Anatomy and Taxonomic Status of the Chasmosaurine Ceratopsid Nedoceratops hatcheri from the Upper Cretaceous Lance Formation of Wyoming, U.S.A

Background The validity of Nedoceratops hatcheri, a chasmosaurine ceratopsid dinosaur known from a single skull recovered in the Lance Formation of eastern Wyoming, U.S.A., has been debated for over a century. Some have argued that the taxon is an aberrant Triceratops, and most recently it was proposed that N. hatcheri represents an intermediate ontogenetic stage between “young adult” and “old adult” forms of a single taxon previously split into Triceratops and Torosaurus. Methodology/Principal Findings The holotype skull of Nedoceratops hatcheri was reexamined in order to map reconstructed areas and compare the specimen with other ceratopsids. Although squamosal fenestrae are almost certainly not of taxonomic significance, some other features are unique to N. hatcheri. These include a nasal lacking a recognizable horn, nearly vertical postorbital horncores, and relatively small parietal fenestrae. Thus, N. hatcheri is tentatively considered valid, and closely related to Triceratops spp. The holotype of N. hatcheri probably represents an “old adult,” based upon bone surface texture and the shape of the horns and epiossifications on the frill. In this study, Torosaurus is maintained as a genus distinct from Triceratops and Nedoceratops. Synonymy of the three genera as ontogenetic stages of a single taxon would require cranial changes otherwise unknown in ceratopsids, including additions of ossifications to the frill and repeated alternation of bone surface texture between juvenile and adult morphotypes. Conclusions/Significance Triceratops, Torosaurus, and likely Nedoceratops, are all distinct taxa, indicating that species richness for chasmosaurine ceratopsids in the Lance Formation just prior to the Cretaceous-Paleocene extinction was roughly equivalent to that earlier in the Cretaceous.


Introduction
In 1889, John Bell Hatcher and his field crew undertook the first of four productive seasons of exploration in the late Maastrichtianaged Lance Formation of Niobrara County, Wyoming, USA, under the auspices of the United States Geological Survey. These expeditions resulted in a massive collection of vertebrate fossils, ranging from fish to mammals to dinosaurs. Most remarkably, over 30 ceratopsian (horned dinosaur) skulls were recovered [1]. These included collections of fragments, disarticulated specimens, and complete, articulated crania, which were used to erect a number of genus-and species-level taxa that have literally defined our concept of ''horned dinosaurs.'' Among the initial series of discoveries were 12 species of Triceratops (later variably sunk into one or two species, Triceratops horridus and Triceratops prorsus; e.g., [2,3]); Sterrholophus flabellatus (synonymous with Triceratops horridus; [3]); Torosaurus latus and Torosaurus gladius (only T. latus is now considered valid; [4]); and Nedoceratops hatcheri [5,6].
Nedoceratops hatcheri, a taxon erected based upon a single, nearly complete skull of a large chasmosaurine (,1.8 m greatest skull length; Figures 1-6), has suffered a long history of proposed synonymy, incomplete description and figuring, and nomenclatural confusion. Hatcher originally intended to name the taxon as a new genus and species, but died before he could do so [7]. Thus, it fell upon Richard S. Lull to attach the name Diceratops hatcheri to a brief description (scarcely a full page of text) penned by Hatcher [5]. Decades later, Lull changed his mind on the generic status of Diceratops and relegated it to a subgenus of Triceratops [7]. In 1986, Ostrom and Wellnhofer [2] posited that Diceratops hatcheri, along with all of the named species of Triceratops, fell within the expected range of variation for a single taxon, Triceratops horridus. This situation remained until 1996, when Forster, in her revision of Triceratops and related forms, suggested that Diceratops was indeed a valid taxon [3]. This opinion has not been unanimously accepted among ceratopsid workers, with many considering the taxon to be synonymous with Triceratops horridus (e.g., [8,9]). Most recently, Scannella and Horner [10] hypothesized that Diceratops hatcheri represented a transitional form between the ''young adult'' and ''old adult'' forms of Triceratops (with Torosaurus as the fully adult form of Triceratops).
Finally, the name ''Diceratops'' itself has experienced a confusing history. After it was discovered that the genus name Diceratops was preoccupied by an extant wasp [11], two competing replacement names were proposed: Nedoceratops [6] and Diceratus [12]. Although initially less widely known, Nedoceratops was the first to be published and hence has priority as the generic name.
The type and only specimen for Nedoceratops hatcheri, USNM 2412, remains poorly understood. In particular, the skull preserves features such as small parietal fenestrae, paired squamosal fenestrae, and erect supraorbital horns (Figures 1-3), that have been variably considered individual variation, ontogenetic features, pathology or valid characters separating the taxon from Triceratops (e.g., [1][2][3]5,7,10,13]). This debate over the validity of N. hatcheri stems largely from inadequately brief and occasionally incorrect descriptions of the specimen, incomplete documentation of restoration on the skull, and differing opinions among ceratopsid workers. Although the last situation continues, this paper addresses the first two issues as a means to move the discussion forward. An understanding of the anatomy and taxonomic status of N. hatcheri is important for answering two Interpretive line drawing in left lateral view, with major reconstructed areas indicated in gray and matrix indicated in yellow. In C, the rostral end of the skull was broken away at the time of photography. Abbreviations: aof, antorbital fenestra; cc, caudal curve of oral margin; if, infratemporal fenestra; j, jugal; jn, jugal notch of squamosal; nh, nasal horncore; o, orbit; p, parietal; pf, parietal fenestra; poh, postorbital horncore; q, quadrate; rc, rostral curve of oral margin; sps, squamosal-parietal suture; sq, squamosal; sqf, squamosal fenestra. Scale bar equals 10 cm. doi:10.1371/journal.pone.0016196.g001 Interpretive line drawing in rostral view, with major reconstructed areas indicated in gray and matrix indicated in yellow. Abbreviations: aob, antorbital buttress; m, matrix and metal supports; nh, nasal horncore. Scale bar equals 10 cm, but note that parallax prevents accurate measurement of parts of the skull caudal to the external naris. doi:10.1371/journal.pone.0016196.g002 questions in particular: 1) How many species of large ceratopsid co-existed in the area of the Lance and Hell Creek formations; and 2) If N. hatcheri is valid, what separates it from related forms such as Triceratops and Torosaurus?
Here, I present the first comprehensive description, diagnosis and illustration of the type specimen for Nedoceratops hatcheri. The degree of restoration of the skull is fully documented (Figures 1-3), considerably altering some previous interpretations of the cranial anatomy (e.g., [1,7]). The holotype skull of N. hatcheri is compared to the morphologies in the coeval Triceratops and Torosaurus in particular, in an attempt to address the validity of N. hatcheri. Finally, I address the recently proposed hypothesis that Triceratops and Torosaurus are two stages of an ontogenetic series for a single taxon, with Nedoceratops possibly representing a transitional form between the two morphs [10].     Diagnosis. Chasmosaurine ceratopsid with the following autapomorphies: nasal horncore nearly completely undifferentiated from the nasal bone; greater portion of procurved postorbital horncores forms a 90 degree angle with tooth row; and parietal fenestrae extremely small (occupying less than five percent of the total surface area of the parietal). Nedoceratops hatcheri is distinguished from Triceratops spp. in the the position of the ventral extremity of the squamosal well above the alveolar process of the maxilla, and in the presence of parietal fenestrae, which are lacking in all Triceratops species. Nedoceratops hatcheri is distinguished from Torosaurus latus in squamosal shape (particularly the reduced jugal notch and lack of a thickened medial margin in N. hatcheri), and that N. hatcheri has extremely reduced parietal fenestrae and a low number of episquamosals in N. hatcheri compared to T. latus.

Description
The holotype skull for Nedoceratops hatcheri, USNM 2412, is nearly complete and virtually undistorted (Figures 1-3). Only minor evidence of crushing (e.g., asymmetry in preservation) is visible. Thus, the shape and orientation of cranial structures are probably largely unaltered relative to the life condition. The skull lacks only the caudal margin of the parietal and most of the right postorbital horncore, with other small absences that may have resulted from excavation, preparation or loss before burial ( Figures 1B,1D,2C,3B). Nearly all of the cranial sutures are obliterated by co-ossification. This is one of several features consistent with USNM 2412 representing an ''old adult'' at its death.
In the following regional descriptions, orientations of cranial structures are described assuming a horizontal tooth row. Most of the palate of the skull is either unprepared or obscured by plaster and paint, and thus that region is not described in detail here. Selected measurements are presented in Table 1.
Rostral and Narial Region. The nasals, premaxillae, and rostral are extensively co-ossified, so sutural relationships among these elements cannot be discerned. The dorsal margin of the ''beak'' (the portion of the face rostral to the nasal horncore, comprising the fused premaxillae and rostral bone) circumscribes a broad arc, with the tip of the rostral bone located just ventral to the level of the maxillary tooth row and the lowest part of the premaxilla ( Figure 4). The ventral margin of the beak consists of two sinuous curves, with the rostral curve approximately twice as broad as the caudal curve ( Figures 1A,1B,4), a feature best seen on the right side due to damage to the specimen ( Figure 1B). Yet, compared to many other chasmosaurines (e.g., Chasmosaurus spp.; most examples of Triceratops), the degree of curvature on the beak as a whole is relatively minor.
The premaxilla displays a number of fossae and foramina consistent with the condition in other chasmosaurines (e.g., see [23]; the anatomical terminology from that paper is used here for the narial region), but plaster restoration obscures some details ( Figures 1B,1D,4). The narial strut is inclined rostrally towards the dorsal end of the element, and enough original bone surface is preserved to indicate that a posterior internarial flange did not project from the caudal surface of this structure ( Figure 4). A prominent triangular process extends from the ventral portion of the narial strut, and the process is situated so that its point projects dorsally into the endonaris (as seen in Triceratops and other closelyrelated chasmosaurines) rather than caudally (as seen in Chasmosaurus spp., Pentaceratops sternbergi, and others). The lateral surface of the triangular process displays a fossa, but plaster and matrix obscure any communications between this fossa and other portions of the narial complex. Similarly, the medial surfaces of the triangular processes are obscured by plaster. A small depression, the premaxillary fossa, occurs immediately rostral to the narial strut. Much of the morphology within the fossa (including a foramen piercing the fossa, as well as a strut within the fossa) is  Table 1. Selected measurements of the holotype skull for Nedoceratops hatcheri, USNM 2412, in millimeters.

Measurement
Basal skull length (rostrum to occipital condyle) 1100 Length from tip of rostrum to distal end of maxillary tooth row 804 Length from rostral rim of orbit to caudal rim of external naris 252 Length from rostral rim of orbit to tip of rostrum 854 Maximum skull width, between orbits 395 Maximum skull width, between distal tips of jugals 701 Height of snout, from bottom of premaxilla to top of nasal 414 Height of skull, from distal tip of jugal to rostral end of supratemporal fenestra restored in plaster ( Figures 1B,1D,4). A canal at the ventral edge of the fossa communicates caudally, although the extent of the canal is not fully prepared ( Figure 4). The portion of the narial complex rostral to the narial strut is approximately the same maximum length as the portion caudal to the strut. Nasal and Nasal Horncore. The nasal horncore of USNM 2412 (Figures 1,2,4) is one of the least pronounced for any known chasmosaurine ceratopsid skull. Rather than a discrete triangular horn (as seen in Triceratops prorsus; e.g., LACM 7207, YPM 1822) or knob of bone (as seen in many Triceratops horridus; e.g., USNM 4720, YPM 1820), the ''horn'' of USNM 2412 is a poorly defined swelling in the dorsal surface of the snout, similar to the condition seen in the basal centrosaurine Diabloceratops eatoni [24]. The snout measures approximately 150 mm in width at the midpoint of the external naris. No trace of an arcuate vessel along the rostral edge of the horncore is preserved, as seen in Eotriceratops and Triceratops [8,25], although plaster reconstruction partly obscures the relevant region. Furthermore, extensive sutural fusion in USNM 2412 obscures any evidence of a separate epinasal ossification, if one existed in the first place. The nasal immediately caudal to the horncore is relatively thin, as seen in Triceratops (e.g., USNM 4720) and contrasting with the condition seen in Coahuilaceratops magnacuerna [26].
Mid-Facial Region. The dorsal surface of the mid-facial region of USNM 2412 is strongly sloped in comparison with most other chasmosaurines and much lower at its rostral end than at its caudal end when seen in lateral view ( Figure 1). Only the Triceratops horridus skull USNM 4928 (holotype of ''Triceratops calicornis'') approaches this morphology. The antorbital fenestra, which is caudally inclined and much longer than wide (Figure 1), is relatively small compared to the primitive condition seen in neoceratopsians (e.g., Protoceratops) and comparable in size to fenestrae seen in most other ceratopsids.
The general morphology of the maxilla is similar to that of Triceratops and Torosaurus. The exact number of maxillary tooth positions cannot be counted due to poor preservation, but at least 21 teeth are preserved on each side. Rostrally, no teeth are preserved, but there is space for approximately five to 10 more alveoli. The teeth are poorly preserved, so nothing can be said of the dental anatomy other than that it appears to be similar to that of other ceratopsids.
The orbits are slightly taller than wide, and a thickened antorbital buttress extends from the dorsal surface of the orbit to the rostro-ventral portion of the orbit (Figure 2). The buttress effectively occludes the orbit from visibility in rostral view of the skull (Figures 2B,2C).
The right postorbital horncore is missing only its distal quarter, but most of the left horncore is restored in plaster (see Figures 1B,1D,2C; contrary to published drawings; compare with plates 47 and 48 in [1]). The horns are remarkably erect, particularly at their bases, more so than in any undistorted Triceratops skull. The left horncore displays only modest rostral curvature towards its distal end, and the mid-section is moderately bowed laterally. The bases of the horncores are longer than wide.
The jugals project nearly ventrally, with only a modest lateral component and virtually no caudal inclination along the long axis of the element. The ventral margin of the jugal ends just barely below the ventral margin of the maxilla.
An epijugal is apparently present, but its sutural relations with the jugal and quadratojugal are completely obscured by fusion. Regardless, the element was not prominent even by the standards of many Triceratops specimens. The bone is approximately tetrahedral in shape, with the rostrodorsal surface being the longest and flattest of the three exposed surfaces. The remaining two surfaces (oriented caudally and ventrally) are comparatively rounded.
The dorsal skull roof is nearly solid, with only a small, circular frontoparietal fontanelle. The fontanelle is positioned towards the caudal half of the bases of the postorbital horncores. Prominent dorsotemporal channels pass caudally from the fontanelle, starting at a single, midline channel wider than the fontanelle itself, before passing into two narrow, caudolaterally trending channels (which then terminate in ''anterior temporal foramina'' [3]). This region has already been described and figured in detail elsewhere (see [27]: fig. 5 and text), so it is not treated further here.
The quadrate extends well below the ventral margin of the maxillary teeth. Its distal cotylus is divided into medial and lateral portions. The distal and caudal margin of the quadrate is clearly visible when the skull is seen in lateral view ( Figure 1B), similar to the condition seen in chasmosaurines such as Pentaceratops, Triceratops, and Utahceratops, but contrasting with the morphology of Chasmosaurus and Mojoceratops.
The infratemporal fenestra has the shape of a rounded triangle, with the ventral margin longest, and the rostral and caudal margins approximately equal in length. The exact relationships of the elements bounding the fenestra cannot be discerned.
Frill. In lateral view, the parietosquamosal frill is remarkably shallow compared to the deep, saddle-shaped form seen in typical Triceratops specimens, particularly because the squamosal is comparatively narrow in USNM 2412 ( Figure 1). The frill is erect, and the dorsal profile, where preserved, is quite straight. In rostral view, the frill is broadly arched from side to side ( Figure 2B). It is shorter than basal skull length, similar to specimens of Triceratops and contrasting with most specimens of Torosaurus.
The ventral extent of the squamosal is roughly level with the ventral margin of the infratemporal fenestra and the top of the alveolar process of the maxilla ( Figure 1); this contrasts with the condition typical of Triceratops, in which the ventral angle of the squamosal extends to the level with the maxillary tooth row or below. Some specimens of Torosaurus latus (e.g., ANSP 15192, MOR 1122) also show a configuration similar to USNM 2412. The squamosal's lateral margin is only gently convex, unlike the prominent crescentic profile seen in many Triceratops specimens (e.g., YPM 1822). Similar to Ojoceratops fowleri, and contrasting with Torosaurus spp. and most Triceratops spp. (MOR 004 is an exception), the ventral corner of the blade of the squamosal is indistinct. The distal end of the squamosal blade tapers to a point. Prominent bilateral fenestra pierce the rostral portion of the blade, although they are asymmetric in size and shape (Figures 1-3; Table 1). The left fenestra is approximately trapezoidal, and the squamosal in this region is greatly swollen around the fenestra's boundaries (up to 67 mm in thickness at the rostral border of the fenestra). The right squamosal fenestra is roughly oval, with the thicknesses of its margins ranging between 14 and 19 mm. These features are further described and interpreted elsewhere [13]. In contrast with previously published drawings (see [1]:plate 47, for example), the distal end of the left squamosal is reconstructed ( Figure 1D).
Episquamosals are present, but the rostral episquamosals are so tightly fused to the squamosal as to be virtually indiscernible. Portions of the lateral edge of the squamosal are restored on both sides, but total episquamosal count is estimated to be five, by comparison of both left and right sides. All preserved episquamosals are approximately equal in size, at around 150 mm in length. Each episquamosal is long, low and approximately ovoid in shape ( Figure 5).
Nearly the entire caudal margin of the parietal is reconstructed (Figures 1B,1D), so it is impossible to determine marginal shape or epiparietal count. Marginal epi-ossifications are restored in plaster along the squamosal-parietal contact, and there is no conclusive evidence for an epi-ossification in this position other than the plaster reconstructions (contra [7]). The midline bar of the parietal is completely unornamented in lateral view, with none of the midline bumps seen variably in other chasmosaurines. A small, elongate parietal fenestra occurs in the middle of the right side (133 mm long by 50 mm wide; Figure 6); the comparable region is restored on the left side of the parietal. The bone surrounding the fenestra is quite thin, between 8 and 10 mm. In caudal view, no major depressions occur on this portion of the parietal (Figure 3); the bone is uniformly flat, except for a midline depression at the rostral end of the bone (presumably for cervical musculature). Although small portions of the rostral, medial, and caudal margin are restored, the entire lateral margin of the fenestra is intact, indicating that this is a genuine feature and not simply the result of incomplete preservation. Importantly, the fenestra was explicitly noted as present by the original preparator [5].
Braincase. The braincase is well-preserved but only partially prepared. All of the component bones of the occipital condyle are well-fused, and no sutures are visible here. Laterally, the wing-like processes of the exoccipitals are extremely broadened (Figure 3), but no more so than is typical of Triceratops and other chasmosaurines (contra [5]). The ventral margins of the process extend below the occipital condyle and even the basisphenoid processes, and are partly visible from a lateral view of the skull (Figure 1). Two foramina for the exits of cranial nerves and associated structures characterize the base of the exoccipital. The contacting pterygoid bones are well-preserved and similar to those seen in Triceratops.

Ontogenetic Status of Nedoceratops hatcheri
The holotype individual of Nedoceratops hatcheri (USNM 2412) is interpreted as an ''old adult,'' based on several features that are comparable with those in specimens considered to be old individuals of Triceratops and other ceratopsians [32,33]. All cranial sutures are completely or nearly completely obliterated, the postorbital horncores are procurved, the bone texture on the frill is deeply vascularized and rugose rather than striated or ''pebbly,'' and the epiossifications are low and elongate, without a triangular peak.

Taxonomic Status of Nedoceratops hatcheri
Opinions on the validity of Nedoceratops hatcheri vary greatly, although most recent work has assumed its synonymy with Triceratops horridus [8][9][10]. The hypothesis that N. hatcheri represents a transitional form between ontogenetic stages (with Triceratops as ''young adult'' and Torosaurus as ''old adult,'' [10]) is considered unlikely for multiple reasons and will be addressed below.
Synonymy of Nedoceratops and Triceratops requires either that any perceived unique features of the former are the result of pathology or developmental anomaly [2,7] or that perceived unique features fall within the expected range of variation for Triceratops [2,7,9]. I consider these hypotheses in turn below.
Undoubtedly, some aspects of the skull of Nedoceratops are abnormal. For instance, the bone around the squamosal fenestra on the left side is massively thickened compared to that on the right side, and the squamosal fenestrae are asymmetric in shape. This, in conjunction with the random occurrence of squamosal fenestrae across chasmosaurines [13], indicates that the squamosal fenestrae are not reliable characters for purposes of alpha taxonomy. It is possible that the parietal fenestra is also abnormal, but the incomplete preservation of the opposite side of the frill prevents a test of this hypothesis. However, the preserved bone texture around the fenestra appears normal relative to the rest of the frill, and the rostrocaudal elongation and caudal placement of the fenestra is consistent with the condition seen in other chasmosaurines with small fenestrae (Anchiceratops and Arrhinoceratops).
Several other features on the skull clearly distinguish Nedoceratops from known specimens of Triceratops. The position of the squamosals, with the ventral border of the squamosal well above the maxillary tooth row, is unique to N. hatcheri. The occurrence of ''anterior temporal foramina'' is also unusual, shared only with some specimens of Torosaurus latus and Torosaurus utahensis [27]. The near absence of a nasal horn in Nedoceratops is of somewhat controversial validity. Nasal horn morphology varies greatly within Triceratops (related at least in part to differences between the two species [28]). However, none of the specimens known or described to date match the extreme condition observed in N. hatcheri. Even USNM 4720 and UCMP 128561, two specimens of Triceratops for which the small size of the nasal horn was used to erect new species [29,30], have more prominent nasal ornamentation than N. hatcheri. Given the ''old adult'' status of USNM 2412 as well as the lack of an open epinasal suture or other evidence of damage, it is improbable that the nasal horn appears small due to traumatic or taphonomic loss of the epinasal ossification (contra [2,3]).
The preponderance of unusual features seen in the type and only skull of Nedoceratops hatcheri may either be explained as a whole suite of abnormalities in a single aberrant individual of Triceratops, or be explained as a suite of autapomorphies characterizing a taxon distinct from Triceratops. I consider the latter explanation to be most likely. This hypothesis could be further bolstered by the discovery and description of additional specimens with similar morphology, or refuted by the identification of undisputed specimens of Triceratops that overlap in all aspects of morphology with N. hatcheri or preserve a mélange of character-states that are intermediate between known Triceratops specimens and Nedoceratops. Additional information relevant to the latter case is presented below and in [10].
Assuming that it is a separate taxon, N. hatcheri is closely allied with the clade including Triceratops and Torosaurus. Nedoceratops shares several features with Triceratops spp. to the exclusion of the coeval Torosaurus latus, including a comparatively short frill relative to skull length, low number of episquamosals, and a lack of a thickened medial margin or prominent concave depression on the dorsal surface of the squamosal. A recent phylogenetic analysis of chasmosaurines [31] recovers N. hatcheri as sister taxon to Triceratops spp., with a decay index of 1 separating the two clades. These taxa are united by a single unambigous synapomorphy, a short parietosquamosal frill (see text S1 in [31]).

Taxonomy of Late Maastrichtian Chasmosaurines
Scannella and Horner [10] proposed that the large chasmosaurine Torosaurus latus was synonymous with coeval Triceratops spp. Although T. latus differs from Triceratops in features such as frill fenestration, frill thickness, and number of epiossifications, it was suggested that these differences were manifested during ontogeny of a single taxon. In other words, Triceratops ontogenetically transformed into Torosaurus as a fully-grown adult. The holotype skull of Nedoceratops hatcheri, USNM 2412, with its small parietal fenestrae and thin frill, was cited as a possible transitional stage between the ''young adult'' and ''old adult'' conditions. This hypothesis (here referred to as the ''Ontogenetic Trajectory Hypothesis,'' or OTH) has major implications for our understanding of ceratopsid evolution and diversity at the Cretaceous-Paleocene boundary. In the section below, the names Triceratops and Torosaurus are used in their traditional sense, as if they were separate taxa.
One assumption of the OTH is that the number of epiossifications on the ceratopsid frill increased through ontogeny. Typical Triceratops spp. frills have between five (e.g., TCM 2001.93.1) and seven (e.g., MOR 1120) episquamosals or undulations for placement of episquamosals, and five or six epiparietals or spaces for epiparietals [10], not counting the ossification spanning the squamosal-parietal suture. In contrast, Torosaurus latus consistently has seven episquamosals and between 10 and 12 epiparietals [34]. Thus, the OTH would require the typical addition of between zero and two episquamosals and between five and seven epiparietals during ontogeny, along with the loss of the midline epiparietal that characterizes nearly all specimens of Triceratops but is absent in all known Torosaurus.
None of the known ontogenetic series for any other ceratopsid suggest that such a large addition of epiossifications ever happened. For instance, the frill of TMP 82.16.11, which likely belongs to a juvenile Centrosaurus apertus approximately one-fifth of adult size, has four scallops on each squamosal and an estimated twelve scallops on the parietal [35]. Assuming that each scallop corresponded to an epiossification, this is within the range of variation for the count seen in full-sized, adult specimens of C. apertus (four [e.g., ROM 767] or five [e.g., TMP 65.12.2] episquamosals and 12 [e.g., CMN 8798] or 14 [e.g., AMNH 5239] epiparietals; personal observation). Thus, C. apertus probably did not add a large number of epiossifications, if any, during ontogeny. In the event that TMP 82.16.11 matured into an individual with five episquamosals and 14 epiparietals, it would have added no more than one or two epiossifications to each of these elements during its entire growth sequence. Similar patterns of constant or near-constant epiossification counts apparently occurred in other centrosaurines for which at least partial ontogenetic series are known (e.g., Styracosaurus albertensis; Einiosaurus procurvicornis).
Although ontogenetic series for the frill are poorly known in chasmosaurine ceratopsids outside of Triceratops, important information is offered by a juvenile squamosal, TMP 98.123.1, from the Dinosaur Park Formation of Alberta, assignable to either Chasmosaurus sp. or Mojoceratops perifania. The element measures 200 mm from the jugal notch to the distal end (compared with a measurement of 910 mm in ROM 839, a presumed adult specimen referable to Chasmosaurus belli; [36]), and has 10 marginal undulations. Presumed adult specimens of Chasmosaurus spp. and Mojoceratops perifania have between 6 and 10 episquamosals or marginal undulations [36]. Thus, there is little evidence that any chasmosaurine consistently added marginal ossifications during ontogeny, and any differences in count between specimens can be attributed to specific differences or individual variation (note that Godfrey and Holmes [36] suggested the number of undulations increased ontogenetically, an interpretation not supported here). Even squamosals from ''baby'' and juvenile Triceratops have between five and seven scallops for attachment of episquamosals [37] (see also figure 3 in [10]), corresponding precisely to the number found in most adult-sized individuals. The smallest known parietal of Triceratops has five places for epiparietals (not counting the ossification spanning the parietal-squamosal suture; [10]), congruent with the count seen in many adult specimens five times its size. The only possible exception may be in Agujaceratops mariscalensis, in which two presumed juvenile or subadult specimens have six episquamosals [38] and a single known presumed adult has ten [39]. Given the variation within Chasmosaurus spp. and the small sample size for A. mariscalensis, this pattern is just as likely the result of individual variation as it is the result of ontogenetic changes.
The OTH also requires that Triceratops acquire parietal fenestrae at the very end of ontogeny. This would contrast sharply with the known conditions in Protoceratops andrewsi [40] and Centrosaurus apertus [35], in which parietal fenestrae appear at a very early ontogenetic stage. Scannella and Horner [10] proposed that the acquisition of fenestrae in Triceratops was associated with a ''striated'' surface texture at the caudal end of the parietal and a ''pebbly'' surface texture at the future site of the fenestrae. Simultaneously, the ventral surface of the parietal developed shallow depressions which later transformed into full fenestrae. Specimens of both Triceratops (AMNH 5116; [10]: fig. 5C) and Torosaurus (MOR 981; [10]: fig. 5D, presumably early within the transition) do indeed show this bone texture. Thus, assuming that the Nedoceratops hatcheri holotype skull, USNM 2412, is transitional between the two morphs, both the striated/pebbly bone texture as well as ventral depressions should be visible on the parietal in this specimen. Yet, USNM 2412 instead shows well-developed neurovascular impressions rather than striated or pebbly bone across the entire preserved ventral and dorsal surfaces of the frill (Figures 5,6), and no depression occurs around the preserved parietal fenestra (although the parietal is very thin in this area). It is far more likely that the depressions on the ventral surface of the parietal observed in some specimens of Triceratops correspond to insertion areas for cervical musculature [41]. Thus, the ''incipient fenestra'' in MOR 2946 ( [10]: fig. 1C) may instead represent an area for muscle attachment.
Although bone resorption (associated with a ''mottled'' or ''pebbly'' texture; [10,42,43]) occurs on both the dorsal and ventral surfaces of the frill in Triceratops, this texture is not unambiguously associated with formation of fenestrae in other ceratopsids. In fact, it is probably a general feature of cranial bone growth in the clade. For instance, this texture also occurs on the squamosals of Centrosaurus, elements which are not normally fenestrated [43], as well as on the midline of the parietal [42]. Furthermore, ontogenetic evidence for Centrosaurus strongly suggests that cranial bone (particularly in the frill) passed through three sequential ontogenetic stages: 1) long-grained texture; 2) mottled texture; and 3) adult texture (characterized by a rugose surface occasionally with neurovascular impressions; [32,42]). Assuming that this is the case in Triceratops (based on the fact that cranial elements from obvious juveniles lack the adult bone texture), and assuming the OTH is correct, finding definitive Triceratops skulls with ''adult'' bone texture across the entire parietal should not be likely. Yet, numerous examples of this condition exist (e.g., USNM 2100, YPM 1822). The fact that a specimen of Torosaurus latus, MOR 1122, shows the most mature bone texture in a histological sample of five individuals [10] is intriguing, but the published sample size is too small to confirm that this is a consistent histological feature of the morphotype or that ''full adult'' Triceratops lack this texture. Additional histological sampling is needed to address this question.
In sum, the OTH requires that Triceratops underwent a sequence of ontogenetic changes that was completely unique among ceratopsids. Addition of numerous epiossifications, acquisition of parietal fenestrae at a very late ontogenetic stage, and reversion of adult bone texture to mottled bone texture and finally a return to adult texture during ontogeny are unlikely (although certainly not impossible). Furthermore, the Nedoceratops hatcheri skull USNM 2412 does not present an intermediate ontogenetic step between Triceratops and Torosaurus morphotypes, particularly based on the rugose surface texture on the parietal in this specimen. A perceived lack of juvenile Torosaurus specimens may simply reflect the fact that this taxon is quite rare, rather than that such specimens only occur as Triceratops. Indeed, baby and juvenile specimens (sensu [33]) of Triceratops (which is otherwise quite common in the Hell Creek Formation) were virtually unknown and unrecognized until recently [44]. YPM 1831, a partial skull, is one possible candidate for a subadult Torosaurus latus. This specimen displays a number of subadult characteristics, including an epinasal that is not fused with the underlying nasals, an epijugal that is not fused to the jugal, and open sutures between the exoccipitals and basioccipital of the occipital condyle [1]. Furthermore, the bone on the surface of the frill is smooth (consistent with subadult status) and no epiossifications are readily visible (suggesting that they may have been disarticulated prior to burial). Thus, it is far more likely that Triceratops, Torosaurus, and probably Nedoceratops, are distinct taxa.

Conclusions
Despite its convoluted taxonomic history, Nedoceratops hatcheri does indeed display several features that distinguish it from typical Triceratops and Torosaurus specimens, as well as other chasmosaurines (such as the profile of the squamosal, lack of a nasal horn, and presence of small parietal fenestrae). Even if N. hatcheri represents an aberrant Triceratops, the anatomy of N. hatcheri is inconsistent with the hypothesis that it is a transitional form between the ''young adult'' (classic Triceratops) and ''old adult'' (classic Torosaurus) morphotypes of a single taxon. Unless Triceratops underwent ontogenetic changes radically different from any other known ceratopsid, it seems most likely that the latter two taxa are also distinct from each other.
In a broader context, the number of valid ceratopsid species living in North America during the latest Cretaceous is of considerable interest for interpreting changes in dinosaur diversity (if any) just prior to their extinction at the Cretaceous-Paleocene boundary. Assuming that Triceratops, Torosaurus, Nedoceratops, and possibly Tatankaceratops are all valid taxa, raw species richness for chasmosaurines in the northern Western Interior of North America during the late Maastrichtian meets or exceeds that for the late Campanian of Alberta. Yet, by sheer number of specimens, Triceratops is clearly most common in the collected sample. The rarity of other chasmosaurines may reflect a true regional predominance of Triceratops, local rarity of non-Triceratops in the Hell Creek and Lance Formations due to ecological factors (i.e., other taxa were more common elsewhere), or a taphonomic artifact.

Materials and Methods
The holotype skull of Diceratops hatcheri, USNM 2412, was studied first-hand on three different occasions in order to map areas of reconstruction, verify morphology, and obtain photographs. Measurements were recorded to the nearest millimeter using a cloth measuring tape and digital calipers.