Foraminacephale brevis, (LAMBE, 1918)

FORAMINACEPHALE BREVIS ( LAMBE, 1918) ( FIGS 1 – 16 View Figure 1 View Figure 2 View Figure 3 View Figure 4 View Figure 5 View Figure 6 View Figure 7 View Figure 8 View Figure 9 View Figure 10 )

Holotype: CMN 1423 About CMN , fairly complete frontoparietal dome missing the anterior portion of the frontonasal boss.

Type locality and horizon: Upper Belly River Group (Dinosaur Park Formation).

Referred material: CMN 12351, parietal;? TMP P70.10.2 partial frontoparietal with articulated partial squamosal, postorbitals, and posterior supraorbitals; TMP 1985.043.0061, parietal; TMP 1986.077.0085, postorbital; TMP 1987.050.0029, frontoparietal; TMP 1992.036.1125, postorbital; UALVP 49440, squamosal. For a complete hypodigm see Sullivan (2003) in addition to Table S1.

Locality and horizon: The vast majority of known specimens assignable to this taxon have been recovered from the Dinosaur Park Formation ( Eberth & Hamblin, 1993), where stratigraphical data are sufficient to be certain from which host formation the specimen was derived (e.g. UALVP 49440). Sullivan (2003, 2006) reported that at least some specimens of this taxon were derived from the Oldman Formation (sensu Eberth & Hamblin, 1993); however, no specimen numbers were cited in support of this claim. Numerous specimens collected since the recognition of the Dinosaur Park Formation as a separate unit from the Oldman Formation ( Eberth & Hamblin, 1993) can be positively sourced to the Dinosaur Park Formation (e.g. TMP 99.55.122, TMP 2000.12.1). The only known specimen ( TMP 2015.044.0041) of F. brevis that can definitively be sourced to the Oldman Formation was recently collected near Onefour, Alberta, and occurs in Unit 3 (upper muddy unit) of the Oldman Formation, which is time equivalent to the Dinosaur Park Formation exposed at Dinosaur Provincial Park ( Eberth & Hamblin, 1993). While we observed? TMP P70.10.2 first hand when it was on loan to Robert Sullivan in 2010, we have not been able to confirm the disposition or provenance data of the specimen, despite contacting numerous museums and Robert Sullivan (pers. comm.).

Diagnosis: Pachycephalosaurine pachycephalosaurid characterized by the following autapomorphies: tall, smooth squamosal bar with a single row of six nodes in the primary node row and a single small corner node just ventral to node row; dorsal surface of frontoparietal and associated peripheral elements covered with small dorsoventrally orientated foramina resulting in a distinct pitted surface texture. Differs from St. validum , Hanssuesia sternbergi , and Colepiocephale lambei in the following characters: ‘down-turned’ posteromedial extension of parietal and closure of supratemporal fenestrae very early in ontogeny. Differs further from Sphaerotholus and Prenocephale in the following characters: prominent grooves on frontal between the supraorbital lobe and the frontonasal boss, and slit-like temporal chamber roof expressed in the parietal.

Comments: We remove CMN 11316 from F. brevis . The specimen was recovered from Horseshoe Canyon Formation exposures along the Red Deer River approximately 3.5 km north of Bleriot Ferry, and north of the town of Drumheller. The specimen consists of fragmentary conjoined frontals that are weakly inflated, and represents a small juvenile that cannot be identified beyond Pachycephalosaurinae.

COMPARATIVE DESCRIPTION

A detailed morphological description of the frontoparietal dome of F. brevis was provided by Sullivan (2000). Here we describe the squamosal and postorbital for the first time. Previously, F. brevis was only known from isolated frontoparietal domes that have a distinctive shape and pitted surface texture. The elements described below are identified as pertaining to F. brevis based on the morphology of the sutural contacts with the frontoparietal dome, and a shared pitted surface texture that is characteristic of the taxon. A previously undescribed specimen of F. brevis ,? TMP P70.10.2, preserves the large, dorsal portions of the postorbital and squamosals in articulation with a nearly complete frontoparietal that confirms the identification of these more complete, but isolated specimens. Combined, this new material also allows a nearly complete 3D reconstruction of the skull roof of F. brevis . We also describe small juvenile and mature adult material referable to F. brevis , and document the first frontoparietal ontogenetic growth series for this taxon.

SQUAMOSAL

An isolated squamosal ( UALVP 49440; Fig. 1A View Figure 1 ) is nearly complete, missing only parts of the ventral processes. It is most similar to the squamosals of Sp. buchholtzae ( TMP 87.113.3; Fig. 1B View Figure 1 ) and P. prenes (Z. Pal. MgD-I/104). The squamosal has sutural surfaces for the parietal medially and anteromedially and the postorbital anteriorly. The sutural surface is continuous, indicating that the supratemporal fenestrae were fully closed. The position and angle of the sutural surface is similar to P. prenes (Z. Pal. MgD-I/ 104) and to Sp. buchholtzae ( TMP 87.113.3). The squamosal would have articulated with the frontoparietal of F. brevis with the parietosquamosal bar at an angle in posterior view. The dorsal surface of the squamosal is fairly continuous with the dome, except for a dorsal expansion at the anterolateral edge. The squamosal of P. prenes (Z. Pal. MgD-I/104) has a similar expansion, although to a lesser degree, whereas the dorsal surface of Sp. buchholtzae ( TMP 87.113.3) lacks this feature. The dorsal surface of UALVP 49440 is very smooth, but is pierced by several small foramina that are identical to those found on the outer margins of F. brevis domes [e.g. CMN 121, CMN 1423 (holotype), and TMP 85.36.292]. Similar foramina are also present on the dorsal surfaces of the dome and peripheral elements of Sp. buchholtzae ( TMP 87.113.3), except to a much lesser degree, whereas P. prenes (Z. Pal. MgD-I/104) has a much more rugose surface texture, somewhat similar to the tuberculate texture of Stegoceras (e.g. CMN 138, TMP 85.4.1, UALVP 2), but also pierced by foramina.

The dorsal surface of the squamosal is separated from the posterior squamosal bar by a row of six similarly sized nodes, including the vertex node. The medial-most posterior node straddles the parietosquamosal suture. The posterior nodes are somewhat flattened posteriorly and pointed dorsally, similar to Sp. buchholtzae ( TMP 87.113.3) and unlike the posteriorly projecting conical nodes of P. prenes (Z. Pal. MgD-I/104) or the pyramidal nodes of St. validum (e.g. CMN 138, TMP 85.4.1, UALVP 2; Schott & Evans, 2012; Fig. 1C View Figure 1 ). The vertex node separates the posterior node row from the lateral node row that, on the squamosal, consists of a single elongated node. A similar, albeit less pronounced, lateral node is found on the squamosal of Sp. buchholtzae ( TMP 87.113.3), which also has a second lateral node that is even more elongate and less pronounced. A mediolateral crack, and associated strip of missing bone, lies along the ventral edge of the posterior node row and obscures the ventral bases of these nodes. Ventral to the vertex node is a lateroventral corner node, as in P. prenes (Z. Pal. MgD-I/104) and Sphaerotholus ( TMP 87.113.3, NMMNH P-27403). This node is slightly smaller than the posterior nodes, but much larger than the tiny corner node of Sp. buchholtzae ( TMP 87.113.3). In contrast, the corner node of P. prenes (Z. Pal. MgD-I/104) is much larger than the posterior nodes. Medial to the corner node is a dorsally angled row of three small, flat, node-like structures. This row may have continued, but is obscured by the strip of missing bone ventral to the primary node row.

Other than the corner node and the flat, node-like structures, the squamosal bar is smooth, devoid of the minute nodes found in St. validum (e.g. CMN 138, TMP 85.4.1, UALVP 2; Schott & Evans, 2012), and pierced by many small foramina adjacent to its ventral margin. The squamosal bar is proportionally much taller than that found in any other pachycephalosaur and maintains the same height throughout its length. This is most similar to P. prenes (Z. Pal. MgD-I/104), which has a tall squamosal bar, but proportionally is not as tall as the much smaller squamosal of F. brevis , and increases slightly in height laterally. In contrast, the squamosal of Sp. buchholtzae ( TMP 87.113.3) is very short and further shortens laterally. Ventrally, the squamosal bars of F. brevis and Sp. buchholtzae ( TMP 87.113.3) are very similar. Both curve dorsally and thin laterally. Most of the ventral surface is missing and the ventral processes of the squamosal are not preserved. The preserved posterior section of the temporal chamber has the slit-like morphology typical of F. brevis [e.g. CMN 121, CMN 1423 (holotype), and TMP 85.36.292] and St. validum (e.g. CMN 138, CMN 515, TMP 85.4.1, UALVP 2) that is distinct from the open morphology of Sp. buchholtzae ( TMP 87.113.3) and other pachycephalosaurids.

POSTORBITAL

Two postorbitals are here referred to F. brevis, TMP 86.77.85 and TMP 92.36.1125 ( Fig. 2A View Figure 2 ). They are both nearly complete, missing only the posterior portion of the lateral bar (both) and the anterior portion of the lateral bar ( TMP 86.77.85). As TMP 92.36.1125 is more complete the description focuses mainly on this specimen, but significant differences are noted.

The medial portion of the postorbital is convex, almost forming a separate dome. This is similar to the condition found in P. prenes (Z. Pal. MgD-I/104) and unlike the flat, dorsoventrally orientated wall formed by the postorbital of Sp. buchholtzae ( TMP 87.113.3; Fig. 2B View Figure 2 ). The dorsal surface is smooth, but pierced by the same small foramina as the squamosal and frontoparietal of F. brevis . Again, this differs from the somewhat rugose texture of the postorbital of P. prenes (Z. Pal. MgD-I/104), which still bears small foramina, and the smooth surface of the postorbital in Sp. buchholtzae ( TMP 87.113.3) that lacks the high density, and the distinctness, of the foramina. The sutural surfaces surrounding this portion of the postorbital are complete with contacts for the squamosal posteriorly, frontoparietal medially, and posterior supraorbital anteriorly. The sutural contacts are orientated such that the postorbital would be mostly confluent with the dome, similar to the condition in P. prenes (Z. Pal. MgD-I/104). The two specimens of F. brevis are similar in the length and height of the medial portion, but TMP 92.36.1125 is much thicker dorsoventrally. In medial view, the dorsal border of TMP 86.77.85 is anteroposteriorly convex whereas TMP 92.36.1125 is much more flat.

Lateroventrally to the medial domed portion of the postorbital is the lateral bar. It is separated from the medial portion by a low anteroposteriorly orientated ridge. This ridge corresponds to the lateral node row. Although no lateral nodes are present on these postorbitals, both are missing the posterior portion of the bar that contacts the squamosal, which, as described above, has a single lateral node. Thus it is possible that one or more nodes would have been present on the missing portion of the postorbital. The ridge in Sp. buchholtzae ( TMP 87.113.3) is less distinct than in F. brevis , but its postorbital also lacks lateral nodes. In P. prenes (Z. Pal. MgD-I/104) the ridge bears four lateral nodes posteriorly. The anterior portion of the ridge in P. prenes (Z. Pal. MgD-I/104) is formed by small tubercles. A second ridge, more prominent in TMP 86.77.85, runs anteroventrally from roughly the midpoint of the main ridge to the anteroventral edge of the postorbital. This ridge is absent in Sp. buchholtzae ( TMP 87.113.3), but in P. prenes (Z. Pal. MgD-I/104) extends from the anterior-most lateral node and is formed by partially connected nodes or tubercles. The surface texture of the lateral bar is somewhat smooth, but bears small tubercles concentrated around the second ridge. The posteroventral portion of the bar is pierced by several small foramina, but lacks the larger foramina found on the medial portion. Anteriorly, the lateral bar forms the posterior margin of the orbit. This is somewhat concave in TMP 86.77.85, but straight in TMP 92.36.1125. The posterior margin of the orbit is not preserved in Sp. buchholtzae ( TMP 87.113.3), but in P. prenes (Z. Pal. MgD-I/104) the morphology is very similar to TMP 92.36.1125. Posterior to the orbital margin, the lateral bar extends ventrally to contact the jugal and form the postorbital – jugal bar. Posterior to the bar, the postorbital forms the anterodorsal margin of the lateral temporal fenestra. This margin is concave, more so in TMP 92.36.1125 than TMP 86.77.85. In Sp. buchholtzae ( TMP 87.113.3) the margin is not as strongly concave and in P. prenes (Z. Pal. MgD-I/ 104) it is only slightly concave. The length of the lateral bar could not be ascertained in either specimen owing to breakage, but the heights of the bars could be compared. Whereas the medial portions of the two postorbitals were roughly the same height the lateral bar of TMP 92.36.1125 is much taller.

In ventral view, the postorbital contains part of the orbital cavity anteriorly. Medial to the orbital cavity is a notch for the articulation of the jugal. Medial to the jugal notch is the roof of the temporal chamber and at the posterior edge of the postorbital is the notch for the squamosal.

JUVENILE MATERIAL

Several new small specimens are here referred to F. brevis , including two small parietals and a frontoparietal dome that each show the diagnostic characters of F. brevis , but exhibit distinct juvenile morphologies.

Parietals: Two small, isolated parietals, CMN 12351 and TMP 85.43.68 ( Figs 3 View Figure 3 , 4 View Figure 4 ), are here referred to F. brevis and represent the smallest known specimens of this taxon. They are both nearly complete, missing only portions of the ventral processes. In TMP 85.43.68, the ventral processes are broken at their bases, but in CMN 12351 most of the processes are preserved, particularly on the left side where the process is nearly complete. TMP 85.43.68 is well preserved, but CMN 12351 is abraded.

The parietals are thickened, but only slightly domed, with both mediolateral and anteroposterior convexity. The dorsal surface texture of TMP 85.43.68 is somewhat more rugose than in larger specimens of F. brevis (e.g. TMP 85.36.292) whereas CMN 12351 is smooth as a result of the water-worn nature of the specimen. Both are covered with small foramina, which, although smaller, are equivalent to those in larger specimens. Anteriorly, the parietal bears the sutural surface for articulation with the frontals. The sutural surface for the postorbital is preserved at the anterolateral edge of the parietal. This surface is extremely small, but proportionally not very different from those in larger specimens. Posterior to the postorbital sutural surface is finished bone surface that corresponds to the anteromedial margin of the supratemporal fenestra. In all larger specimens (except for CMN 12351) the supratemporal fenestra are closed and this area of the parietal instead bears the sutural surface for the squamosal. The margin of the supratemporal fenestra angles strongly medially until it straightens out posteriorly where it meets the medial extension of the parietal. The size of the open supratemporal fenestra would have been proportionally large as the margin encompasses more than half the total length of the parietal. The lateral edges of the medial extension of the parietal are straight and formed at the anterior end by the margin of the supratemporal fenestra and posteriorly by the sutural surface for the squamosal. The sutural surface for the squamosal is missing in TMP 85.43.68, but is discernible in CMN 12351.

The medial extension of the parietal is not as ‘down-turned’ as in larger specimens of F. brevis (e.g. TMP 85.36.292), but rather continues the slight curvature of the parietal ventrally. Despite this, the morphology of the supratemporal border and medial extension of the parietal are clearly distinguishable from the flat parietals of St. validum [e.g. UCMZ ( VP) 2008.01, UCMZ( VP) 2008.02; Schott et al. 2011]. These parietals all have triangular medial extensions and the supratemporal fenestrae are much smaller and circular. In lateral view, CMN 12351 reveals the dorsoventrally narrow, slit-like temporal chamber characteristic of F. brevis and St. validum (e.g. CMN 138, CMN 515, UALVP 2). TMP 85.43.68 has a similar morphology, although the ventral process of the parietal is broken at its base. In CMN 12351, the ventral process of the parietal expands ventrally and laterally to form part of the ventral surface of the temporal chamber. On the left side of the parietal the process is nearly complete and a portion of the sutural surface for the squamosal, which would have contacted with the lateral edge of the ventral process, is discernible, albeit heavily water worn.

Frontoparietals: Two nearly complete frontoparietals, TMP 91.36.265 ( Figs 5 View Figure 5 , 6 View Figure 6 ) and TMP 99.55.122 ( Figs 7 View Figure 7 , 8 View Figure 8 ), are identified here as juvenile F. brevis owing to their small size and distinct morphologies. These specimens were previously referred to the species F. brevis by Sullivan (2003), but their juvenile morphology was not noted or described. The specimens are larger than the small isolated parietals described above, but smaller than other specimens of F. brevis . These frontoparietals have domed parietals with closed supratemporal fenestrae, but anteriorly uninflated frontals. Of the two specimens described below, TMP 91.36.265 is slightly smaller (frontoparietal length 57.1 mm; compared with 61.1 mm for TMP 99.55.122), but thicker (frontoparietal thickness 27 mm, compared with 24 mm for TMP 99.55.122). TMP 91.36.265 has thinner frontals that are almost flat (height of nasal suture 2.1 mm compared with 4.3 mm in TMP 99.55.122).

In dorsal view, the specimens are easily identifiable as F. brevis by the general morphology of the frontoparietal, the presence of foramina typical of F. brevis , and the ‘down-turned’ medial extension of the parietal with slit-like temporal chambers. The dorsal surface texture of the frontoparietal is somewhat more rugose than in larger specimens (e.g. TMP 85.36.292; Figs 9 View Figure 9 , 10 View Figure 10 ), especially in TMP 91.36.265. This is in part because of the many furrows that run along the surface of the parietal that appear to be pathological. The frontals, which lack these canals, still show some external rugosity. In dorsal view, the frontonasal boss is identifiable through anterior advancement beyond the supraorbital lobes that flank it on either side. In TMP 91.36.265, the frontonasal boss is almost completely flat except that it is angled dorsally and therefore thickens posteriorly. The frontonasal boss and the supraorbital lobes have a minor amount of mediolateral convexity resulting in ‘incipient grooves’ that make distinction between the two structures possible. In TMP 99.55.122, the frontonasal boss is less flat with a greater dorsal angle and more mediolateral convexity such that the frontal grooves that separate the boss from the supraorbital lobes are discernible. In both specimens, the supraorbital lobes are largely flat and separate from the posteromedial doming of the frontal, which thickens posteriorly before it contacts the welldomed parietal. The interfrontal suture is visible both dorsally and ventrally, whereas the frontoparietal suture is visible ventral and laterally, but indiscernible dorsally.

The morphology of the parietal in TMP 99.55.122 is the same as that in larger specimens of F. brevis . It has the same sloping parietal with a ‘down-turned’ posteromedial extension bearing two partial nodes (although the nodes are mainly restricted to the squamosal in this specimen). The sutural surface for the postorbital and squamosal are the same, as is the slit-like temporal chamber. TMP 91.36.265 conforms to these features except that the sutural surface for the squamosal is extremely short dorsoventrally. It seems likely that this is because of the recent closure of the supratemporal fenestra in this specimen and that the suture would further increase in height through ontogeny with inflation of the dome. Evidence to support this may be found on the right side of the parietal where a portion of the supratemporal fenestra may still be open. However, owing to breakage in this area, it is difficult to be certain. Beyond the short squamosal suture, TMP 91.36.265 has the same ‘down-turned’ parietal bearing two small nodes and slit-like temporal openings that support its referral to F. brevis .

ADULT MATERIAL

The majority of F. brevis specimens appear to be subadult and exhibit the morphology that has typically been attributed to this taxon (e.g. TMP 85.36.292, Figs 9 View Figure 9 , 10 View Figure 10 ; Williamson & Carr, 2002; Sullivan, 2000, 2003). One specimen, TMP 87.50.29 ( Figs 11, 12 View Figure 12 ), is identified here as an adult F. brevis . Previously, Williamson & Carr (2002) suggested that TMP 87.50.29 might be an adult representative of F. brevis , but noted that the specimen was rather large, with a high frontonasal boss, and lacked nodes on the medial extension of the parietal. Sullivan (2003) assigned this specimen to the genus Prenocephale , correctly identified the presence of partial nodes on the posteromedial extension of the parietal, and noted several differences between this specimen and Sp. buchholtzae ( TMP 87.113.3). He did not support the referral of this specimen to F. brevis , but this was apparently solely on the basis of size, as no characters were described to distinguish TMP 87.50.29 from F. brevis . Ryan & Evans (2005) figured this specimen as an adult ‘ Stegoceras brevis ’, but gave no comment on its morphology.

TMP 87.50.29 consists of a nearly complete frontoparietal dome missing a section of the dorsal surface near the centre of the dome, the ventral processes, and portions of the lateral sutural surfaces. The specimen is much larger than other specimens referred to this species (frontoparietal length, 95.1 mm; compared with 73.1 mm for UALVP 8508, the next largest specimen). In general, the morphology of this specimen is consistent with F. brevis . It has the same dorsal outline and all the characters used to diagnose this species, including the down-turned parietal with slit-like temporal chamber and smooth dorsal surface texture punctured by many small foramina. Additionally, TMP 87.50.29 has several morphological features that support its identification as an adult F. brevis .

The foramina in TMP 87.50.29 are less numerous than in smaller specimens of F. brevis (e.g. TMP 99.55.122, TMP 85.36.292) and are further restricted to the periphery. The frontonasal boss is more inflated and convex than in smaller specimens, and the supraorbital lobes are further incorporated into the dome, but still maintain mediolateral convexity. Distinct grooves separate the supraorbital lobes from the frontonasal boss as in all other specimens of F. brevis . Posteriorly, the dome is further inflated, which results in a stronger slope of the parietal. The medial extension of the parietal maintains the same ‘down-turned’ morphology of all other F. brevis specimens. The nodes on the posteromedial extension of the parietal would have been almost fully on the squamosal, as only slight depressions indicating the bases of the nodes are visible on the lateral edge of the extension. In lateral view, the morphology of the sutural surfaces is largely consistent with other specimens of F. brevis (e.g. TMP 99.55.122, TMP 85.36.292), except for the increased inflation of the supraorbital lobes and frontonasal boss. Additionally, the temporal chamber maintains a slit-like appearance in lateral view.

PARTIALLY ARTICULATED SPECIMEN

Previously F. brevis has only been known from isolated frontoparietal domes. Here we identify?TMP P70.10.2 ( Figs 13 View Figure 13 , 14 View Figure 14 ) as the first F. brevis specimen with articulated peripheral elements, which provides additional support for our assignment of the isolated squamosal and postorbitals to this taxon. Dorsally the frontoparietal dome is nearly complete, missing only the anterior-most portion of the frontonasal boss, the posterior extension of the parietal, and the anterolateral edges of the supraorbital lobes. Much of the posteroventral portion of the frontoparietal is missing, as is a portion of the right lateral supraorbital lobe. The right squamosal, left and right postorbitals, and left and right posterior supraorbitals are in articulation with the frontoparietal dome. The right squamosal is partially complete, missing most of the posterior bar and ventral surface. Some remnants of nodes are visible, but are largely abraded. The right postorbital is largely complete, but the postorbital bar is partially missing its lateral and ventral surfaces, which are highly abraded. Only the dorsomedial-most portion of the left postorbital is preserved. The posterior portion of the right posterior supraorbital is largely complete with a finished dorsal surface of the orbit ventrally and a lateral bar with a ridge running posteroventrally to anterodorsally in lateral view. The left posterior supraorbital is nearly complete and preserves a complete dorsal surface of the orbit ventrally and the sutural surface for the anterior supraorbital anteriorly, as seen in anterior view. The lateral bar and part of the dorsal surface of the left postorbital are heavily abraded and worn.

Overall?TMP P70.10.2 is intermediate in size between TMP 85.36.292 and the adult specimen TMP 87.50.29 and this holds true for its ontogenetic development as well. Posteriorly, the dome is fully inflated, as seen in dorsal view, more so even than TMP 87.50.29, but the supraorbital lobes are mostly flat and ventral to the level of the dome as in smaller specimens (e.g. TMP 99.55.122, TMP 85.36.292). The dorsal surface of the dome largely lacks the many foramina typical of smaller, less mature specimens. Instead these are restricted to the periphery of the dome, as in the adult specimen TMP 87.50.29. In many cases the foramina are partially obscured by the highly abraded and worn nature of the periphery of the dome. In dorsal view, the dome has an overall round appearance similar to TMP 99.55.122, but this is confounded by the incomplete frontonasal boss, missing posteromedial extension of the parietal, and the articulated squamosal, postorbitals, and posterior supraorbitals. Thus, the dome shape was probably more similar in shape to the larger F. brevis specimens (TMP 85.36.292, TMP 87.50.29).

The squamosal, postorbitals, and posterior supraorbitals are all highly incorporated into the dome. The squamosal is similar in overall appearance to the isolated squamosal UALVP 49440, but is further incorporated into the dome and correspondingly has less of the characteristic foramina along its dorsal surface. Unfortunately, as noted above, the squamosal bar of?TMP P70.10.2 is not preserved and the ornamentation between the two squamosals cannot be compared. The morphology of the postorbital closely matches that of the isolated postorbital TMP 92.36.1125. Again, it is more incorporated into the dome, has fewer foramina, and is missing the lateral bar owing to abrasion. The posterior supraorbital is less incorporated into the dome than the squamosal or postorbitals, but more so than in St. validum (UALVP 2) or P. prenes (Z. Pal. MgD-I/104). The lateral bar is tall, similar to that seen in the isolated F. brevis postorbital (TMP 92.36.1125), but unlike St. validum (UALVP 2) or P. prenes (Z. Pal. MgD-I/104). However, unlike the postorbital, the posterior supraorbital has a ridge that starts on the dorsal edge of the bar (but the anterior half of the posterior supraorbital is missing) and drops ventrally posteriorly so that is it not continuous onto the postorbital. This condition may be unique to F. brevis as typically the lateral ridge or node row is continuous along the squamosal, postorbital, posterior supraorbital, and further anteriorly (e.g. St. validum , P. prenes ). Overall the morphology of the specimen is consistent with other specimens of F. brevis including the referred isolated squamosal and postorbital.

3D RECONSTRUCTION

Detailed 3D laser surface scans of the isolated squamosal (UALVP 49440), one of the postorbitals (TMP 92.36.1125), and two nearly complete frontoparietal domes (CMN 121 and TMP 85.36.292) were taken by ARIUS 3D, scaled to the same size, and articulated in three dimensions based on the morphology of their sutural surfaces. This produced a 3D reconstruction of the skull of F. brevis that provides the most complete picture of the taxon to date ( Fig. 15).

The squamosal has a primary row of six nodes on its posterior border with the medial-most node straddling the parietosquamosal suture. The posterior region of the squamosal is devoid of the minute nodes found in St. validum (e.g. CMN 138, TMP 85.4.1, UALVP 2), but does have a prominent lateroventral corner node, as in P. prenes (Z. Pal. MgD-I/104) and Sphaerotholus (TMP 87.113.3, NMMNH P-27403). The squamosal bar ventral to the node row is much taller than that found in any other pachycephalosaur and maintains the same height throughout its length. In posterior view, the squamosal is distinctly sloped ventrolaterally, as in Sphaerotholus (TMP 87.113.3, NMMNH P-27403) and P. prenes (Z. Pal. MgD-I/104), but maintains the same depth throughout its width. The postorbital is narrow and tall, as in Sp. buchholtzae (TMP 87.113.3), but more convex dorsally, as in P. prenes (Z. Pal. MgD-I/104). The postorbital bar is smooth and devoid of nodes, as in Sphaerotholus (TMP 87.113.3, NMMNH P-27403). The overall morphology is very similar to Prenocephale – Sphaerotholus grade taxa (e.g. smooth squamosal bar, corner node, down-turned parietal), but also contains several characters shared with Stegoceras (e.g. more than five primary nodes, groves in frontal, slit-like temporal chamber).

FRONTOPARIETAL VARIATION AND ONTOGENY

Sullivan (2006) claimed that there is no known growth series for F. brevis ; however, we have identified specimens with both juvenile and adult morphologies that differ from the previously described morphology for this species ( Sullivan, 2000, 2003, 2006), but which are consistent within an ontogenetic growth series ( Fig. 16 View Figure 16 ). In order to test the hypothesized growth series of F. brevis , growth of the frontoparietal was analysed allometrically and compared with the allometric growth in the other three pachycephalosaur taxa from the Belly River Group, Alberta. Additionally, variation in frontoparietal morphology was analysed quantitatively using multivariate morphometrics. This allowed us to quantitatively describe the ontogeny of F. brevis and to test the hypothesis that it, and the other Belly River Group taxa, represent distinct morphological species. We also note quantitative differences in the morphologies of these taxa, which support F. brevis and the other three Belly River group taxa as distinct.

Frontoparietal ontogeny of St. validum was analysed previously ( Schott & Evans, 2012) and it was found, contra Chapman et al. (1981), that the frontoparietal dome thickened dorsoventrally with positive allometry. Additional ontogenetic changes were identified, such as the frontoparietal widening posteriorly at a faster rate than anteriorly, which results in a triangular appearance of the dome in large specimens, and the negative allometric growth of the frontal with respect to the parietal ( Schott & Evans, 2012). Evans et al. (2013a) examined more general morphometric patterns across pachycephalosaurs with complete frontoparietal domes and found that positive allometry of frontoparietal dome thickness was common to all species studied and did not show significant variation. However, other aspects of frontoparietal growth did vary and could be used to distinguish species ( Evans et al., 2013a).

METHODS

Linear measurements were taken from the majority of pachycephalosaur specimens from the Belly River Group of Alberta. The specimens include both fossils and casts (when fossils were not available). Following the methodology of Schott & Evans (2012) a total of 21 linear measurements was taken between 27 homologous landmarks on the frontoparietal, based, in part, on those identified by Goodwin (1990) and Schott et al. (2011), as shown in Figure 17 View Figure 17 . Owing to the fragmentary nature of fossil remains not all measurements could be taken on all specimens. Most landmarks are positioned at the dorsal margin of the contact between sutural surfaces for the peripheral skull bones on the frontoparietal. These landmarks are as follows: n/n, n/prf, prf/aso, aso/pso, pso/po, po/ stf/sq. The widths, heights, and length of the sutural surfaces were all taken between the corresponding landmarks. The remaining landmarks are on the ventral surface of the dome. These are located at the anterior edge of the frontoparietal at the sutural contact of the nasals, along the midline at the anterior extent of the braincase, along the midline at the contact between the frontals and parietal in the endocranial fossa, along the frontoparietal suture at the lateral extent of the endrocranial fossa, and at the posterior edge of the dome along the midline. Frontoparietal thickness was measured using proportional callipers between landmarks located within the endocranial fossa at the contact of the frontals ss n ss aso ss pso

ss prf

ss po stf ss sq and parietal and on the dorsal surface of the frontoparietal at the contact of the frontals and parietal. All other measurements were taken using standard digital callipers. Measurements were transformed logarithmically so that they fit the linear allometric growth function, which has the additional benefit of normalizing the distribution of the data. A complete list of measurements and specimens used can be found in Table S1.

Regressions were calculated using reduced major axis regression (RMA, also known as standard major axis regression) implemented in the LMODEL2 software package ( Legendre, 1998; see also Legendre & Legendre, 1998) for the statistics program R (R Core Team, 2013). RMA has been considered an appropriate regression model for tests of allometry ( Warton et al., 2006; Smith, 2009). Statistical significance of the allometric pattern was determined based on the confidence intervals (i.e. if the confidence interval encompasses a slope of one, it is statistically isometric).

Owing to the varying completeness of specimens, we used four different standard (x) variables: frontoparietal length, frontoparietal width, parietal length, and frontal length. Frontoparietal length is perhaps the most intuitive index of size; however, very few specimens have complete frontoparietals. In order to include juvenile specimens and maximize sample size (and thus statistical power), we also used frontal and parietal lengths and frontoparietal width as standard variables for comparisons. In these cases, even isolated frontals and/or parietals could be incorporated into the bivariate analyses.

A detailed analysis of frontoparietal growth was performed for F. brevis using the majority of variables available. Dodson (1975a) found that intraspecific allometries in a growth series of extant Alligator mississipiensis showed very high correlation coefficients and thus we consider a high correlation coefficient as support for the hypothesis that the specimens of F. brevis represent an ontogenetic series of a single species. Following this, presence of the proposed juvenile (TMP 85.43.68) and adult (TMP 87.50.29) specimens of F. brevis within the confidence limits for the regression lines would support the hypothesis that they represent juveniles and adults of F. brevis , respectively.

A second analysis was performed comparing frontoparietal growth in the four potential Belly River group taxa, F. brevis , St. validum , H. sternbergi , and C. lambei . Dodson (1975b) found that the growth of two sympatric species of Sceloporus was very similar, but that differences were detectable in characters that showed strong positive or negative allometry. Thus, we might expect similar results in the analysis of the Belly River Group assemblage, with major differences amongst species evident in measurements that are either strongly negatively or positively allometric.

Finally, a principal component analysis was performed using the linear measurements. Because standard multivariate techniques cannot accommodate missing data, a subset of the available measurements and specimens were used that balanced sample size and the amount of morphological variation captured by the measurements. The standard (x) variable used for the bivariate analyses (frontoparietal width) was excluded from the PCA in order to test for size dependence of the variables through a linear correlation with principal components. These data were analysed using a PCA on both covariance and correlation matrices implemented in the statistical program PAST ( Hammer, Harper & Ryan, 2001). In order to statistically test the morphological distinctness of each taxa, a multivariate analysis of variance (MANOVA) was also performed in PAST.

All analyses were performed on log-transformed measurements, as log-transformed morphometric data fit a linear function (the allometric growth function) and so will not violate the assumption of linearity inherent in a PCA. A list of measurements and specimens used can be found in Table S2.

RESULTS

Results of the bivariate regressions for F. brevis are given in Tables S3 – S5 and graphs of select indices are presented in Figure 18 View Figure 18 . Regressions using parietal width as the standard variable had the highest sample size and thus the most statistical power. Comparisons using frontoparietal length and parietal length had lower sample sizes that may have hindered the ability to detect positive and negative allometry statistically. The correlation coefficients were generally high (most are>0.9, see Tables S3 – S5), which supports the hypothesis that the specimens of F. brevis are an ontogenetic series of a single taxon. Additionally, the proposed juvenile (TMP 85.43.68) and adult (TMP 87.50.29) specimens of F. brevis fell within the confidence intervals for the regression lines, and in most cases fell on the regression line (see the smallest and largest specimens in Fig. 18 View Figure 18 ), which supports the hypothesis that they represent juveniles and adults of F. brevis , respectively.

The height of the dome generally exhibited positive allometric growth (Table S3, Fig. 18A View Figure 18 ). The positive allometry was weakest when compared with frontoparietal width. The height of the frontonasal boss tended to show the strongest positive allometry, although the slope was not generally significantly greater than the other slopes. There was a general trend for the slopes of the heights to decrease posteriorly, but the differences in the individual slopes were generally not significant. Following this trend, the slope for thickness of the frontoparietal was generally the lowest.

The growth of the width of the frontoparietal did not show a single definitive pattern (Table S4, Fig. 18B View Figure 18 ). When compared with frontoparietal width some of the dome widths showed isometry [width (W):n/prf, W:prf/aso, W:pso/po], whereas the width across the frontoparietal at the contact between the anterior and posterior supraorbital suture showed negative allometry, and the width across the frontoparietal at the contact between postorbital and squamosal sutures showed positive allometry. When frontoparietal length was used as the standard variable the two anterior-most measurements showed isometry (W:n/prf and W:prf/aso), whereas the posterior measurements all showed positive allometry. If parietal length was used as the standard variable, only the two posterior-most measurements were positively allometric.

The lengths of the frontoparietal sutures all exhibited statistically isometric growth and in most cases the slopes were near one (Table S5, Fig. 18C, D View Figure 18 ). The length of the frontoparietal, and the frontals and parietal individually, showed negative allometry when compared with frontoparietal width. The frontal and parietal showed isometric growth with respect to each other and to frontoparietal length.

Results of the bivariate regressions comparing the four Belly River Group taxa are given in Tables S6 – S8 and graphs of select indices are presented in Figures 19 – 21 View Figure 19 View Figure 20 View Figure 21 . Comparisons amongst the Belly River Group taxa were almost exclusively made using the width of the frontoparietal as the standard variable because of the small sample sizes for both C. lambei and H. sternbergi , which also lack specimens with complete frontals and/or parietals.

Foraminacephale brevis , St. validum , and C. lambei showed relatively high correlation coefficients (r> 0.8; see Tables S6 – S8) for most variables, supporting the hypothesis that they are distinct species. The correlation coefficients for H. sternbergi are generally much lower (with many below 0.5, see Tables S6 – S8). This is largely because of the very small size range for this taxon (14 mm range in frontoparietal width compared with 67 mm for St. validum ), but also suggests the possibility that the H. sternbergi sample represents more than one species, of which some specimens may be adult St. validum . Unfortunately, owing to the small sample size, this possibility cannot be properly tested at this time.

The height of the frontonasal boss (Table S6, Fig. 19A View Figure 19 ) showed similar positive allometry in all four taxa. H. sternbergi had a greater slope (5.48) than the other three taxa, but not significantly so, and with a very low correlation coefficient (r = 0.32). The height of the supraorbital lobes (Table S6, Fig. 19B View Figure 19 ) showed similar positive allometry in both F. brevis and St. validum (1.84 and 1.9, respectively). H. sternbergi had a higher slope (3.6) than the other taxa, but again the difference was not significant and the correlation coefficient was very low (r = 0.03). C. lambei shows isometric growth of the supraorbital lobe, but the slope (1.18) was not significantly lower than in the other taxa. Both F. brevis and St. validum had positive allometry for thickness of the frontoparietal compared with width of the frontoparietal (Table S6, Fig. 19C View Figure 19 ), but the slope in St. validum (2.72) was significantly greater than in F. brevis (1.84). Both C. lambei and H. sternbergi showed isometry. The results were the same if frontal length was used as the standard variable instead of parietal width (Table S7, Fig. 19D View Figure 19 ).

The width of the frontonasal boss was isometric in all taxa (Table S8, Fig. 20A View Figure 20 ). Stegoceras validum and F. brevis showed negative allometry for the width across the supraorbital lobes (Table S8, Fig. 20B View Figure 20 ). Stegoceras validum also showed negative allometry of the width anterior to the frontoparietal suture (W:pso/po; Table S8, Fig. 20C View Figure 20 ). Both F. brevis and H. sternbergi showed positive allometry for the width posterior to the frontoparietal suture (W:po/stf/sq) with the slope in H. sternbergi (4.82) being significantly greater than that in F. brevis (1.16; Table S8, Fig. 20D View Figure 20 ).

The allometry of the lengths of the sutures was similar amongst all taxa, with all showing isometry (Table S9, Fig. 21A – C View Figure 21 ). The length of frontal was negatively allometric in F. brevis and St. validum , but isometric in C. lambei and H. sternbergi (Table S9, Fig. 21D View Figure 21 ).

Overall, many of the allometric trends were similar amongst the four taxa and this is to be expected when comparing closely related species ( Dodson, 1975b). However, differences in the allometric coefficients (slopes) are expected for strongly allometric variables ( Dodson, 1975b). The thickness of the frontoparietal, a trait that shows strong positive allometry, was significantly different between F. brevis and St. validum and between St. validum and C. lambei . Additionally, the slopes for the width posterior to the frontoparietal suture were significantly different between F. brevis and H. sternbergi . These differences further support the hypothesis that these samples represent distinct taxa within the Belly River Group assemblage. It is likely that increased sample sizes and ranges will allow additional differences in the allometries of these species to be detected.

The results of the PCA on the Belly River Group taxa are shown in Figures 22 View Figure 22 and 23. The first principal component (PC) explained 81.2% of the variation in the covariance matrix with PC 2 and 3 explaining 6.5 and 4.4% of the variation, respectively (Table S10). With the correlation matrix, PC 1 explained 71.4% of the variation in the covariance matrix with PC 2 and 3 explaining 7.8 and 7.0% of the variation, respectively (Table S10). All variables had positive loadings on the first component for the covariance and correlation matrices (Table S11), and this variable was strongly correlated with size (r = 0.91 and 0.96, P <0.000) for the covariance and correlation matrices, respectively; Fig. 23). However, the height of the anterior portion of the frontoparietal [height (H):n/n, H:pfr/aso, H:aso/pso] loaded most strongly on this component for the covariance matrix (Table S11), whereas in the correlation matrix the loadings were more similar except for those that described the shape of the endocranial fossa [length (L):ecf and W:ecf], which were weakest (Table S11). PC 2 and 3 both had more variable loadings and neither was significantly correlated with size (P> 0.05, Fig. 23). For the covariance matrix PC 2 was most strongly influenced by the height of the posterior supraorbital – postorbital sutural contact (H:pso/po), followed by the height of the nasal suture and the length of the anterior supraorbital (H:n/n, L: aso). PC 3 was most strongly influenced by the thickness of the frontoparietal [thickness (T):fp] followed by the height of the nasal suture and the prefrontal/ anterior supraorbital sutural contact (H:n/n, H:prf/ aso), and the length of the posterior supraorbital (L: pso). For the correlation matrix, PC 2 was most strongly influenced by the width of the endocranial fossa (W:ecf) followed by the length of the anterior and posterior supraorbitals (L:aso, L:pso). The width of the endocranial fossa also loaded most strongly on PC 3, but was followed by the height of the posterior supraorbital – postorbital sutural contact (H:pso/po).

The covariance matrix generally showed better separation of the taxa than the correlation matrix. Good separation of most taxa was obtained with the covariance plot of PC 1 and 2. However, there was no separation between St. validum and C. lambei . The plot of PC 2 and 3 also showed separation of the taxa with some overlap between F. brevis and Hanssuesia , and between St. validum and Colepiocephale .

The MANOVA found significant differences amongst the four taxa (Wilks’ k = <0.001, F = 25.4, P <0.001), which supports the hypothesis that the four taxa are indeed morphologically distinct. Unfortunately, the sample sizes were too small for pairwise comparisons between the species.

FRONTOPARIETAL HISTOLOGY

Cranial histology has previously been examined in pachycephalosaurs by Brown & Schlaikjer (1943), Goodwin, Buchholtz & Johnson (1998), Goodwin &