A chasmosaurine ceratopsid premaxilla from the basal sandstone of the Hell Creek Formation, Montana

: A well-preserved large chasmosaurine ceratopsid premaxilla (MOR 1122 7-22-00-1) collected from the basal sandstone of the Cretaceous Hell Creek Formation (HCF) represents one of the stratigraphically lowest ceratopsid occurrences in the formation. The specimen was discovered in 2000, during the excavation of a large Torosaurus latus skull (MOR 1122) which was later hypothesized to represent an advanced growth stage of the more commonly recovered HCF ceratopsid Triceratops . MOR 1122 7-22-00-1 compares favorably with the incomplete premaxillae of the MOR 1122 skull and reveals details of premaxilla morphology of ceratopsids from this stratigraphic zone. It preserves large, closely spaced ventromedial foramina, a narrow triangular process, and a thin septal flange at the base of the narial strut. The nasal process is narrow, caudally inclined and has a forked dorsal surface which appears to represent a morphology intermediate between that of the slightly stratigraphically lower ceratopsid Eotriceratops xerinsularis from the Horseshoe Canyon Formation of Alberta and specimens recovered higher in the HCF. MOR 1122 7-22-00-1 expresses a deep recess extending medial to the strut of the triangular process, a feature shared with other HCF ceratopsid specimens but not Eotriceratops or other earlier occurring triceratopsin taxa. The morphology of MOR 1122 7-22-00-1 is consistent with noted stratigraphic trends in HCF ceratopsids and highlights the increased complexity of the narial region in uppermost Cretaceous triceratopsins.


INTRODUCTION
The uppermost Cretaceous Hell Creek Formation (HCF) of Montana and surrounding regions contains the remains of some of the last non-avian dinosaurs (hereafter referred to as 'dinosaurs') to roam western North America Clemens and Hartman 2014). An extensive survey of dinosaurs in the HCF near Fort Peck Lake in eastern Montana revealed that approximately 40% of dinosaur skeletons recorded represent the chasmosaurine ceratopsid Triceratops . The abundance of Triceratops in the HCF, combined with detailed locality data for most specimens, allows these fossils to be placed within the stratigraphic framework of the formation. Two recognized species (Triceratops horridus and T. prorsus  are stratigraphically separated, with T. prorsus being found in the upper unit (U3, sensu  and T. horridus is restricted to the lower unit (L3) and the lower part of the middle unit (M3) . Data are consistent A chasmosaurine ceratopsid premaxilla from the basal sandstone of the Hell Creek Formation, Montana Abstract: A well-preserved large chasmosaurine ceratopsid premaxilla (MOR 1122 7-22-00-1) collected from the basal sandstone of the Cretaceous Hell Creek Formation (HCF) represents one of the stratigraphically lowest ceratopsid occurrences in the formation. The specimen was discovered in 2000, during the excavation of a large Torosaurus latus skull (MOR 1122) which was later hypothesized to represent an advanced growth stage of the more commonly recovered HCF ceratopsid Triceratops. MOR 1122 7-22-00-1 compares favorably with the incomplete premaxillae of the MOR 1122 skull and reveals details of premaxilla morphology of ceratopsids from this stratigraphic zone. It preserves large, closely spaced ventromedial foramina, a narrow triangular process, and a thin septal flange at the base of the narial strut. The nasal process is narrow, caudally inclined and has a forked dorsal surface which appears to represent a morphology intermediate between that of the slightly stratigraphically lower ceratopsid Eotriceratops xerinsularis from the Horseshoe Canyon Formation of Alberta and specimens recovered higher in the HCF. MOR 1122 7-22-00-1 expresses a deep recess extending medial to the strut of the triangular process, a feature shared with other HCF ceratopsid specimens but not Eotriceratops or other earlier occurring triceratopsin taxa. The morphology of MOR 1122 7-22-00-1 is consistent with noted stratigraphic trends in HCF ceratopsids and highlights the increased complexity of the narial region in uppermost Cretaceous triceratopsins.
with the presence of an anagenetic lineage of Triceratops in which T. horridus evolved into T. prorsus over the course of the latest Cretaceous . Although Triceratops is extremely common in the HCF (e.g., Brown 1917;, remains of this animal are less common in the lower unit of the formation . The large HCF Triceratops dataset includes members of different ontogenetic stages and reveals a dramatic cranial transformation throughout growth Goodwin 2006, 2008). As individuals matured, the postorbital horn cores changed orientation (from caudally to rostrally curved) and the initially triangular epiossifications of the parietal-squamosal frill flattened onto the frill margin Goodwin 2006, 2008;Wilson and Fowler 2017). Further morphological and histological evidence indicates that the parietal-squamosal frill of Triceratops underwent expansion and eventual fenestration, resulting in a mature morphology previously considered to represent the distinct chasmosaurine Torosaurus latus (Scannella andHorner 2010, 2011;Horner and Lamm 2011). This synonymy hypothesis has been the subject of ongoing study and Torosaurus is considered a distinct taxon by other authors (see Farke 2011;Longrich and Field 2012;Maiorino et al. 2013). Specimens previously referred to Torosaurus latus are relatively rare, being represented by fewer than 20 individuals (e.g., Marsh 1891;Colbert and Bump 1947;Farke 2007;Scannella and Horner 2010;Longrich and Field 2012;McDonald et al. 2015) and are more commonly recovered from the lower half of the HCF . In 2000, Merl and Gladys Busenbark discovered Museum of the Rockies (MOR) locality HC-258 ('TORO II') in the basal sandstone of the HCF in Fergus County, Montana, USA Fig. 1). A large, well-preserved partial ceratopsid skull representing one of the most complete examples of 'Torosaurus' at the time was collected from the site (Farke 2007;Scannella and Horner 2010). Initial excavation was undertaken by Ken Olson and MOR; later excavation was completed by Bob Harmon, Nels Peterson, and the 2000 MOR paleontology field crew. This specimen (MOR 1122;Fig. 2) represents one of the stratigraphically lowest ceratopsid specimens recovered from the HCF of Montana . In addition to the MOR 1122 skull, The TORO II quarry also produced the well preserved right premaxilla of a second chasmosaurine ceratopsid (MOR 1122 7-22-00-1, see Figs. 3, 4;Farke 2007). As the premaxillae of MOR 1122 are only partially preserved (Farke 2007), MOR 1122 7-22-00-1 represents the stratigraphically lowest occurring well-preserved ceratopsid premaxilla from the HCF. Premaxillae are taxonomically informative in chasmosaurine ceratopsids (e.g., Wu et al. 2007;Sampson et al. 2010;Loewen et al. 2010) including Triceratops. In Triceratops prorsus, the premaxilla contributes to a convex rostrum typically supporting an elongate nasal horn whereas in Triceratops horridus, the rostrum is more elongate with the caudally inclined nasal process of the premaxilla (NPP) typically supporting a smaller nasal horn Longrich and Field 2012; fig. 1; Fig. 5). The base of the HCF has a maximum age of 68 Ma (Fowler 2017). Eotriceratops xerinsularis, a slightly earlier occurring triceratopsin (~68.8 Ma, see Wu et al. 2007;Longrich 2011;Eberth and Kamo 2020) from the Horseshoe Canyon Formation of Alberta, is primarily distinguished from Triceratops based on the morphology of its premaxilla. Given the rarity of well-preserved ceratopsid premaxillae from the basal sandstone of the HCF, MOR 1122 7-22-00-1 is described here and compared with other latest Maastrichtian specimens.

MATERIALS AND METHODS
The specimen MOR 1122 7-22-00-1 was examined at MOR. Measurements were taken using a combination of digital calipers, sliding metal calipers, and a tape measure (measurements over 50 cm). Specimen photos were taken primarily using either a Nikon 1 J2 or Nikon Coolpix B500 camera. Photographs were prepared into figures using Adobe Photoshop and Adobe Illustrator (CC 2018). HCF ceratopsid material at MOR (see  was examined for direct comparison with MOR 1122 7-22-00-1. Additional material was studied at YPM and the holotype of Eotriceratops (TMP 2002.057.007) was examined at TMP. In order to assess the systematic position of MOR 1122 7-22-00-1, it was added to the specimen level cladistic analysis of HCF ceratopsid specimens conducted in . The analysis was performed using PAUP* 4.0a (build 168; Swofford 2002) and cladograms were displayed using FigTree (Rambaut 2012). MOR 1122 7-22-00-1 could be coded for 8 of 30 morphological characters (multistate character list from  presented in Appendix 1, codings presented in Appendix 2). Initially, a branch-and-bound search was performed with MOR 1122 7-22-00-1 added to the Scannella et al. 2014 specimen set which excludes individuals that cannot be coded for at least 10 characters or characters of the frill (18 specimens with Arrhinoceratops designated as the outgroup , fig. 3B]). Additional analyses were performed using the reduced specimen set (following removal of MOR 2924 from the matrix [Scannella et al. 2014, fig. 3C] and with an alternative coding of the positioning of ventromedial foramina in MOR 981 (see below). An analysis was also performed on the reduced specimen set using only charac-ters of the premaxilla and excluding specimens which could not be coded for at least three characters.

Geological background
The HCF is divided into three units: the lower third (L3), middle third (M3), and upper third (U3) (Fowler 2009;Fig. 1). At the base of the formation is an amalgamated channel sandstone termed the basal sandstone, first noted by Brown (1907) and subsequently examined by authors further describing the stratigraphy of the formation (e.g., Flight 2004;Hartman et al. 2014;Fowler 2016Fowler , 2017. The base of the HCF has a maximum age of 68 Ma (Fowler 2017). MOR locality HC-258 is positioned within the basal sandstone ), approximately three meters above the Colgate Member of the underlying Fox Hills Formation (J. Horner, pers. comm. 2019). Dinosaur fossils are relatively rare in the basal sandstone compared to strata higher in the HCF ; however, some exceptionally preserved specimens have been recovered from this sandstone including a partial articulated Edmontosaurus (MOR 1142) and a specimen of Tyrannosaurus rex which preserves evidence of soft tissues (MOR 1125;Schweitzer et al. 2005;). In addition to remains of the two ceratopsid specimens, the HC-258 quarry also preserved a hadrosaur cervical vertebra (MOR 1122(MOR C-2020 and assorted microvertebrate material including a partial theropod phalanx, partial crocodilian osteoderm, piece of trionychid shell, and fragmentary theropod tooth. DESCRIPTION MOR 1122 7-22-00-1 ( Fig. 3; Tab. 1) is well preserved and nearly complete; it is missing only the caudoventral 'prong' and a portion of the medial surface which permits a view of the interior chambers. Some of the thin bone surrounding the septal fossa is missing dorsally, but the rostral margin is preserved. The maximum preserved length is 54.1 cm ( Fig. 4; Tab. 1). A pronounced interpremaxillary process divides the ventral margin of the interpremaxillary fenestra into two regions. The caudal region has a maximum width of 7.7 cm and the smaller rostral portion is approximately 5.8 cm wide. This is unlike the condition in the stratigraphically lower Eotriceratops, in which the interpremaxillary process appears contiguous with the rostroventral margin of the fenestra, but closely resembles the condition in Triceratops found higher in the formation (e.g., MOR 1120, MOR 1625; Fig. 5). Ventral to the interpremaxillary fenestra, several sulci descend from the triangular process towards a shallow depression ventral to the rostral-most portion of the fenestra. The rostral-most portion of this depression exhibits slightly mottled surface texture, similar to that noted on some specimens of ceratopsid frill (e.g., Brown et al. 2009;Scannella and Horner 2010). A single marked sulcus, which is parallel to those descending from the triangular process, extends from the depression to the rostroventral margin of the premaxilla. Much of the ventral portion of the lateral surface of the premaxilla exhibits a subtle rugose texture with many faint, obliquely oriented striations. Several small foramina are present on the lateral surface of the bone ventral to the interpremaxillary fenestra. A pronounced sulcus located approximately 3 cm ventral to the interpremaxillary process leads to a relatively large foramen (maximum width approximately 1 cm). A similarly sized foramen is positioned approximately 4 cm ventral to the caudal margin of the interpremaxillary fenestra. The narial strut is relatively broad ventrally and tapers dorsally. Fine striations present on the lateral surface of the strut are oriented roughly parallel to the ventral margin of the premaxilla. A septal flange is present along the caudal margin of the base of the narial strut. The absence of this septal flange has been used to distinguish Triceratops from some other chasmosaurines, including Pentaceratops and Chasmosaurus (e.g., Forster et al. 1993). However, a septal flange restricted to the base of the narial strut is here noted to be present in some specimens of Triceratops, including juveniles (e.g., MOR 1199) and the large subadult YPM 1821 (Fig. 6). Several small foramina are present just lateral to the septal flange. The triangular process (narial process of Wu et al. 2007) bears a prominent dorsocaudally directed strut-like projection which tapers caudally. This process is lateral to the septal flange and exhibits a recess on its lateral surface. The dorsal margin of the triangular process is positioned slightly above the ventral margin of the interpremaxillary fenestra. This is consistent with the position noted in some specimens of Triceratops (e.g., AMNH FARB 5116, YPM 1821 ); the dorsal margin of the triangular process is more distinctly elevated above the ventral margin of the interpremaxillary fenestra in Eotriceratops than in Triceratops ). A deep recess is positioned ventromedial to the strut of the triangular process; this feature is shared with Triceratops found higher in the HCF but is not present (or is greatly reduced) in the stratigraphically lower Maastrichtian triceratopsin Eotriceratops . Similarly, this feature is not present in Regaliceratops peterhewsi (C. Brown, pers. comm. 2020), which has been recovered within and more recently outside of triceratopsini (Brown and Henderson 2015; Mallon et al. 2016). The nasal process of the premaxilla (NPP) is elongate, narrow, and caudally inclined, meeting the narial strut at an angle of approximately 124 degrees ). It bears a shallow facet for articulation with the nasal; the facet exhibits patches of striated and mottled surface textures as well as a series of shallow sulci oriented roughly parallel to the primary trend of the NPP. The dorsal surface of the NPP is forked, bearing a pronounced caudal projection and a smaller rostral projection. These projections are separated by a shallow fossa. In Eotriceratops, which slightly predates the HCF ), the NPP is similarly bifurcated although the rostral projection is less pronounced (Fig. 7). In Eotriceratops, the NPP is in line with the rostrodorsal surface of the premaxilla, whereas in MOR 1122 7-22-00-1 it is deflected slightly dorsally, consistent with the morphology of T. horridus . In some HCF ceratopsid specimens (e.g., MOR 1120, MOR 1625), a prominence is noted just rostral to or descending from the narial strut and directed into the interpremaxillary fenestra . In MOR 1122 7-22-00-1, there are at least two closely spaced struts directed into the interpremaxillary fenestra, the caudal-most of which originates well rostral to the narial strut, at the rostral-most extent of the facet on the NPP for articulation with the nasal. The caudal-most strut or prominence is similarly placed rostral to the narial strut in YPM 1820, the holotype of Triceratops horridus (Fig. 8), and MOR 3011, a specimen from M3 of the HCF ).  Measurements presented in Table 1. Abbreviations: NPPL, nasal process of the premaxilla length; NPPW, nasal process of the premaxilla width; IFL, interpremaxillary fenestra length; DBN, dorsoventral breadth of narial strut; DBT, dorsoventral breadth of triangular process; MPL, maximum preserved length. Much of the internal anatomy of the premaxilla is visible in medial view (Fig. 3B). The medial surface of the NPP is relatively flat and smooth where it would have contacted the opposite premaxilla. A prominent sulcus (here termed the interpremaxillary channel) extends from the rostral portion of the interpremaxillary fenestra towards a series of thin septa which separate the sulcus from dorsomedial and ventromedial premaxillary chambers. The dorsomedial chamber is approximately 15 cm long and the ventromedial chamber extends to the caudal-most preserved extent of the premaxilla. The chambers are separated caudally by a recess located medial to the triangular process (Fig. 3B). The dorsomedial chamber is bound rostrodorsally by a ridge caudoventral to the interpremaxillary fenestra. Two smaller (approximately 1 cm wide), shallow sulci extend from the ventral surface onto the lateral surface of the interpremaxillary channel. The caudal-most sulcus connects the interpremaxillary channel to the dorsomedial chamber and leads to the largest foramen noted on the lateral surface, ventral to the interpremaxillary fenestra. The two largest ventromedial foramina (here termed the primary and secondary foramen) are contiguous with the premaxillary chambers (Fig. 3B). The larger primary foramen (approximately 4 cm wide) leads into the interpremaxillary channel which connects to the dorsomedial chamber; the smaller secondary foramen (approximately 2.8 cm wide) leads directly into the ventromedial chamber. These foramina are closely spaced (2.9 cm distance between the caudal margin of primary foramen and rostral margin of the secondary foramen). Although the right premaxilla of the articulated MOR 1122 skull is incomplete, the morphology of the ventromedial foramina compares favorably with MOR 1122 7-22-00-1 (Figs. 9, 10B, C). In MOR 1122, these foramina are 3.4 cm apart. The primary ventromedial foramen is partially bordered rostrally by an obliquely oriented margin (Fig. 9). Though incomplete, it indicates a minimum diameter of 2.4 cm for the primary ventromedial foramen. The ventromedial foramina of MOR 3081, a 'Torosaurus' specimen from the upper part of the lower HCF, are also relatively large and closely spaced (Fig. 10D). The stratigraphically lower Eotriceratops also exhibits closely spaced ventromedial foramina ( Fig. 10A; . Triceratops recovered from higher in the HCF typically exhibit more widely spaced ventromedial foramina (Figs. 10E−G, 11; Table 2); though in some specimens of Triceratops the foramina may be more closely spaced (Hatcher et al. 1907, fig. 28). A small additional foramen is located in the caudolateral margin of the primary foramen of MOR 1122-7-22-00-1; a shallow sulcus extends rostroventrally from this foramen. Approximately 3.8 cm rostral to the primary foramen, an additional foramen is present and extends into a relatively wide (approximately 2 cm) sulcus rostroventrally.
Ventrally, a groove for articulation with the rostral bone extends lateral to the ventromedial foramina for approximately

PHYLOGENETIC ANALYSIS
In order to assess the systematic position of MOR 1122 7-22-00-1, it was added to the specimen level cladistic analysis of HCF ceratopsid specimens conducted in . Initially, a branch-and-bound search was performed with MOR 1122 7-22-00-1 added to the Scannella et al. 2014 specimen set which excludes individuals that cannot be coded for at least 10 characters or characters of the frill. The analysis resulted in 49,252 most parsimonious trees with a length of 53 steps, a Consistency Index of 0.74 and a Retention Index of 0.82 (Fig. 12A). Eotriceratops was recovered as the most basal member of the ingroup, but in an unresolved position relative to Arrhinoceratops. The strict consensus recovers MOR 1122 7-22-00-1 in a polytomy with specimens exhibiting the expanded 'Torosaurus' frill morphology (MOR 981, MOR 1122, MOR 3081). These specimens exhibit a caudally oriented NPP combined with closely spaced ventromedial foramina (the position of  these foramina in MOR 981 was initially coded as '?' (see below). This polytomy is recovered as basal to a stratigraphic succession of other specimens including a large polytomy of specimens from upper M3 and U3. MOR 1122 and MOR 3081 were collected from L3, but the exact stratigraphic position of MOR 981 is unknown (Scannella et al. 2014). MOR 1120, a subadult specimen collected from the upper part of L3, exhibits a caudally oriented NPP and ventromedial foramina which are relatively more widely spaced than those preserved in specimens lower in the formation ). Specimens recovered from the upper half of the formation express a more vertically oriented NPP (character 3). A branch-and-bound search performed using the reduced specimen set (following removal of MOR 2924 from the matrix   fig. 3C]) resulted in 1,974 most parsimonious trees with a length of 52 steps, a Consistency Index of 0.75 and a Retention Index of 0.82 (Fig. 12B). The strict consensus tree distinguished upper M3 specimens (MOR 3027, UCMP 113697, MOR 3045) from U3 specimens (Fig. 9B). MOR 3045 and U3 specimens exhibit an expanded NPP (character 13). Large primary foramina are preserved in MOR 981; however, the relationships of these foramina to surrounding features are somewhat obscured by crushing (see Fig. 13). As such, the character describing the positioning of these foramina (character 19;) was initially coded as '?' following . Farke (2007 noted the presence of foramina positioned "just rostral to the maxillae" in MOR 981. A version of the analysis using the reduced data set was run with the ventromedial foramina of this specimen coded as being positioned far apart. This analysis resulted in 3,384 most parsimonious trees with a length of 53 steps, a Consistency Index of 0.74 and a Retention Index of 0.81. The strict consensus tree recovers MOR 981 and all HCF specimens from the lower half of the formation (including MOR 1122 7-22-00-1) in a polytomy (Fig. 12C). An analysis of the reduced dataset using only premaxilla characters was run after removal of specimens which did not exhibit at least three characters. This analysis resulted in 528 most parsimonious trees with a length of 11 steps, a Consistency Index of 0.91 and a Retention Index of 0.93. The strict consensus tree recovers MOR 1122 7-22-01-1 and MOR 1122 as basal to MOR 1120 and MOR 2982 (from upper L3 and lower M3, respectively) and a polytomy of specimens from the upper half of the formation (Fig.  12D). MOR 1122 7-22-00-1 and MOR 1122 share a deep recess of the triangular process but do not exhibit a prominence just rostral to or descending from the narial strut that is observed in some specimens recovered from higher in the formation . Specimens from the upper half of the formation are united by the expression of a more vertically oriented NPP (character 3).

DISCUSSION
The fact that MOR 1122 7-22-00-1 is unfused to surrounding cranial bones might suggest ontogenetic immaturity despite its large size, (see   Table 2. VF, ventromedial foramina. Composed using Microsoft Excel. Table 2. Measurements of ventromedial foramina (VF) for specimens in Figure 10. Diameter (measured rostrocaudally) and distance between the closest points of the primary and secondary VF. Data presented graphically in Figure 11. Measurements in cm.

Premaxilla
Primary Cladistic analyses recover MOR 1122 7-22-00-1 with HCF specimens exhibiting the expanded 'Torosaurus' frill morphology (Fig. 12). MOR 1122 7-22-00-1 was found associated with MOR 1122, in which the premaxillae are not well preserved (Farke 2007); however, that skull expresses a narrow, caudally inclined strut extending from the triangular process and closely spaced ventromedial foramina, consistent with the morphology of MOR 1122 7-22-00-1 (Fig. 10). The incomplete, obliquely oriented rostral margin of the primary ventromedial foramen is positioned rostrodorsal to the secondary foramen and extends from within a subtle fossa on the medial surface of the premaxilla (Fig. 9). This may suggest that this margin does not indicate the full extent of the foramen but instead represents a subdivision of a larger primary foramen. Similarly, a smaller foramen is noted within the primary ventromedial foramen of MOR 1122 7-22-00-1. Closely spaced ventromedial foramina are also observed in MOR 3081, a specimen that exhibits the expanded 'Torosaurus' frill morphology and was collected from the mudstones of the upper part of L3 . MOR 981, another large 'Torosaurus' skull, expresses a very large (7 cm long) primary foramen on each premaxilla. The relationship of these foramina to surrounding features is partly obscured by crushing (Fig. 13). Farke (2007) (Ullmann and Lacovara 2016). If Triceratops and 'Torosaurus' are distinct but closely related taxa, the morphology of these foramina might suggest MOR 1122 7-22-00-1 holds affinities with the latter. However, MOR 981 demonstrates that the size and position of these foramina may not be consistent in all 'Torosaurus' specimens. Further, the placement of the ventromedial foramina can vary between the right and left sides of unfenestrated specimens (e.g., YPM 1821 figured in Hatcher et al. 1907, fig. 28). Similarly, the right premaxilla of TMP 2002.057.0007 (Eotriceratops) exhibits two closely spaced ventromedial foramina whereas the left premaxilla expresses three. Despite this variation, relatively large, closely spaced ventromedial foramina appear to be more commonly recovered low in the HCF ( Fig. 10; ). The morphology of MOR 1122 7-22-00-1 is consistent with its stratigraphic position in the basal sandstone of the HCF. Eotriceratops, recovered from Unit 5 of the Horseshoe Canyon Formation (~ 68.8 Ma;Wu et al. 2007;Eberth and Kamo 2020), has a comparatively simple premaxilla. Eotriceratops is diagnosed by its tall, unrecessed triangular process of the premaxilla; a feature which distinguishes it from all HCF specimens recorded to date. The premaxilla of Eotriceratops further does not exhibit a septal flange and the base of the narial strut is not expanded (Fig. 5). In contrast, MOR 1122 7-22-00-1 has a deeply recessed triangular process, a narial strut with an expanded base, and a septal flange positioned along the base of the strut. The recessed triangular process and expanded narial strut morphologies are consistent with other HCF specimens. The presence of a septal flange appears to become less common in HCF ceratopsids, though it is expressed in early ontogeny. MOR 1122 7-22-00-1 is distinguishable from Triceratops prorsus, which is found in the upper unit of the HCF and expresses a much wider and more upright NPP Longrich and Field 2012;. It compares favorably with specimens of Triceratops horridus from the lower half of the HCF, in which the NPP is narrow and caudally inclined. The NPP is not everted caudally to the extent observed in Eotriceratops and other earlier occurring chasmosaurines where this process is aligned with the rostral ramus of the premaxilla. The forked surface of the NPP of MOR 1122 7-22-00-1 appears to be intermediate between the morphology of Eotriceratops and specimens found stratigraphically higher in the HCF (Fig. 7); however, direct comparisons to other specimens from the lower half of the HCF are complicated by the fact that the morphology of this process in similarly large individuals is unknown due to non-preservation (e.g., MOR 1186), taphonomic damage (e.g., MOR 6653; ), or obfuscation due to fusion with adjacent elements (e.g., MOR 981, MOR 1122). Similarly, details of the morphology of the NPP in the holotype of Triceratops horridus (YPM 1820) are partially obscured due to articulation with the nasals (Fig. 8) and the specimen does not preserve the ventromedial foramina of the premaxillae. These factors highlight the significance of MOR 1122 7-22-00-1 as a well preserved specimen from a critical stratigraphic zone. In MOR 1122 7-22-00-1, the dorsal surface of the triangular process is elevated above the ventral White arrows indicate the enlarged primary foramina. Yellow arrow indicates the possible position of the secondary foramen, well caudal to the primary foramina (Farke 2007). Scale, 10 cm. margin of the interpremaxillary fenestra, but not quite to the degree observed in Eotriceratops. The result is a deeper ventral portion of the premaxilla compared to specimens from higher in the formation (Fig. 5). It is possible that MOR 1122 7-22-00-1 represents a transitional morphology between the stratigraphically lower Eotriceratops and specimens found higher in the HCF. Unfortunately, as this individual is represented by a single bone, details of the morphology of the rest of the animal and how they might compare to the associated MOR 1122 skull and other HCF specimens are unknown. It is here referred to Triceratops sp.; the discovery of additional well-preserved specimens from the lower HCF will further resolve the range of variation within premaxillae from these strata and allow for direct comparisons with large, unfused specimens Fowler 2017). The elaborate morphology of the premaxilla is one of the features that distinguishes chasmosaurine ceratopsids from centrosaurines, in which the premaxilla is a comparatively simple bone (Dodson et al. 2004). MOR 1122 7-22-00-1 highlights an apparent increase in complexity of the narial region relative to earlier occurring chasmosaurines. Witmer (2001) hypothesized that a rostroventral placement of the nostril in Triceratops would provide airflow over the full narial apparatus, facilitating physiological functions. The increased complexity of the chasmosaurine narial region at the onset of HCF deposition may indicate increased physiological or other capabilities in the latest occurring triceratopsins. Further exploration of the basal sandstone of the HCF will continue to clarify ontogenetic sequences throughout the formation, illuminate heterochronic trends, and test evolutionary hypotheses.

ACKNOWLEDGEMENTS
Thank you to the Busenbark and Weaver families for allowing specimens from HC-258 to be collected by MOR.