rod-like. Based on com
parisons with other specimens, we believe that the nine
elongate triangular epoccipitals belong to the squamosal and
the four rod-like epoccipitals belong to the parietal. The
most elongate of the squamosal epoccipitals is 14.5 cm long
and is estimated to have been -15.2 cm long. This element
has a length to width ratio of 4:2, and the ratio reaches about
4:4 in the most slender epoccipital of the series, although it
is relatively short. In a recently collected torosaur skull,
MOR 1122, the epoccipitals, which are in place, have a
length/width ratio of about 5 and do not contact one another.
In Torosaurus latus, no squamosal epoccipitals have been
described for the type specimen (YPM 1830) or referred
skull (ANSP 15192; Colbert and Bump 1947). Our examina
tion of ANSP 15192 reveals that the squamosal epoccipitals
are obscured because of fusion with the frill. The
epoccipitals of Anchiceratops differ in being strongly trian
gular and the epoccipitals of Arrhinoceratops differ in being
small. Further, the lateral margin of the squamosal in those
taxa is straight rather than convex as in Eotriceratops. Varia
tion is present in the shape of the squamosal epoccipitals in
chasmosaurines, and the shape of these bones is often ob
scured because of fusion with the frill in late stages of life.
However, we consider the elongate crescent or spindle shape
of epoccipitals on the squamosal to be one of the diagnostic
features of Eotriceratops because epoccipitals of this shape
are not seen in other ceratopsids we examined. Furthermore,
the presence of a contact between the epoccipitals in
Eotriceratops is also a rare phenomenon because, based on a
growth series of the squamosal frills of the chasmosaurine
Pentaceratops sternbergii (Lehman 1990, fig. 16.10), the
epoccipitals do not change in mutual relationships, remain
ing separate from one another through ontogeny among the
chasmosaurines we examined.
Eotriceratops has an elliptical fossa on the anterior por
tion of the ventral surface of the squamosal frill, a feature
that may be unique within chasmosaurines. This fossa is
bordered with thickened margins, so it cannot be an artifact
of preservation (Figs. 10F, 10G). The ventral surface of the
squamosal frill in other chasmosaurines we examined, as in
ceratopsids generally, is usually very smooth or has a broad,
shallow depression with no consistent shape and no clearly
demarcated borders. Such an ill-defined, shallow depression
is present in some specimens of Triceratops (YPM 1821,
1824) and Torosaurus (YPM 1831), and an oval depression
along the longitudinal axis is present in Arrhinoceratops
(Parks 1925, pl. I). Our examination of a specimen of
Arrhinoceratops (ROM 5135) reveals that this depression is
smoothly transitional with the concave ventral surface of the
frill and lacks the well-distinguished border seen in
Eotriceratops. Based primarily on the skulls of Torosaurus,
Lull (1908) interpreted these depressions as the site of the
insertion of the M. latissimus dorsi. We believe that different
shapes and different marginal conditions of the depression or
fossa reflect different bone–muscle relationships and
biomechanical differences associated with features such as
the size and weight of the skull; therefore, we consider the
structure of this fossa to be a diagnostic feature of
The presence of an obliquely dorsoventrally extending
vascular trace meeting a transversely arched vascular trace
ventrally on the anterior surface of the nasal horn core of
Eotriceratops is an unusual feature within chasmosaurines.
A transversely arched vascular trace on the anterior surface
of the nasal horn core is consistently present in Triceratops
(Forster 1990, 1996b); among the chasmosaurine specimens
we examined, a pattern of vascular traces similar to that of
Eotriceratops is seen only in USNM 1201, a specimen of
The epijugal differs from that of other chasmosaurines in
the combination of the following two structures: (1) a
sharply cone-shaped or dorsoventrally elongated, triangular
configuration with a pronounced proximoposterior process
in lateral view; and (2) separate fossa- or cup-like articular
facets on its proximal surface and its proximomedial surface
that cap the jugal and quadratojugal, respectively. A cone-
shaped epijugal with a triangular outline in lateral view is
present in Arrhinoceratops (Parks 1925) and Torosaurus
utahensis (Sullivan et al. 2005), but it is proximally broader
than deep in those taxa. Anchiceratops, as documented by
specimen TMP 83.1.1, is like Eotriceratops in having a
cone-shaped epijugal that is dorsoventrally elongate, but dif
fers in lacking a pronounced proximoposterior process. In
most other chasmosaurines, the epijugal shows an irregular
shape in lateral view, or a trihedral shape (see CM 1221,
YPM 1822, YPM 1821 for Triceratops; see TMP 81.19.175,
TMP 87.45.1, AMNH 5401, ROM 843 for Chasmosaurus;
© 2007 NRC Canada
1260 Can. J. Earth Sci. Vol. 44, 2007
see Lehman (1996, fig. 9) for Pentaceratops and
Torosaurus; see USNM 2412 for Diceratops). In other
chasmosaurines, it is typical for the epijugal to have a bipar
tite medial surface for articular contact with the jugal and
quadratojugal (e.g., Torosaurus utahensis (Sullivan et al.
2005, fig. 6.4)). However, none of the specimens of those
taxa we examined shows a facet for the quadratojugal in the
form of a cup-like fossa. Thus we interpret the presence of
this feature as diagnostic for Eotriceratops.
Phylogenetic relationships of Eotriceratops are estab
lished on the basis of a cladistic analysis that employs the
data matrix recently used by Dodson et al. (2004). Three ad
ditional characters of Eotriceratops were included (Appen
dix A) so that the complete data matrix consists of 76
characters and 16 taxa. Character 45 is slightly modified by
the addition of a state 2 (elongate spindle-shaped squamosal
epoccipitals). The data matrix is subjected to maximum par
simony analyses using PAUP* (Swofford 2002). These apply
100 branch and bound searches with input order of terminal
taxa randomized in each search. Characters are given uni
form weight and multistate characters are unordered. Nodal
support is assessed through calculation of bootstrap propor
tions based on 1 000 replicates using branch and bound al
gorithm in PAUP*.
The parsimony analysis produces six shortest trees, with a
tree length of 92, a CI of 0.880, a RI of 0.925, and a RC of
0.815. In all of the six trees, Eotriceratops is hypothesized
to be a chasmosaurine (Fig. 14). The monophyly of the
Chasmosaurinae is supported by 12 synapomorphies includ
ing five unequivocal characters. Among the 12
synapomorphies (based on tree 1), seven can be scored for
Eotriceratops. They include the large rostral with deeply
concave posterior margin, and hypertrophied dorsal and ven
tral rami (1); the anteriorly elongate premaxillary septum
(4); the presence of the thickened narial strut along the pos
terior margin of the premaxillary septum (5); the perforation
of the premaxillary septum (8); the presence of a narial pro
cess projecting into the external naris from the
posteroventral margin of the premaxillary septum (11); the
supraorbital horn core anteriorly curved (25); and the shape
of epoccipitals (45). Among these seven synapomorphies,
characters (1), (5), and (11) are unequivocal.
Eotriceratops is located successively more crownward in
phylogenetic position than the clade including
Chasmosaurus and Pentaceratops and the clade consisting
of Anchiceratops and Arrhinoceratops; it forms, together
with Triceratops, Diceratops, and Torosaurus,a
monophyletic group in all the six trees. This is supported by
five synapomorphies including two that are unequivocal
(character 9, the presence of a recess in ventral portion of
the premaxillary septum, and character 74, the nasal horn
core anteriorly positioned), and three that are equivocal
© 2007 NRC Canada
Wu et al. 1261
Fig. 14. Cladograms showing phylogenetic relationships of Eotriceratops. (A) Strict consensus tree computed from the six most parsi
monious trees (MPTs) recovered by PAUP* analysis of a modified version of the character–taxon matrix by Dodson et al. (2004). (B,
C) Alternative relationships of Eotriceratops with crowner taxa among the six most parsimonious trees. Nodal numbers indicate boot
strap proportions based on 1 000 replicates using the branch and brand algorithm in PAUP*.
(character 43, the epoccipital crossing squamosal-parietal
contact; character 75, the presence of transversely arched
vascular trace across the anterior surface of the nasal horn
core; and character 76, the presence of interpremaxillary
process). Within the group, Eotriceratops either forms a
sister-group relationship with Triceratops (Fig. 14B), which
is primarily supported by two unequivocal synapomorphies
(character 43, an epoccipital crossing the squamosal-parietal
contact, and character 75, the presence of a transversely
arched vascular trace on the anterior surface of the nasal
horn core), or is hypothesized to be the most basal taxon of
the group because of the small size of the recess along the
ventral aspect of the premaxillary septum (character 10) and
the lack of a recess on the lateral surface of the narial process
of the premaxilla (character 12, Fig. 14C). Thus the phylo
genetic relationships among Eotriceratops, Triceratops, and a
subgroup consisting of Diceratops and Torosaurus are uncer
tain. However, because Triceratops, Diceratops, and Toro
saurus are all from the strata younger than those that yielding
Eotriceratops, the hypothesis that Eotriceratops occupies a
basal position within the group appears likely.
We thank G. Guthrie for rediscovering the specimen and a
team of technicians and volunteers of the Royal Tyrrell Mu-
seum of Palaeontology (RTMP), led by K. Kruger and
G. Harding, for collecting the specimen. M. Mitchell skill-
fully prepared this difficult specimen and graciously took
extra time to provide the original sketch of Fig. 2, which was
instrumental in describing the specimen. P. Currie kindly
provided information on the original discovery of the speci-
men 91 years ago by B. Brown and provided field support
for this project with the help of British Petroleum Canada.
P. Dodson provided very helpful reference materials, and
R. Sullivan provided beautiful photos of a new torosaur skull
(MOR 1122) for comparison. This study benefitted from
discussions with S. Sampson, M. Ryan, R. Sullivan, and
R. Holmes. T. Sato helped with the initial computer analysis
of the data matrix, and D. Tanke and A. Farke kindly pro
vided references and information. X.-c. Wu is very grateful
to J. Gardner and the collection personnel of the RTMP for
assistance during his visits, and thanks W. Joyce, M. Norell,
C. Mehling, R. Purdy, N. Gilmore, T. Daeschler, Z.-x. Lou,
J. Eberle, and B. Small for opportunities to examine specimens
in their care. C.A. Forster, T.M. Lehman, and H.-D. Sues care
fully reviewed the manuscript and made helpful comments
and suggestions, which improved it significantly. This pro
ject was supported by grants from the RTMP and CMN, and
the RTMP Cooperating Society.
Braman, D.R., and Sweet, A.R. 1999. Terrestrial palynomorph
biostratigraphy of the Cypress Hills, Wood Mountain, and Turtle
Mountain areas (Upper Cretaceous—Paleocene) of western Can
ada. Canadian Journal of Earth Sciences, 36: 725–741.
Colbert, E.H., and Bump, J.D. 1947. A skull of Torosaurus from
south Dakota and a revision of the genus. Proceedings of the
Academy of Natural Sciences of Philadephia, XCIX(49): 93–
Currie, P.J., and Koppelhus, E.B. 2005. Dinosaur Provincial Park.
Indiana University Press, Bloomington and Indianapolis, Ind.
Dodson, P., Forster, C.A., and Sampson, S.D. 2004. Ceratopsidae.
In The Dinosauria. 2nd ed. Edited by D.B. Weishampel, P.
Dodson, and H. Osm`lska. University of California Press,
Berkeley, Los Angeles, and London, pp. 494–513.
Eberth, D.A. 2004. A revised stratigraphy for the Edmonton Group
(Upper Cretaceous) and its potential sandstone reservoirs.
Pre-conference field trip #7. Canadian Society of Petroleum
Geologists – Canadian Heavy Oil Association Joint Conference,
Calgary, Alta. 31 May – 4 June 2004. Canadian Society of Pe
troleum Geology, Calgary, Alta., 44 pp.
Eberth, D.A., and O’Connell, S.C. 1995. Notes on changing
paleoenvironments across the Cretaceous–Tertiary boundary
(Scollard Formation) in the Red Deer River valley of southern
Alberta. Bulletin of Canadian Petroleum Geology, 43: 44–53.
Eberth, D.A., Currie, P.J., Brinkman, D.B., Ryan, M.J., Braman,
D.R., Gardner, J.D., et al. 2001. Alberta dinosaurs and other fos
sil vertebrates: Judith River and Edmonton Groups (Campanian-
Maastrichtian). In Guidebook for the Field Trips. Mesozoic and
Cenozoic Paleontology in the Western Plains and Rocky Moun
tains. Society of Vertebrate Paleontology 61st Annual Meeting,
Bozeman, Mont. 3–6 October 2001. Edited by C.L. Hill. Mu
seum of the Rockies, Bozeman, Montana, Occasional Paper 3,
Farabee, M.J., and Canright, J.E. 1986. Stratigraphy, palynology of
the lower part of the Lance Formation (Maestrichtian) of Wyo-
ming. Palaeontographica, Abteilung B, 199, pp. 1–89.
Forster, C.A. 1990. The cranial morphology and systematics of
Triceratops with a preliminary analysis of ceratopsid phylogeny.
Ph.D. dissertation, University of Pennsylvania, Philadelphia, Pa.
Forster, C.A. 1996a. A species resolution in Triceratops: cladistic
and morphometric approaches. Journal of Vertebrate Paleontol-
ogy, 16: 259–270.
Forster, C.A. 1996b. New information on the skull of Triceratops.
Journal of Vertebrate Paleontology, 16: 246–258.
Gibson, D.W. 1977. Upper Cretaceous and Tertiary coal-bearing
strata in the Drumheller-Ardley region, Red Deer River Valley,
Alberta. In Geological Survey of Canada, Paper 76-35, 41 pp.
Hamblin, A.P. 2004. The Horseshoe Canyon Formation in southern
Alberta: Surface and subsurface stratigraphic architecture,
sedimentology, and resource potential. Geological Survey of
Canada, Bulletin 578.
Hatcher, J.B., Marsh, O.C., and Lull, R.S. 1907. The Ceratopsia.
US. Geological Survey, Monograph 49, pp. 1–300.
Hicks, J.F., Obradovich, J.D., and Tauxe, L. 1999. Magnetostra
tigrahy, isotopic age calibration and intercontinental correlation
of the Red Bird section of the Pierre Shale, Niobrara County,
Wyoming, USA. Cretaceous Research, 20: 1–27.
Hicks, J.F., Johnson, K.R., Obradovich, J.D., Tauxe, L., and Clark,
D. 2002. Magnetostratigraphy and geochronology of the Hell
Creek and basal Fort Union formations of southwestern North
Dakota and a recalibration of the age of the Cretaceous–Tertiary
boundary. In The Hell Creek Formation and the Cretaceous–
Tertiary boundary in the northern Great Plains: an integrated
continental record of the end of the Cretaceous. Edited by J.H.
Hartman, K.R. Johnson, and D.J. Nichols. Geological Society of
America, Special Paper 361, pp. 35–55.
Leffingwell, H.A. 1971. Palynology of the Lance (Late Creta
ceous) and Fort Union (Paleocene) formations of the type Lance
area, Wyoming. In Symposium on Palynology of the Late Creta
ceous and Early Tertiary. Edited by R.M. Kosanke. Geological
Society of America, Special Paper, 127: 1–64.
© 2007 NRC Canada
1262 Can. J. Earth Sci. Vol. 44, 2007
Lehman, T.M. 1990. The ceratopsian subfamily chasmosaurinae:
sexual dimorphism ans systematics. In Dinsoaur systematics:
approach and perspectives. Edited by K. Carpenter and P. J. Cur
rie. Cambridge University Press, Cambridge, UK., pp. 211–230.
Lehman, T.M. 1996. A horned dinosaurs from the El Picachu For
mation of west Texas and review of ceratopsid dinosaurs from
the American Southwest. Journal of Paleontology, 70: 494–508.
Lehman, T.M. 1998. A gigantic skull and skeleton of the horned
dinosaur Pentaceratops sternbergi from New Mexico. Journal of
Paleontology, 72: 894–906.
Lerbekmo, J.F. 1985. Magnetostratigraphic and biostratigraphic
correlations of Maastrichtian to early Paleocene strata between
south-central Alberta and southwestern Saskatachewan. Bulletin
of Canadian Petroleum Geology, 33: 213–226.
Lerbekmo, J.F., and Braman, D.R. 2002. Magnetostratigraphic and
biostratigraphic correlation of late Campanian and Maastrichtian
marine and continental strata from the Red Deer Valley to the
Cypress Hills, Alberta, Canada. Canadian Journal of Earth Sci
ences, 39: 539–557.
Lerbekmo, J.F., and Coulter, K.C. 1985. Late Cretaceous to early
Tertiary magnetostratigraphy of a continental sequence: Red
Deer Valley, Alberta, Canada. Canadian Journal of Earth Sci
ences, 22: 567–583.
Lull, R.S. 1908. The cranial musculature and the origin of the frill
in the ceratopsian dinosaurs. The American Journal of Sciences,
Lull, R.S. 1933. A review of the Ceratopsia or horned dinosaurs.
Memoirs of the Peabody Museum of Natural History, 3, pp. 1–
Lund, S.P., Hartman, J.H., and Banerjee, S.K. 2002.
Magnetostratigraphy of interfingering upper Cretaceous–
Paleocene marine and continental strata of the Williston Basin,
North Dakota and Montana. In The Hell Creek Formation and
the Cretaceous–Tertiary Boundary in the northern Great Plains:
an integrated continental record of the end of the Cretaceous.
Edited by J.H. Hartman, K.R. Johnson, and D.J. Nichols. Geo-
logical Society of America, Special Paper 361, pp. 57–74.
Marsh, O.C. 1888. A new family of horned dinosaurs from the
Cretaceous. American Journal of Science, Series 3, 36: 477–
Marsh, O.C. 1890. Additional characters of the Ceratopsia, with
notice of new Cretaceous dinosaurs. American Journal of Sci
ence, 39: 419–426.
McCabe, P.J., Strobl, R.S., Macdonald, D.E., Nurkowski, J.R., and
Bosman, A. 1989. An evaluation of the coal resources of the
Horseshoe Canyon Formation and laterally equivalent strata, to
a depth of 400 m, in the Alberta Plains area. Alberta Research
Council, Open File Report 1989-07.
Nichols, D.J. 2002. Palynology and palynostratigraphy of the Hell
Creek Formation in North Dakota: a microfossil record of plants
at the end of Cretaceous time. In The Hell Creek Formation and
the Cretaceous–Tertiary boundary in the northern Great Plains:
an integrated continental record of the end of the cretaceous.
Edited by J.H. Hartman, K.R. Johnson, and D.J. Nichols. Geo
logical Society of America, Special Paper 361, pp. 393–456.
Norton, N.J., and Hall, J.W. 1969. Palynology of the Upper Creta
ceous and Lower Tertiary in the type locality of the Hell Creek
Formation, Montana, USA. Palaeontographica, Abteilung B,
125, pp. 1–64.
Ogg, J.G., Agterberg, F.P., and Gradstein, F.M. 2004. The Creta
ceous period. In A geologic timescale 2004. Edited by F.M.
Gradstein, J.G. Ogg, A. Smith, A.G., Bleeker, W., and
Lourens, L.J. Cambridge University Press, Cambridge, UK.,
Parks, W.A. 1925. Arrhinoceratops brachyops, a new genus and
species of Ceratopsia from the Edmonton Formation of Alberta.
University of Toronto Studies, Geological Series, 19: 1–15.
Russell, D.A. 1967. A census of dinosaur specimens collected in
Western Canada. National Museum of Canada, Natural History
Papers, 36: 1–13.
Russell, L.S. 1975. Mammalian faunal succession in the Creta
ceous system of western North America. In The Cretaceous sys
tem in the western interior of North America. Edited by W.G.E.
Caldwell. The Geological Association of Canada, Special Paper
13, pp. 137–160.
Ryan, M.G., and Russell, A.P. 2001. Dinosaurs of Alberta (exclu
sive of Aves). In Mesozoic vertebrate life. Edited by D.H. Tanke
and K. Carpenter. Indiana University Press, Bloomington, Ind.,
Seeley, H.G. 1888. The classification of the Dinosauria. Report of
the British Association for the Advancement of Science, 57: 31–
Srivastava, S.K. 1970. Pollen biostratigraphy and paleoecology of
the Edmonton Formation (Maestrichtian), Alberta, Canada.
Palaeogeography, Palaeoclimatology, Palaeoecology, 7: 221–
Sternberg, C.M. 1929. A new species of horned dinosaur from the
Upper Cretaceous of Alberta. Bulletin of National Museum of
Canada, 54: 34–37.
Sullivan, R.M., and Lucas, S.G. 2003. The Kirtlandian, a new land-
vertebrate “age” for the Late Cretaceous of Western North
America. In New Mexico Geological Society Guidebook: Geol-
ogy of the Zuni Plateau 2003, 54th Field Conference. pp. 369–
Sullivan, R.M., and Lucas, S.G. 2006. The Kirtlandian Land-
Vertebrate “Age” faunal composition, temporal position and
biostratigraphic correlation in the nonmarine Upper Cretaceous
of western North America. New Mexico Museum of Natural
History and Science Bulletin, 35: 7–30.
Sullivan, R.M., Boere, A.C., and Lucas, S.G. 2005. Redescription
of the ceratopsid dinosaur Torosaurus utahensis (Gilmore 1946)
and a revision of the genus. Journal of Paleontology, 79: 564–
Swofford, D.L. 2002. PAUP*. Phylogenetic analysis using parsi
mony (*and other methods), version 4.0b10. Sinauer associates,
Weishampel, D.B., Dodson, P., and Osm`lska, H. (Editors). 2004.
The Dinosauria. 2nd ed. University of California Press, Berke
ley, Los Angeles, and London.
Appendix A. Characters and character state
The first 73 characters are from Dodson et al. (2004),
character 74 is from Forster (1990), and the last two are
1. Rostral, size and shape: triangular in lateral view with
short dorsal and ventral processes (0); enlarged with
deeply concave caudal margin and hypertrophied dorsal
and ventral processes (1).
2. External naris, size and position: small, restricted to
dorsal one-third of snout (0); large, expanded to occupy
most of the depth of the snout (1).
3. Premaxillary septum: absent (0); present, rostral
premaxillae lie directly on one another (1).
4. Premaxillary septum, shape: subcircular (0); rostrally
© 2007 NRC Canada
Wu et al. 1263
5. Premaxilla, thickened narial strut along posterior margin
of the premaxillary septum: absent (0); present (1).
6. Premaxilla, narial strut orientation: rostrally inclined
(0); caudally inclined (1).
7. Premaxilla, bony flange on caudal margin of narial
strut: absent (0); present (1).
8. Premaxilla, interpremaxillary fossa perforating
premaxillary septum: absent (0); present (1).
9. Premaxilla, recess in ventral portion of septum: absent
(0); present (1).
10. Premaxilla, size of recess in septum: small, slight in-
pocketing along ventral aspect of septal fossa (0); large,
deeply recessed into premaxilla (1).
11. Premaxilla, premaxillary (narial) process extending into
the external naris from the caudoventral margin of the
premaxillary septum: absent (0); present (1).
12. Premaxilla, recess on lateral surface of the premaxillary
(narial) process: absent (0); present (1).
13. Premaxilla, ventral expansion of the caudoventral oral
margin: absent, caudoventral oral margin of premaxilla
level with alveolar margin of the maxilla (0); present,
expanded ventrally to extend well below alveolar mar
gin of the maxilla (1).
14. Premaxilla, caudal tip of caudoventral process inserts
into an embayment in the nasal and is surrounded by the
nasal: present (0); absent (1).
15. Premaxilla, forked distal end of caudoventral process:
absent (0); present (1).
16. External antorbital fenestra, size: large, 20% or more
length of body of maxilla (0); greatly reduced to less
than 10% length of body of maxilla, or absent (1).
17. Nasal, ornamentation: absent (0); horncore (1);
pachyostotic boss (2).
18. Nasal, ornamentation position: centered caudal or
caudodorsal to internal naris (0); shifted forward, cen-
tered dorsal or rostrodorsal to endonaris (1).
19. Nasal, horncore length: small, length of horncore less
than 15% basal skull length (0); moderate to large,
length of horncore 20% or more of basal skull length
20. Lacrimal, size: large, forms 50% or more of the rostral
orbital margin (0); small, forms 40% or less of the
rostral orbital margin (1).
21. Postorbital, supraorbital ornamentation: absent (0); pres
22. Postorbital, supraorbital ornamentation: horncore (0);
rugose boss (1).
23. Postorbital, position of supraorbital ornamentation: cen
tered rostrodorsal or dorsal to orbit (0); centered
caudodorsal or caudal to orbit (1).
24. Postorbital, length of supraorbital horncore: elongate,
greater than 35% basal skull length (0); short, less than
15% basal skull length (1).
25. Postorbital, curvature of supraorbital horncore in adults:
caudally recurved (0); straight or rostrally curved (1).
26. Prefrontal–prefrontal contact: absent (0); present (1).
27. Jugal infratemporal flange: absent (0); present, contacts
jugal process of squamosal below the infratemporal
28. Quadratojugal, position of contact with jugal and
quadrate: jugal, quadratojugal, quadrate contact in a
rostral to caudal order (0); jugal, quadratojugal,
quadrate contact in a lateral to medial order (1).
29. Supracranial cavity complex: absent (0); present,
supracranial cavities narrow and shallow, do not under
lie supraorbital ornamentation (1); present, supracranial
cavities broad, underlie supraorbital ornamentation and
may be confluent with extensive cornual sinuses (2).
30. Frontal, contribution to orbital margin: present (0); ab
31. Frontal fontanelle, shape: rostrocaudally long and uni
formly narrow transversely (0); key-hole shaped with a
broad rostal half (1); broadly oval to circular (2).
32. Parietosquamosal frill, length relative to basal skull
length: elongate, 0.80 or more (0); shortened, 0.70 or
33. Parietosquamosal frill, marginal undulations: absent (0);
34. Parietal, imbrication effect on lateral margin of frill in
adults: absent (0); present (1).
35. Squamosal, length relative to parietal: equal or subequal
in length (0); squamosal much shorter than parietal (1).
36. Squamosal, rostromedial lamina forming the
caudolateral floor of dorsotemporal fossa: absent (0);
37. Squamosal–quadrate contact: socketlike cotylus on
squamosal for ball-like quadrate head (0); elongate
groove on squamosal to receive lamina of quadrate (1).
38. Parietal fenestra, orientation: long axis directed trans-
versely (0); long axis directed axially (1).
39. Parietal fenestra, maximum proximodistal diameter:
40% or less total parietal length (0); 45% or more total
parietal length (1).
40. Parietal, accessory fenestra medial to the rostral end of
the infratemporal fenestra: absent (0); present (1).
41. Parietal, median bar: narrow and straplike, width less
than 0.05 total parietal length (0); wide, 0.15 or more
total parietal length (1).
42. Epoccipitals on parietal and squamosal: absent (0); pres
43. Epoccipital crossing squamosal–parietal contact: absent
(0); present (1).
44. Epoccipitals, pattern of fusion to frill margin: occurs
from rostral to caudal (0); occurs from caudal to rostral
45. Squamosal epoccipital, shape: crescentic or ellipsoidal
(0); triangular (1); elongate spindle-like (2).
46. Parietal epoccipitals, number per side: one to three (0);
six to eight (1).
47. Parietal epoccipital, most median epoccipital (locus 1)
developed into a curved process with pronounced exter
nal sulci and ridges: absent (0); present (1).
48. Parietal epoccipital, orientation at locus 1 relative to pari
etal: dorsal to rostrally directed (0); caudally directed (1).
49. Parietal epoccipital, locus 2 developed into a curved
process, with pronounced external sulci and ridges: ab
sent (0); present (1).
50. Parietal epoccipital, orientation at locus 2: caudally di
rected from and perpendicular to parietal margin (0);
medially directed, comes off the parietal at an angle (1).
51. Parietal epoccipital, locus 3 modified into a large horn
like process: absent (0); present (1).
© 2007 NRC Canada
1264 Can. J. Earth Sci. Vol. 44, 2007
52. Basioccipital, contribution to occipital condyle: large,
forms more than one-third of condyle (0); reduced,
forms ventral one-third of condyle only (1).
53. Frontal, contribution to exit for cranial nerve I: present
(0); absent, enclosed entirely by ossifications of the
interorbital septum (1).
54. Supraoccipital, contribution to foramen magnum: pres
ent (0); absent, eliminated from margin by exoccipital–
exoccipital contact on midline (1).
55. Exoccipital, number of exits for cranial nerves X, XI,
XII: three (0); two (1).
56. Lower jaw, level of articulation with quadrate: same as
occlusal surface of tooth row (0); substantially ventral to
tooth row (1).
57. Predentary, orientation of triturating surface: nearly hor
izontal (0); inclined steeply laterally (1).
58. Dentary, caudal extent of tooth row: terminates medial
to the coronoid process (0); terminates caudal to the
coronoid process (1).
59. Dentary, shape of the coronoid process: low, with gently
convex apex and no neck (0); high, expanded into a
rostrally projecting hook at apex, constricted neck pres
60. Tooth, number of roots: one (0); two (1).
61. Tooth vertical replacement series: one or two replace-
ment teeth (0); more than two replacement teeth (1).
62. Tooth ornamentation: subsidiary ridges present, extend
from margin to base of tooth (0); subsidiary ridges re-
duced, present only at margin of teeth (1).
63. Atlas, neural arch orientation: nearly vertical (0);
steeply inclined caudally (1).
64. Sacrum, number of fused vertebrae: eight or less (0); ten
or more (1).
65. Sacrum, deep longitudinal channel on ventral surface:
present (0); absent (1).
66. Sternum, shape: elongate and narrow (0); short and
67. Manual and pedal unguals, shape: clawlike (0); hooflike
68. Ilium, laterally everted shelf on dorsal margin: absent
(0); present (1).
69. Ilium, supracetabular process on dorsal margin over
caudal part of acetabulum: absent (0); present (1).
70. Ischium, cross-sectional shape of shaft: ov oid (0); laterally
compressed and bladelike, narrow along dorsal margin (1).
71. Ischium, orientation of shaft: nearly straight (0); slightly
decurved (1); broadly and continuously curved (2).
72. Femur, coalescence of greater and cranial trochanters:
absent (0); present (1).
73. Femur–tibia proportion: tibia longer than femur (0); fe
mur longer than tibia (1).
74. Position of nasal horn core: posterior (0); over front of
nares (1) (Forster 1990).
75. Transversely arched vascular trace across the anterior
surface of the nasal horn core: absent (0); present (1).
76. A process projecting into the interpremaxillary fenestra:
absent (0); present (1).
© 2007 NRC Canada
Wu et al. 1265
Protoceratops 000?0 ???0? 0?000 00000 0???? 00000 ?0000 00000 00??? ????? ?0000 00000 00000 00000 000NN ?
Zuniceratops ????? ????? ????? ????? 10000 ????? ????? ????? ????? ????? ????? ??010 00??? ????? ????? ?
Achelousaurus 01100 ??00? 0?110 120?1 1101? ?1111 01111 11100 11110 10001 11111 11111 11110 11110 1110? N
Anchiceratops 11111 101?0 10010 11101 10101 ?0121 20100 01100 11001 00000 01111 10111 111?? 11111 21100 ?
Arrhinoceratops 11111 1010? 10010 11101 10101 ?01?1 ?0100 01100 1???? 00000 01?11 1???1 ????? ????? ???00 ?
Avaceratops ?1100 ??00? 0?11? ????? ????? ??1?? ??101 01??? ????? ????? ?1?11 1?111 111?? ?1??? ?11?? N
Centrosaurus 01100 ??00? 0?110 11011 10010 11111 01111 11100 11110 11111 01111 11111 11110 11110 11100 N
Chasmosaurus 11111 011?0 10001 11?01 100?0 0?111 10100 01110 01001 00100 01111 10111 11111 11111 21100 0
Diceratops 11111 00111 11010 11101 10101 ?0121 20100 01101 110?1 00000 01?11 1???1 11??? ????? ???10 1
Einiosaurus 01100 ??00? 0?110 11011 1001? ?1111 01111 11100 11110 10001 11111 11111 11110 11110 11100 N
Pachyrhinosaurus 01100 ??00? 0?110 120?1 1101? ?1111 01111 11100 11110 10011 11111 11111 11110 11110 11100 N
Pentaceratops 11111 011?0 10001 11?01 10001 ?01?1 10100 01110 01001 00100 01111 10111 11111 ?1111 21100 1
Styracosaurus 01100 ??00? 0?110 11011 1001? 11111 01111 11100 11110 11111 11111 11111 11110 11110 11100 N
Torosaurus 11111 00111 ??010 11101 10101 00121 20100 01101 11001 00000 01111 10111 11??? ????? ???10 ?
Triceratops 11111 00111 11010 11101 10101 00121 21100 01100 11101 00000 01111 10111 11111 11111 21111 1
Eotriceratops 11111 00110 100?? 1??0? 10101 ??121 ??1?? ?1??? ?1102 ????? ?1??? ????1 111?? ????? ???11 1
Table A1. Character state data matrix.