Etymology

The genus is named for Yama, a Tibetan tantric Buddhist deity, who is the Lord of Death and one of the eight Dharmapalas, or protectors, of Buddhist teaching. Yama has the head of a water buffalo and bears horns, a trait from which ceratopsians derive their name. The specific name refers to the Eastern Gobi provenance of this taxon.

Locality and Geological Setting

The specimen derives from a fluvial, pale red sandstone layer at Khugenetslavkant. These redbeds appear to overlie the Khar Khutul ( =  Khar Hötöl) beds and underlie an unnamed unit with a fauna of segnosaurs, advanced iguanodontians, and ornithomimosaurs reminiscent of the Iren Dabasu fauna (Gilmore, 1933; Currie and Eberth, 1993). The Khar Khutul beds may be correlated with the Khuren Dukh ( =  Hüren Dukh) beds that have been interpreted as Aptian–early Albian in age (Jerzykiewicz and Russell, 1991; Khand et al., 2000), but have also been dated as Barremian (~128 MYA) by Shuvalov (2000). The age of the Iren Dabasu fauna has been poorly constrained with ages ranging from the Cenomanian to the early Campanian (Jerzykiewicz and Russell, 1991; Currie and Eberth, 1993; Khand et al., 2000). If the correlations outlined above are correct, the Khugenetslavkant sandstone facies appears to be late Early Cetaceous in age. This is roughly similar or perhaps slightly younger than the Albian age ascribed to Archaeoceratops (Tang et al., 2001). Stratigraphic relationships among central Asian continental sediments are still poorly understood and much work remains to be done in this respect.

Differential Diagnosis

Small basal ceratopsian taxon with unkeeled rostral unlike all other Neoceratopsia (optimization-dependent unique trait); rostral patterned with anastomosing ridges and grooves (present in neoceratopsians more derived than Liaoceratops); subtemporal part of jugal lobate and deeper than suborbital section (plesiomorphic for Neoceratopsia, shared only with Liaoceratops); epijugal ossification present (plesiomorphic for neoceratopsians other than Liaoceratops); angular with two tubercles along ventral edge (unique trait; three tubercles present in Liaoceratops); surangular forming vertical wall to glenoid (plesiomorphic for neoceratopsians more derived than Liaoceratops); all cheek teeth with distinct, offset primary ridge (plesiomorphic for neoceratopsians more derived than Liaoceratops).

Holotype

The holotype (IGM 100/1315) consists of a partial skull missing the premaxillae, rostral, and predentary, as well as much of the frill and the left cheek and posterior part of the left mandible. The holotype skull appears to be from a mature individual, as witnessed by closure of the sutures among the occipital and basicranial braincase elements. The epijugal is not fused to the jugal, a feature that appears to be related to advanced maturity in Protoceratops andrewsi and ceratopsids. The skull is only dorsoventrally compacted to a small degree.

Referred Materials

A referred specimen (IGM 100/1303) collected in 2002 comprises the rostral, left maxilla, right jugal and quadrate, the left surangular, both articulars, a few isolated teeth, three partial vertebrae, the left ilium, and several hindlimb fragments. Additionally, a large number of fragmentary ceratopsian bones were surface collected from the Khugenetslavkant locality by the 2002 and 2003 AMNH-MAS field crews. An isolated predentary, IGM 100/1867, is described here and can safely be referred to Yamaceratops based on its similarity with the predentaries of other basal neoceratopsians.

Openings of the Skull

The bones that circumscribe the naris are missing from both the holotype and the referred specimen (figs. 1–Figure 23). The antorbital fossa is subtriangular in outline, but does not extend as far below the anterior margin of the orbit as in the holotype of Archaeoceratops (IVPP V11114). At its ventral end, the fossa forms a deep pocket in the body of the maxilla above and lateral to the tooth row (fig. 2). The dorsal half of the fossa is recessed into the lacrimal, which meets the maxilla in an oblique suture along the floor of the fossa (fig. 1). Neither specimen preserves the rim of the antorbital fenestra, which is highly reduced in most ornithischians.

Figure 1

Left lateral view of the holotype skull of Yamaceratops dorngobiensis (IGM 100/1315). Abbreviations are explained in appendix 3.

Figure 2

Rostral view of the holotype skull of Yamaceratops dorngobiensis (IGM 100/1315). Abbreviations are listed in appendix 3.

Figure 3

Dorsal view of the holotype skull of Yamaceratops dorngobiensis (IGM 100/1315). Abbreviations are listed in appendix 3.

As in other ceratopsians, the orbit is circumscribed by the jugal, lacrimal, prefrontal, the frontal, and postorbital (figs. 1–Figure 23). The jugal forms the entire ventral orbital rim, which is shallowly concave. Anteriorly, the jugal bears a short spur that overlaps the base of the lacrimal and forms a distinct angle at the anteroventral corner of the orbit (fig. 1). A similar spur occurs in Archaeoceratops (IVPP V11114), Protoceratops (AMNH 6429, 6466), Leptoceratops (CMN 8887, 8889), and some other ceratopsians. The lacrimal forms only a small part of the anterior rim of the orbit, which is mainly bounded by the prefrontal. The frontal contribution is relatively short and bridges the small gap between the prefrontal and the rostral process of the postorbital (fig. 3). The postorbital provides the largest single contribution to the orbit. The posterodorsal corner of the orbit is formed at the intersection of the short rostral process and the long descending process, which bounds the entire posterior rim.

Both the infratemporal and supratemporal fenestrae are preserved only on the right side of the holotype (figs. 1, 3). As in other basal neoceratopsians, the postorbital is excluded from the margin of the infratemporal fenestra by the long dorsal process of the jugal. The subtemporal portion of the jugal is relatively deep and the ventral border of the fenestra is level with the ventral-most point of the orbit. A small embayment in the jugal border of the fenestra, adjacent to the ventral end of the postorbital, is observed on the right side of the specimen (fig. 1). This embayment appears to have finished edges and is not observed in any other basal ceratopsian taxon. It may be a unique feature in this individual, perhaps due to pathology, or may represent a derived character of Yamaceratops.

The supratemporal fenestra is only partially preserved because most of the frill is missing in the holotype. The rostral end of the fenestra extends as a shallow depression on to the caudal edge of the frontal (fig. 3). Laterally, the postorbital overhangs the fenestra and does not form a shelf along the fenestral edge as in Leptoceratops (CMN 8887, 8889), Montanoceratops (Makovicky, 2001), and an undescribed leptoceratopsid specimen from Alberta (TMP 88.11.1; Ryan and Currie, 1998). The anterior edge of the fenestra forms a wide concave curve and is not constricted in dorsal view.

Rostral

The rostral of IGM 100/1303 is nearly complete, missing only part of the buccal margin and the left buccal process (fig. 4). The rostral is relatively thick compared to the other skull bones, and it has a rugose, highly vascularized outer texture traversed by anastomosing grooves (fig. 4A), a feature seen in most other neoceratopsians except for Liaoceratops. Although it is more acutely U-shaped in ventral view than the rostral of psittacosaurs (i.e., AMNH 6542), the rostral of Yamaceratops is not sharply keeled as in more advanced neoceratopsians. Dorsally, the rostral bears a slender process that would have wedged between the two halves of the premaxillary internarial bar. This process continues as a thin midline crest about halfway down the internal surface of the rostral (fig. 4B). A similar crest is present on the internal midline of the rostral of Brachyceratops (Gilmore, 1917; USNM 8072, USNM 14765) and is more prominent on the smaller juvenile rostral (USNM 8072) than on the larger one (USNM 14765). It is probably a widespread feature among neoceratopsians, but its presence is often concealed by the articulated nature of specimens.

Figure 4

Rostral bone of the referred specimen of Yamaceratops dorngobiensis (IGM 100/1303) in lateral (A) and internal (B) views. Abbreviations are listed in appendix 3.

The right buccal (lateral) process is preserved, but appears to be incomplete caudally (fig. 4). It is grooved dorsally where it fits onto the ventral edge of the premaxilla. The preserved buccal margin appears relatively blunt, which suggests that the cropping edge was formed entirely by a thick keratinous beak. A buccal process was proposed as a potential synapomorphy uniting leptoceratopsids with Ceratopsidae to the exclusion of other neoceratopsians by You and Dodson (2003). As demonstrated in Yamaceratops, this character has a broader distribution and is present in all neoceratopsian taxa that preserve the rostral, including Liaoceratops. Sereno (2000) and You and Dodson (2003) regarded this structure as absent in Protoceratops, but juvenile specimens of Protoceratops sp. from Ukhaa Tolgod reveal that it is present as a differentiated process in ontogenetically young individuals, and its apparent absence is related to expansion of the lateral sheet of bone between the buccal and dorsal processes through ontogeny.

Maxilla

The maxilla is a complex bone, which forms the majority of the preorbital portion of the face. Rostrally, the body of the maxilla meets the premaxilla along a straight, but oblique, suture near the buccal margin (figs. 1, 5A). Neither the holotype nor the referred specimen preserves the dorsal edge of the maxilla where it would meet the nasal and lacrimal. The maxilla bears a long rostral process that extends forward to form the palate between the premaxillae (figs. 2, 5), which were apparently excluded from the palate as in other marginocephalians (Sereno, 1986). The rostral process is triangular in cross section with a vertical midline sutural surface where it meets its counterpart anterior to the choana. The sutural surface is marked by a series of horizontal, parallel striations in the disarticulated maxilla of IGM 100/1303 (fig. 5B); these would have interlocked with striations from the opposite maxilla. The roof of the palatal shelf is dorsally concave (fig. 3). Posterior to the maxillary palatal roof, the palatal shelves diverge and form an elongate spindle-shaped choanal vacuity between them (fig. 3) as in other basal neoceratopsians (Osmólska, 1986). The maxilla meets the palatine along an oblique suture on the palate as in most ornithischians.

Figure 5

Left maxilla of the referred specimen of Yamaceratops dorngobiensis in lateral (A) and medial (B) views. Abbreviations are listed in appendix 3.

The body of the maxilla widens caudally beneath the antorbital fossa. Above the tooth row, the face of the maxilla curves laterally to form a deep “cheek” recess lateral to the tooth row, but there is no sharp crest as in some basal ornithopods (fig. 5A). The floor of the antorbital fossa is excavated by a deep pocket, which appears to end just dorsal and lateral to the tooth roots in the holotype. The tooth row is partially preserved in both maxillae of the holotype, as well as in the left maxilla of the referred specimen. There are about 11 or 12 tooth positions in both the holotype and the referred specimen, representing a number similar to Liaoceratops (Xu et al., 2002) and Archaeoceratops (You and Dodson, 2003).

Caudally, the maxilla has a complex articulation with the jugal, the details of which are best seen in the disarticulated referred specimen (figs. 5, 6). Laterally, the concave caudal end of the maxilla slots onto the anterior end of the jugal along a curved and oblique suture that extends from the caudal wall of the antorbital fossa to below the orbit (fig. 5A). Adjacent to this facial process, the medial edge of the maxilla bears a long fingerlike process, which slots into a groove on the medial face of the jugal (fig. 5B). A third apparent process just rostral to the medial process (fig. 5A) is actually a separate piece of bone adhering to the maxilla, perhaps derived from the lacrimal.

Figure 6

Right partial jugal of the referred specimen of Yamaceratops dorngobiensis (IGM 100/1303) in dorsal view, showing details of the maxillary articulation on rostral end. Abbreviations are listed in appendix 3.

The anatomy of the jugal–maxillary contact is difficult to evaluate in a majority of basal ceratopsians, either because the elements remain in tight articulation (Leptoceratops 8887, 8889; most specimens of Protoceratops and Bagaceratops; Archaeoceratops IVPP V11114; Liaoceratops IVPP V12738) or are abraded or missing as in Montanoceratops (AMNH 5464). Disarticulated specimens reveal a broad overlapping suture in Montanoceratops (MOR 542), Protoceratops (IGM 100/1130), and Prenoceratops (Chinnery, 2004). In Udanoceratops (PIN 3907/11) the maxilla bears a robust jugal process that is triangular in cross section and which fits into a wide slot on the medial face of the jugal, thus resembling the condition in Yamaceratops.

Jugal

Only the right jugal of the holotype is preserved in its entirety (fig. 1). It appears to be less symmetrical in lateral view than the jugal of other basal neoceratopsians. Rostrally, the jugal tapers to a point that contacts the rim of the antorbital fossa, whereas the caudal end is deeply expanded and lobate in shape. This rostrocaudal asymmetry contrasts with the jugals of taxa such as Archaeoceratops (IVPP V11114), Leptoceratops (CMN 8887, 8889), Protoceratops (AMNH 6466), and Udanoceratops (PIN 3907/11), which exhibit a greater symmetry between the depth of the suborbital and subtemporal parts of the jugal. In these taxa, the rostral tip of the jugal is expanded and participates in the rim of the antorbital fossa, while the caudal end tapers to a point below the infratemporal fenestra. The jugal of Yamaceratops resembles the jugal of Liaoceratops (Xu et al., 2002) more closely than those of other basal neoceratopsians. In Liaoceratops the suborbital part of the jugal tapers strongly, whereas the subtemporal part is markedly deep. Unlike Yamaceratops and other more derived neoceratopsians, however, the suborbital part of the jugal is truncated and the quadratojugal is visible in lateral view in Liaoceratops. An equal or greater depth of the suborbital part of the jugal relative to the temporal ramus has been proposed as a synapomorphy of Ceratopsia by various authors (Sereno, 2000; You and Dodson, 2003), but is not present in Liaoceratops and Yamaceratops.

The articulation for the maxilla is visible in the partial right jugal of IGM 100/1303 (fig. 6). A narrow groove that rises anteriorly marks the lateral articulation with the maxilla. This groove is backed by a smooth wall of bone that forms a narrow ridge separating the lateral and medial parts of the maxillary articulation in ventral view. The medial part of the articulation is a broad sulcus that received the posteromedial fingerlike process of the maxilla. Just dorsal to the medial maxillary articulation, the medial face of the jugal exhibits a low pyramidal mound that articulates with the tip of the lateral process of the ectopterygoid (fig. 6).

A narrow but deep notch invades the infratemporal border of the jugal at the base of the dorsal process and reaches almost halfway to the orbital border on the holotype skull (fig. 1). This embayment is surrounded by a smooth-walled depression in the bone, and no similar feature has been reported from other ceratopsian specimens. In several respects, this structure is similar to bone lesions interpreted as pathologies in some dinosaur specimens, including a mature Psittacosaurus mongoliensis (IGM 100/1132). However, without at least one more complete jugal, it remains uncertain whether this feature is a phylogenetic character or pathologic in origin.

The dorsal process of the jugal is elongate and contacts the squamosal along the dorsal rim of the infratemporal fenestra and excludes the postorbital from participating in the border of the fenestra (fig. 1). It appears to reach almost as far caudally as the postorbital. In Protoceratops (AMNH 6418, 6429, 6466), Archaeoceratops (IVPP V11114), and Leptoceratops (CMN 8887), the dorsal process of the jugal terminates level with or just rostral to the postorbital.

Caudal to the dorsal process, the subtemporal wing of the jugal flares and has a roughly rhomboid shape. The posterior edge is expanded laterally to form an oblique ridge, in place of the more pointed hornlike structure seen in coronosaurs such as Protoceratops and ceratopsids. This ridge is more laterally directed and not as prominent as the epijugal ridge of Udanoceratops and other leptoceratopsids, and is most similar to Archaeoceratops in its orientation and relative size. The convex caudal face of the ridge is rugose for reception of the epijugal (fig. 3).

The external surface of the jugal is lightly textured and marked by a number of small neurovascular foramina. One of these, located near the base of the dorsal process, appears to pass directly through the jugal, unlike in other ceratopsians with the exception of an unpublished specimen of Liaoceratops (P. J. Makovicky, personal obs.).

Epijugal

A well-preserved epijugal (fig. 7) was discovered with the holotype skull and presumably belongs with the complete right jugal. It is an elongate, crescent-shaped bone as in leptoceratopsids (AMNH 5464; CMN 8887, 8889), and would have attached to a long stretch of the caudodorsal edge of the jugal ridge. Its lateral surface is gently convex and is marked by vascular grooves (fig. 7B) that are typical of skull parts that bore a keratinous cover. The medial face is smooth and is concave (fig. 7A). An epijugal is absent in Liaoceratops and more basal ceratopsian taxa, but appears to be present in all other neoceratopsians.

Figure 7

Right epijugal of the holotype specimen of Yamaceratops dorngobiensis (IGM 100/1315), in medial (A) and lateral (B) views.

Lacrimal

Only the posterior part of each lacrimal, bearing the antorbital fossa rim and the orbital margin, is preserved on the holotype (figs. 1, 2). The antorbital fossa is deeply excavated on to the lateral surface of the lacrimal, and it presumably formed the caudal border of the antorbital fenestra as in Protoceratops andrewsi (Brown and Schlaikjer, 1940b). The lacrimal contacts the maxilla ventrally and the jugal caudally and abuts a transverse ridge of the palatine along the cruciform roof of the palate (fig. 2) as in protoceratopsids (Osmólska, 1986) and ceratopsids (Hatcher et al., 1907).

Prefrontal

The prefrontal is L-shaped with a descending ramus that bounds the orbit rostrally, and a posterior process that forms much of the dorsal rim of the orbit (figs. 1, 3). The two rami are perpendicular to each other, and the posterior ramus is the longer of the two. The prefrontal is flat dorsally, and has a gently concave lateral border, but is not sinuous as in the juvenile specimens that were used to diagnose Protoceratops kozlowskii ( =  Bagaceratops) (Maryanska and Osmólska, 1975). At the juncture of the ventral and posterior rami, the medial face of the prefrontal is deeply concave.

Frontal

The frontals of the holotype specimen are fused along the midline, although the sutural line is still faintly perceptible (fig. 3). At the rostral midline, the frontals form a sharp wedge between the posterior ends of the nasals. Each nasal would have overlapped a striated, sutural surface on the frontal adjacent to the prefrontal. The skull table is slightly concave between the sutures for the postorbitals. The supratemporal fenestra has a gentle emargination on to the posterior edge of the frontal (fig. 3), but it is not incised as markedly as in Protoceratops andrewsi (AMNH 6429), Leptoceratops (CMN 8889), or Liaoceratops (IVPP V12738). The suture with the postorbital is broad and sinuous in dorsal view with a lappet of the postorbital overlapping the frontal rostrally and a lappet of the frontal overlying the postorbital caudally. This sinuous morphology is seen in other basal ceratopsians, including Psittacosaurus mongoliensis (AMNH 6254), Liaoceratops (IVPP V12738), and Leptoceratops (CMN 8889), but appears to be a plesiomorphic character state and is also seen in Hypsilophodon (BMNH R2477) and Lesothosaurus (BMNH RU B23).

Postorbital

As in other neoceratopsians, the postorbital is a triangular plate of bone that extends from the posterior edge of the orbital margin and onto the temporal bar. Its external surface is slightly rugose, as in other neoceratopsians, suggesting tight adherence between the bone and the overlying skin. Along the orbital rim, the postorbital overlaps the jugal along an oblique suture that extends dorsomedially from the lower part of the orbit to its midheight, above which the postorbital exclusively forms the orbital rim. The posterodorsal corner of the orbit is transversely concave, and bounded laterally by a small convex lip. Dorsally, the upper edge of the postorbital overhangs the supratemporal fenestra as in most neoceratopsians with the exception of the leptoceratopsids Leptoceratops, Montanoceratops (Makovicky, 2001), and Prenoceratops.

The ventral edge of the postorbital is slightly concave overall as in Liaoceratops (Xu et al., 2002) and is divided into two gently concave sections separated by a small, shallow point located just dorsal to the embayment along the front edge of the infratemporal fenestra. Caudally, the postorbital inserts into the bifid rostral end of the squamosal as in Liaoceratops, Archaeoceratops, and leptoceratopsids.

Squamosal

Only the incomplete right squamosal is preserved on the holotype specimen (fig. 1). It is missing the caudalmost edge that forms the lateral corner of the frill (fig. 3). The squamosal bears two ventral processes that wrap around the head of the quadrate, a long, bifid rostral process that articulates with the postorbital and a broad, medial lappet that overlaps the parietal within the supratemporal fenestra. The rostral process wraps around the caudal end of the postorbital, such that a longer lateral and a shorter medial exposure are visible between the prongs of the squamosal (fig. 1). The ventral prong contacts the jugal along the rim of the infratemporal fenestra, but the location of the suture is difficult to determine given the large number of cracks along the edge of the temporal bar. We interpret the contact as lying just rostral to the caudalmost point of the postorbital. The rostral process of the squamosal is deeply bifid in the leptoceratopsids Montanoceratops (AMNH 5464) and Leptoceratops (CMN 8887). The sutural boundaries among the bones forming the temporal bar in Archaeoceratops are obscure, but the reconstruction presented by You and Dodson (2003) in which the squamosal has no rostral process is unlike the squamosal of any ornithischian and likely incorrect. Protoceratops (AMNH 6466) has a squamosal with a long rostral process that is slightly forked for reception of the postorbital. Ceratopsids have a wide zigzag suture between the postorbital and squamosal.

The medial process of the squamosal is lappet-like, but relatively less expanded than in Liaoceratops and psittacosaurs (fig. 3). The more rostral of the two processes bracing the head of the quadrate arises from the ventral edge of the medial process, but because of breakage, its length cannot be determined. The main body of the squamosal at the juncture of the processes forms a gently convex lateral surface (fig. 1) at the caudal end of the temporal bar, which suggests that the dorsal edge of the squamosal curved gently to join the parietal frill as in Liaoceratops (CMN 8887, 8889). This observation is bolstered by the shape of the caudolateral corner of the supratemporal fenestra in dorsal view (fig. 3). The squamosal makes a smooth, open transition to the frill margin, unlike the acute, slitlike frill corner in coronosaurs (Sereno, 2000; Makovicky, 2001). The broken laterodorsal edge of the fenestra is relatively robust, suggesting that the frill may have had a thickened edge as in Liaoceratops. The ventral edge of the squamosal frill margin is caudally concave in lateral view and gently convex transversely.

Parietal

What little is present of the parietal in the holotype skull is poorly preserved (fig. 3). The parietal meets the frontal along a zigzag suture that is oriented transversely and lies just caudal to the rostral emargination of the supratemporal fenestra. It forms the steep-sided medial walls of the supratemporal fenestrae, which are separated by a sagittal crest. The height of the sagittal crest is indeterminate because of breakage. Caudally, the frill portion of the parietal meets the dorsal edge of the paroccipital process along a straight, horizontal suture. The length and shape of the frill cannot be determined, as the entire margin is lost to erosion. Poor preservation also precludes observations on the nature of the contact between the parietal and the elements that constitute the sidewall of the braincase. On the occipital surface of the skull, the dorsal end of the supraoccipital fits into a slot along the ventral edge of the parietal as in other basal neoceratopsians (fig. 8).

Figure 8

Occipital view of the holotype skull of Yamaceratops dorngobiensis (IGM 100/1315). Abbreviations are listed in appendix 3.

Quadratojugal

As in neoceratopsians more derived than Liaoceratops, the quadratojugal is transversely wide and more exposed in occipital view than it is laterally (figs. 1, 8). The quadratojugal is crescent shaped in lateral aspect with a convex caudal surface that follows the curve of the jugal crest and a concave rostral surface that walls off the infratemporal fenestra caudally. In this respect, the quadratojugal appears similar to that of Montanoceratops (AMNH 5464), which also appears curved to follow the outline of the jugal horn. Dorsal to the overlap by the jugal, the quadratojugal apparently bears a short dorsal spine that adheres to the lateral edge of the quadrate. This process is mostly missing in the holotype, but its length can be deduced from the corresponding contact on the quadrate. A similar overlap of the quadrate by a pointed dorsal process of the quadratojugal is observed in Liaoceratops (IVPP V12738).

Quadrate

The quadrate has a nearly straight, rostrocaudally compressed shaft as in other neoceratopsians (figs. 1, 9). The quadrate head is tall and narrow and flanked caudally by a straight ridge that marks the contact with the caudoventral process of the squamosal and the tip of the paroccipital process. A broad, triangular pterygoid wing extends from the medial edge of the shaft for most of its length and terminates at the quadrate head (fig. 9). The distal condyles are asymmetric and separated by a shallow groove into a larger lateral and smaller medial articulation, unlike the flat condylar articulation of psittacosaurids. A sinuous ridge runs along the lateral face of the quadrate shaft marking the caudal edge of the sutural surface for the quadratojugal. It extends from the lateral condyle up the caudal edge of the shaft, curves rostrally at the midheight of the shaft, and terminates along its rostral edge about one-third of the way down from the quadrate head. A similar extensive ridge is present along most of the lateral surface of the quadrate shaft in Liaoceratops, but the extent of this ridge becomes relatively shorter in more derived taxa, such as Protoceratops, in which it only reaches about halfway up the quadrate shaft. As in the basal ceratopsian Chaoyangsaurus, the lateral edge of the shaft bears a short rostral process that underlaps the quadratojugal (fig. 9). Its length is indeterminate due to breakage, but it is unlikely that it reached the jugal. This plesiomorphic feature has not been reported in more derived ceratopsians, and is apparently also absent in Psittacosaurus mongoliensis (IGM 100/1032).

Figure 9

Right quadrate of the referred specimen of Yamaceratops dorngobiensis (IGM 100/1303) in oblique rostrolateral view. Abbreviations are listed in appendix 3.

Braincase

Erosion has obscured many braincase features, including many of the sutural contacts. The laterosphenoids form the rostrolateral walls of the braincase dorsal to the hypophysial fossa. They meet at the midline to separate the optic nerve exits from the olfactory nerve tract. Each laterosphenoid bears a prominent capitate process that reaches laterally to underlap the postorbital–frontal contact. A small foramen for exit of the trochlear nerve (CN IV) pierces the wall of the laterosphenoid dorsally, just rostral to the base of the capitate process. The sutures between the laterosphenoid and prootic or the prootic and opisthotic cannot be made out. A long external auditory meatus extends laterally from the middle ear as a groove along the ventral edge of the paroccipital process as in other neoceratopsians (fig. 10). A triangular slip of the supraoccipital is exposed inside the supratemporal fenestra along the base of the paroccipital process and wedges between the exoccipital and the parietal (fig. 3).

Figure 10

Ventral view of the holotype skull of Yamaceratops dorngobiensis (IGM 100/1315). Abbreviations are listed in appendix 3.

The supraoccipital is formed as a low, wide triangle and participates in the margin of the foramen magnum (fig. 10). A low, vertical midline ridge extends from the edge of the foramen magnum to the contact with the parietal. A midline ridge is variously developed among some basal neoceratopsian taxa including Leptoceratops, Protoceratops, and Montanoceratops, where it is split by a drop-shaped depression ventrally (Makovicky, 2001).

The condylar region of the holotype specimen is poorly preserved, but the exoccipitals appear to meet along the floor of the medullar region (fig. 8). This suggests that they formed the dorsal part of the occipital condyle, and the basioccipital did not join in flooring the medulla oblongata. The basioccipital flares laterally below the condyle, to form the large basioccipital tubera characteristic of ceratopsians (Makovicky, 2001). A shallow but wide notch separates the tubera at the ventral midline and forms the caudal border of a small pit between the basioccipital and basisphenoid (fig. 10). Such a pit is also observed in some but not all specimens of Protoceratops and Bagaceratops and was identified as a median Eustachian system as in crocodilians (Maryanska and Osmólska, 1975). It does not connect to the middle ear region, however, and its variable intraspecific occurrence within several neoceratopsian species argues against such a homology (Makovicky, 2001).

The basipterygoid processes are relatively robust and short as in Liaoceratops and unlike the more elongate and slender processes of Protoceratops or the recurved ones that are diagnostic of leptoceratopsids (Chinnery and Weishampel, 1998; Makovicky, 2001). They project lateroventrally below the hypophysial fossa as in other neoceratopsians, but unlike psittacosaurids and other outgroups, in which they are directed rostrolaterally. A triangular, bladelike cultriform process projects rostrally from between the basipterygoid processes.

Palatine

The palatines are inclined to form the posterior part of the highly vaulted palate (figs. 2, 10). Each palatine is a large subtriangular plate that overlaps the medial surface of the maxilla along an oblique suture extending from the caudal end of the choana to the ectopterygoid as in Protoceratops (PIN 614/7; Osmólska, 1986). Medially, the palatines meet to form the base of the interchoanal septum, which is broken in the holotype (fig. 2). The missing vomers presumably slotted into the narrow roof of the palate in this region. Caudal to the choana, the dorsal surface of the palatine bears a transverse lamina that abuts the lacrimal laterally. The transverse laminae of both palatines meet the sharp apex of the palatal roof in a cruciform juncture that is a derived character of many neoceratopsians (Osmólska, 1986). Along its caudomedial edge, each palatine forms the rostral rim of the pterygo-palatine foramen (Brown and Schlaikjer, 1940b), which is a reduced homolog of the sauropsid suborbital fenestra. The pterygo-palatine fenestra (figs. 10, 11) is proportionately quite large in Yamaceratops as in Liaoceratops, Psittacosaurus, and Chaoyangsaurus, unlike the relatively diminutive foramen observed in coronosaurs.

Figure 11

Oblique close-up of the right caudal region of the palate of the holotype skull of Yamaceratops dorngobiensis (IGM 100/1315). Abbreviations are listed in appendix 3.

Pterygoid

The pterygoid is a complex tetraradiate bone that contacts the palatine, ectopterygoid, basipterygoid, and quadrate, as well as the other pterygoid along the midline (fig. 10). The four rami converge on the main moiety of the pterygoid, which forms the caudal end of the palate just ventral to the braincase. A narrow gap separates the palatal bodies of the pterygoids, but it is unclear how much of this may be due to postmortem distortion of the skull. Such a gap is also observed in Prenoceratops (B. Chinnery-Allgeier, personal commun.). Each rostral process is triangular, inclined rostrodorsally, and meets its counterpart at the midline. Together, the triangular rostral processes slot into the vaulted palatal roof formed by the palatines, which partly overlap the rostral pterygoid processes laterally. A small gap between the tips of the pterygoids and the apex of the vault formed by the palatines presumably held the caudal ends of the vomers.

Just caudal to the pterygoid rim of the pterygopalatine fenestra, a long, slender mandibular process projects lateroventrally from the lateral edge of the pterygoid. It extends along the caudal rim of the palate and projects beyond the caudal end of the tooth row. The mandibular process underlaps the ventral process of the ectopterygoid along its full length, and together they form the palatal mandibular process that channels M. pterygoideus towards its insertion on the mandible (Galton, 1974).

The strongly flared quadrate wing of the pterygoid rises abruptly from the caudal part of the palatal body. At its base, the quadrate wing forms the narrow pocket for articulation with the basipterygoid processes. It flares widely at its distal end to overlap the pterygoid wing of the quadrate. An elongate, slitlike opening extends between the dorsal edge of the quadrate wing of the pterygoid and the base of the paroccipital process. A short caudal process projects along the midline, caudal to the facet for reception of the basipterygoid process as in other basal ceratopsians (fig. 10). It appears to be pointed rather than lobate as in Liaoceratops (IVPP V12738).

Ectopterygoid

The ectopterygoid is widely exposed on the palate in contrast to the condition in more derived neoceratopsians (fig. 11). It is triradiate with a wide but short lobe that extends on to the pterygoid, a strongly curved lobe that abuts the jugal, and a long ventrally curved process that is sutured to the pterygoid to form the ventral process that channeled the M. pterygoideus dorsalis (figs. 3, 11). In Yamaceratops, as in Chaoyangsaurus (IGCAGS V371) and Liaoceratops (IVPP V12738), a ventral trough excavates the base of the ectopterygoid for reception of the caudal end of the maxilla. The ectopterygoid thus separates the maxilla and pterygoid as in Liaoceratops and outgroup taxa such as Hypsilophodon (Galton, 1974; BMNH R2477). In more derived taxa, especially in ceratopsids, the ectopterygoid is reduced and the pterygoid contacts the caudal edge of the maxilla (Brown and Schlaikjer, 1940b).

In its robustness and participation in formation of the ventral process, the ectopterygoid is similar to those of Chaoyangsaurus (IGCAGS V371), Liaoceratops (IVPP V12738), and Leptoceratops (Sternberg, 1951) and juvenile Protoceratops (Brown and Schlaikjer, 1940b). In adult Protoceratops, the ectopterygoid has a very reduced exposure on the roof of the palate (Brown and Schlaikjer, 1940b; AMNH 6466), and this element is entirely excluded from palatal view in ceratopsids (Hatcher et al., 1907; Brown and Schlaikjer, 1940b). In Yamaceratops (fig. 11), as in Liaoceratops, Leptoceratops (Sternberg, 1951), and Protoceratops (Brown and Schlaikjer, 1940b), the ectopterygoid participates in the border of the pterygopalatine fenestra, as is primitive among dinosaurs. The pterygopalatine is much reduced in adult Protoceratops and ceratopsids relative to more basal forms, and an ectopterygoid contribution to the border of this opening is entirely lost in ceratopsids.

Predentary

Neither the holotype nor the referred partial specimen preserves the predentary, but an isolated partial basal neoceratopsian predentary (IGM 100/1867) was collected at the Kughenetslavkant locality and is referred to Yamaceratops (fig. 12). The element is missing most of its caudal end, including the labial and ventral processes. In dorsal view, the rostral end forms an acute point. The predentary is relatively solid and the buccal surface is only slightly concave as in the basal taxa Chaoyangsaurus, Liaoceratops, and Archaeoceratops in contrast to the scoop-shaped predentary of Leptoceratops, Protoceratops, and ceratopsids. The tip is slightly recurved in profile (fig. 12) and would probably have curved dorsally in life. As in most other basal neoceratopsians, the edges of the predentary are beveled and not sharp or grooved as in ceratopsids.

Figure 12

Isolated predentary (IGM 100/1867) referred to Yamaceratops dorngobiensis in lateral (A) and dorsal (B) views.

Dentary

The dentary has a gently curved ventral edge, but does not bear an everted flange as in Liaoceratops and some psittacosaurid species (fig. 1). Each dentary bears a prominent lateral ridge that extends from the base of the coronoid process to the symphysis, and the tooth row is deeply inset from the buccal margin. The mandibular symphysis is broad and faces anteriorly (fig. 2). On the better preserved left dentary of the holotype, a small groove is visible at the rostrodorsal tip of the dentary for reception of the buccal process of the predentary, but unlike in leptoceratopsids, there is no large pit in this region. As in other basal ceratopsians, there is little separation between the articular surface for the predentary and the first tooth position, in contrast to ceratopsids in which a distinct diastema occurs at the tips of the dentary. The coronoid process is well developed and separated from the tooth row by a wide sulcus unlike Chaoyangsaurus, Psittacosaurus, and Liaoceratops. The tooth row terminates level with and medial to the base of the coronoid process (figs. 1, 2) as in higher neoceratopsians.

Splenial

Both splenials are preserved in the holotype (fig. 10). As in other ceratopsians, the moiety of the splenial is a deeply expanded sheet of bone that overlaps the lingual face of the dentary (figs. 10, 11). The ventral margin of the splenial parallels the ventral edge of the dentary, but the dorsal margin slopes toward the symphysis so that the main body of the bone attenuates rostrally. The rostral end broadens transversely to form a symphyseal process that projects forward to meet its counterpart at the symphyseal midline.

Coronoid

Both coronoids are preserved in the holotype specimen. As in other neoceratopsians, the coronoid consists of an expanded lobe that forms the medial face of the coronoid process and a slender ventral process that curves around the rostrodorsal corner of the adductor fossa. The ventral process is relatively more robust than in Protoceratops. A separate intercoronoid is not discernible, but is known to occur in some other neoceratopsians, including Protoceratops, some ceratopsids (Brown and Schlaikjer, 1940b), and the basal taxon Liaoceratops (P.J. Makovicky, personal obs.).

Surangular

Both surangulars are preserved on the holotype skull, although the left one is missing the glenoid. A disarticulated, but well-preserved, left surangular was found with the referred specimen (fig. 13). The surangular is subtriangular in lateral view (fig. 13A) with the coronoid process forming the dorsal apex. Rostral to the coronoid process, the surangular bears a long, pointed rostral process, which slots into the caudal end of the dentary coronoid process. The lateral surface of this process is striated for sutural contact with the dentary, and grooves extend along the medial and lateral faces of the process adjacent to its dorsal edge. The rostral edge of the coronoid prominence is sharply keeled where it slots between the medial edge of the dentary and the expanded lobe of the coronoid. The caudal edge of the prominence is thickened and rounded and is caudally confluent with a convex wall of bone that delimits the lateral edge of the glenoid (fig. 13D). This lateral glenoid wall is observed in all neoceratopsians with the exception of Liaoceratops.

Figure 13

Left surangular of the referred specimen of Yamaceratops dorngobiensis (IGM 100/1303) in lateral (A), medial (B), ventral (C), and dorsal (D) views. Abbreviations are listed in appendix 3.

A robust lateral ridge extends along the labial face of the surangular (fig. 13A) as in Liaoceratops, Archaeoceratops, protoceratopsids, and Montanoceratops (Chinnery and Weishampel, 1998). Ventral to the caudal end of the edge and the glenoid, the lateral face of the surangular is depressed to receive the dorsal edge of the angular (fig. 13C). The surangular forms most of the lateral cotyle of the glenoid, which occupies the space between the lateral wall and articulation with the articular. The surangular bears a thin sheetlike medial process that wraps around the rostral end of the articular (fig. 13B). A thin crest that lies medial to the articulation for the angular forms a long, almost flat surface for articulation with the body of the articular (fig. 13A,C). This crest adheres to the full length of the articular and participates in the retroarticular process. The medial face of the coronoid prominence is concave, and a small vascular foramen is present anterior to the medial wing of the surangular, as in other neoceratopsians.

Articular

The articular is an oblong bone, triangular in cross section with an elliptical articular surface forming the glenoid dorsally and robust ridge forming the ventral edge of the bone (fig. 14). Rostral to the glenoid, the articular bears a short, blunt process that fits against the concave face of the surangular medial process (fig. 14B,C). The elevated glenoid surface is elliptical and gently concave dorsally. Its lateral edge is raised and forms the low but wide ridge that divides the glenoid into medial and lateral cotyla. The articular surface of the glenoid is rugose, and a sinuous line separates it from a small, cup-shaped fossa that lies at the caudal end of the articular dorsal to the retroarticular process (fig. 14B). The fossa, and a small mediodorsal tuber that lies adjacent to it, probably mark the insertion of the mandibular abductors as in crocodilians (Schumacher, 1973). A small foramen pierces the base of the tuber adjacent to its rugose apex, as in Xuanhuaceratops (Zhao et al., in press). It may connect to a small foramen and groove on the medial face of the bone (fig. 14A), and could be the trace of the Meckelian chord or, less likely, a foramen aerosum. The tuber is continuous ventrally with the blunt retroarticular process of the articular (fig. 14A).

Figure 14

Left articular of the referred specimen of Yamaceratops dorngobiensis (IGM 100/1303) in medial (A) and dorsal (B) views and in articulation with left surangular (C). Abbreviations are listed in appendix 3.

Angular

Only the right angular of the holotype is nearly complete (fig. 1), as the left one is missing the caudal end. The angular forms the caudoventral edge of the mandible and overlaps the dentary rostrally and the surangular dorsally. Rostrally, a process of the angular overlaps the caudal end of the dentary near its ventral edge on the lateral face of the jaw (fig. 1). A lappet of the angular also overlaps the ventral edge of the dentary in Liaoceratops (IVPP V12633). This lappet is more dorsally placed in Archaeoceratops (IVPP V11114), and a narrow, more fingerlike process of the angular slots into the caudal edge of the dentary in Leptoceratops (CMN 8889) well above the ventral edge. The contact between the angular and dentary differs in Chaoyangsaurus (IGCAGS V371) and Psittacosaurus, in which the dentary appears to overlap the angular along the ventral edge of the jaw.

The angular does not bear an everted ridge along its caudoventral edge as in Leptoceratops (CMN 8889) and Protoceratops (AMNH 6466). Two small bony tubercles are present near the caudal end of the angular, one adjacent to the suture with the surangular and the other, larger one along the ventral margin. In Liaoceratops three tubercles are present, but are all linearly arranged along the ventral margin of the bone (Xu et al., 2002). At its caudal end, the angular is deflected medially and embraces the articular from below.

Dentition

Neither specimen preserves the premaxillary bones, but a single premaxillary tooth appears to be preserved with the referred specimen (fig. 15C). The tooth resembles the premaxillary teeth of Liaoceratops in having rostral and caudal carinae, unlike the more cylindrical premaxillary tooth crowns of more derived taxa such as Protoceratops (Sereno, 2000). Minute, irregular serrations are present along the basal, unworn section of each carina. Minute serrations are also present on the distal carina of one of the premaxillary teeth in the holotype of Liaoceratops (P.J. Makovicky, personal obs.). Both sides of the crown bear a thin layer of enamel. The labial face of the tooth is highly convex, but the lingual side is less convex and bears a small wear facet distally. The root is columnar and almost as wide as the crown. Features that identify this tooth as a premaxillary tooth include the presence of enamel on both crown faces, the near-absence of denticles, lack of a primary ridge, and absence of notch at the distal crown base where the following tooth crown would fit. All “cheek” teeth possess a distinct primary ridge in the holotype specimen of Yamaceratops. The unexpected presence of a small wear facet on the lingual face indicates occlusion with the horny ramphotheca of the predentary.

Figure 15

Isolated teeth from the referred specimen of Yamaceratops dorngobiensis (IGM 100/1303). A: Larger cheek tooth in nonocclusal view. B: Probable anterior maxillary tooth in labial view. C: Premaxillary tooth in labial view. Abbreviations are listed in appendix 3.

The maxillary teeth increase in size caudally along the tooth row, with the anteriormost crown being only about half as tall as the crowns near the caudal end of the tooth row. All maxillary crowns are asymmetric around the primary ridge, with the mesial part wider than the distal part. Secondary ridges mark the labial crown faces and terminate in small denticles along the unworn crown tips (fig. 15B). There are five or six denticles along the mesial side of the crown and three along the distal face on the unworn maxillary teeth near the middle of the left tooth row of the holotype. The primary ridges are low and rounded in the smaller, rostral tooth positions, but become better defined on the larger more posterior teeth.

With the exception of the rostralmost tooth, all maxillary teeth have ovoid crowns with mesial and distal edges that are subparallel for much of the crown height. A shallow, mesial groove extends from the tooth onto the base of the crown between the labial and lingual crown faces. It accommodates the rounded distal edge of the root of the preceding tooth. The teeth are closely packed as in other neoceratopsians (Sereno, 2000) and appear to erupt in two, highly coordinated series of Zahnreihen (Edmund, 1960). The roots are elliptical in cross section and wider transversely than mesiodistally. In contrast to the premaxillary tooth, all maxillary teeth have crowns that are distinctly wider than their respective roots (fig. 15). Although many crowns are either obscured by occlusion with the lower jaw or suffer from poor preservation, all appear to have single wear facets, angled at about 30° to the vertical axis of the tooth. All maxillary and dentary teeth have enamel restricted to the nonocclusal crown faces.

The dentary teeth resemble the maxillary ones, but are proportionately wider and more rounded in lingual aspect (fig. 16). A crestlike primary ridge is present on all dentary teeth, and it subdivides the lingual crown face into a smaller mesial and larger distal parts. Distal secondary ridges are parallel with the primary ridge, but the mesial ones are oriented obliquely to the primary ridge. The number of denticles is 4–5 on the distal edge and 3–4 on the mesial one.

Figure 16

Left dentary tooth row of the holotype specimen of Yamaceratops dorngobiensis (IGM 100/1315). Abbreviations are listed in appendix 3.

Vertebrae

One cervical vertebra, an isolated centrum, and a midcaudal neural arch are the only vertebral elements preserved with the referred specimen. The cervical centrum is poorly preserved, but the neural arch is in better shape, and reveals a short, erect neural spine and highly angled zygapophyseal articular facets. The prezygapophyses only project for a very short distance rostral to the spine, and do not extend beyond the anterior intercentral articulation. The neural canal is very wide. Comparison with vertebral column of Protoceratops suggests that these anatomical features are found near the front end of the cervical series in the third and fourth cervicals. In neoceratopsians the third cervical fuses to the axis to form part of the syncervical early in ontogeny, which suggests that the present element is the fourth cervical.

The isolated centrum has an open suture for contact with the arch, indicative of immaturity (Brochu, 1996). The centrum lacks a chevron facet indicating that it is not a caudal. There is no sign of a parapophysis, which suggests it may be a trunk vertebra in which the parapophyses are situated on the arch.

The isolated midcaudal neural arch (fig. 17) is typical of basal Neoceratopsia in having an elongate and erect neural spine, the tip of which is lost. The closely spaced zygapophyses have near-vertical articular facets and are close together, and the neural canal is narrow.

Figure 17

Caudal neural arch of the referred specimen of Yamaceratops dorngobiensis IGM 100/1303) in lateral view. Abbreviations are listed in appendix 3.

Pelvic and Hindlimb Elements

A partial left ilium (fig. 18) was associated with the skull bones of the referred specimen of Yamaceratops. It preserves the pubic and ischial peduncles as well as the postacetabular blade, but the narrow preacetabular blade has been lost. The pubic peduncle is slender and directed anteroventrally, whereas the ischiadic peduncle is robust and bears a large convex antitrochanter. Unlike in Archaeoceratops, the pubic peduncle has a more rounded lateral edge, and the antitrochanter is not excavated by a fossa (You and Dodson, 2003). The iliac blade is low and vertical above the acetabulum. It tapers very slightly toward the distal end, which is crushed, but appears rounded as in other ceratopsians. On its medial face (fig. 18B) impressions for the sacral ribs and transverse processes of sacrals 2–5 are visible and resemble the pattern in other basal neoceratopsians.

Figure 18

Left ilium of the referred specimen of Yamaceratops dorngobiensis (IGM 100/1303) in lateral (A) and medial (B) views. Abbreviations are listed in appendix 3.

A distal end of a right femur was recovered with the referred specimen. It does not differ markedly from the femur of other basal ceratopsians. A pedal phalanx and an ungual (fig. 19) were also recovered. The phalanx is longer than wide and broader than tall in proximal view with a transversely convex articular surface. The distal articular surface is marked by wide, shallow ginglymus. The ungual is missing the tip, but would have terminated in a pointed claw rather than a rounded hoof. It is only slightly curved in lateral view, and has deeply incised claw sheath grooves (fig. 19A).

Figure 19

Pedal ungual of the referred specimen of Yamaceratops dorngobiensis (IGM 100/1303) in dorsal (A) and ventral (B) views.

Taxonomic Sampling

The matrix comprises 20 putatively valid taxa at generic level, except for two species of Psittacosaurus. The validity of all but two of these taxa was recently reviewed and supported by autapomorphies (Makovicky, 2002). The two taxa Asiaceratops and Graciliceratops are of more questionable status. Asiaceratops salsopaludalis was erected for an assemblage of isolated elements from the Cenomanian Khodzhakul Svita of Uzbekistan (Nessov et al., 1989; Nessov, 1995). The holotype is a left maxilla containing four teeth, but a large number of other isolated ceratopsian elements from the same formation have been referred to this taxon (Nessov et al., 1989; Nessov, 1995). Although not directly associated with the holotype, these elements correspond roughly in size, but referral is tentative at best and is solely based on cooccurrence within the same geological formation and on overall size. More compelling is that a number of elements are replicated (two maxillae, at least four partial dentaries, two basicrania, several surangulars) and all belong approximately to a single size range and bear identical morphologies. Although hardly conclusive evidence for a distinct taxon, interpreting these remains as pertaining to a single taxon is more conservative than interpreting them as deriving from multiple species (Makovicky, 2002). The possibility that this material stems from juveniles, as indicated by low tooth counts in the preserved jaw elements, detracts from this argument, however, as juvenile ceratopsians often lack diagnostic characters (Sampson et al., 1997). The holotype maxilla is not diagnostic by itself, but assuming that all referred material belongs to the same taxon as the maxilla, one character may prove diagnostic. In an isolated basicranium, tentatively referred to Asiaceratops salsopaludalis (TsNIGRI 496/12457), a triangular process of the basioccipital fits into a corresponding gap on the posterior face of the basisphenoid on the ventral surface of the braincase. In other taxa, this suture is either flat or there is an embayment on the basisphenoid, but no corresponding process on the basioccipital. Therefore, with the caveat that its status as a monophyletic taxon has yet to be determined, Asiaceratops is tentatively included in the phylogenetic analysis.

Graciliceratops is based on a single immature individual (MgD I/75) comprising a partial skull and postcranium. Sereno (2000) recently reviewed the status of this taxon and separated it from Microceratops, which is a nomen dubium. Although Sereno (2000) diagnosed Graciliceratops based on the slender nature of its frill and a high tibiofemoral ratio, these features are known to vary with ontogeny in the closely related Protoceratops (Brown and Schlaikjer, 1940b). Makovicky (2002) regarded Graciliceratops as a metaspecies that is provisionally diagnostic based on differential character distributions. Although similar to Protoceratops in most respects, it retains clawlike pedal unguals unlike any other coronosaur. It is included in the present analysis.

Several recently named taxa were not included in the analysis, because their validity is considered questionable and all may represent junior synonyms of known taxa. These taxa in question are Bainoceratops (Tereschenko and Alifanov, 2003), Lamaceratops, and Platyceratops (Alifanov, 2003) from the Late Cretaceous of Mongolia, and Magnisrostris (You and Dong, 2003) from the Bayan Mandahu locality in Inner Mongolia. The three latter taxa have been suggested as close relatives of Bagaceratops, but purportedly differ from the ontogenetic series of Bagaceratops at the Paleobiological Institute at the Polish Academy of Sciences described by Maryanska and Osmólska (1975). Unfortunately, these comparisons are insufficient, as the specimens in the Polish collections are all juveniles and very poorly preserved. Some of the characters listed by Alifanov (2003) to distinguish Lamaceratops and Platyceratops from Bagaceratops, such as various skull proportions, jugal horn shape, and nasal horn size and shape, are developmentally variable characteristics in ceratopsians and are observed to change with ontogeny in the closely related Protoceratops andrewsi. Other supposed diagnostic features of the three Bagaceratops relatives are artifacts of preservation (e.g., incipient orbital horn cores described in Magnirostris; You and Dong, 2003), are observed to vary individually among specimens of Bagaceratops (e.g., profile of the nasal horn), or cannot be compared to any of the Polish Academy of Sciences specimens, because the latter do not preserve the relevant parts (e.g., rostral of Magnirostris, frill fenestrae of Platyceratops). At present, all of the newly described specimens display autapomorphies of Bagaceratops (Makovicky, 2002) such as a fused nasal horn, presence of an accessory antorbital fenestra, a long maxillary diastema rostral to the tooth row, and a crestlike mandibular ridge on the dentary. They lack convincing character differences from Bagaceratops, however, that can be established across a wide range of specimens from the known growth series of this taxon. Moreover, all three taxa were collected at localities from which specimens of Bagaceratops have been reported, namely Khulsan (Maryanska and Osmólska, 1975), Khermeen Tsav (Maryanska and Osmólska, 1975), and Bayan Mandahu (Dong and Currie, 1993). Although it is not unreasonable to expect multiple species of Bagaceratops across various localities (and presumably different time intervals), a more thorough survey of characters across the range of known specimens is required to uphold the proposed generic distinctions. Therefore, we take the conservative view and consider these taxa as junior synonyms of Bagaceratops in this generic level analysis.

The recently named taxon Bainoceratops (Tereschenko and Alifanov, 2003), from the Shabarakh Usu (Flaming Cliffs) locality, was distinguished from the cooccurring Protoceratops based on details of vertebral morphology, including the presence of undulatory prezygapophyseal articular facets on the posterior trunk vertebrae. This feature is, however, observed in some, though not all, specimens referred to Protoceratops (AMNH 6424) and in the holotype of Archaeoceratopstalic>. It is also present in a juvenile specimen of Montanoceratops (MOR 542), although not in the larger holotype, suggesting that this feature is subject to individual or perhaps ontogenetic variation. The remaining vertebral features used to distinguish Bainoceratops are rather subtle and seem to fall within the variation observed in Protoceratops (P.J. Makovicky, personal obs.).

Character Sampling

The list of characters (appendix 1) represents an expanded version of the Xu et al. (2002) character set, which in turn was largely derived from Makovicky (2001). We have augmented those character sets with several characters developed in other recent treatments of basal ceratopsian relationships (Sereno, 2000; Makovicky, 2002; You and Dodson, 2003), although some of these were rescored relative to their original usage. Other characters from these works were reviewed but not added, either because of overlap with characters already employed or because they were too vaguely defined to capture the morphological complexity at hand. Almost all observations were made firsthand on original specimens and personal photographs thereof.

Analysis and Results

The complete matrix (appendix 2) comprises 133 characters and 20 taxa and was analyzed using the branch-and-bound parsimony algorithm in PAUP* 4.0b10 (Swofford, 1998). All characters were parsimony informative, equally weighted, and considered unordered. Branches were collapsed if their minimum length equals zero, as such branches are not supported under all optimizations. Strict consensus methods were used to assay the commonality of branching patterns among all most parsimonious trees (MPTs). The taxa Hypsilophodon and Stegoceras were used as outgroups with trees rooted on the former according to current hypotheses of ornithischian relationships (e.g., Sereno, 1999). A NEXUS version of the matrix can be downloaded from  http://fm1.fieldmuseum.org/aa/Files/pmakovicky/Yamaceratops matrix.

Phylogenetic analysis resulted in four MPTs, the strict consensus of which is shown in figure 20A. Yamaceratops occupies a phylogenetic position as a basal neoceratopsian intermediate between Liaoceratops and Archaeoceratops. Tree length is 222 steps, the consistency index and rescaled consistency index are 0.703 and 0.566, respectively, and the retention index is 0.806. Seven unambiguous synapomorphies (chars. 29, 40, 57, 82, 84, 86, 105) unite Yamaceratops with higher neoceratopsians, and four unambiguous synapomorphies (chars. 28, 94, 100, 113) separate Yamaceratops from the clade comprising Archaeoceratops and more derived neoceratopsians.

Figure 20

A: Strict consensus of four most parsimonious trees (MPTs) resulting from analysis of the 133-character, 20-taxon matrix in appendix 2. B: Strict consensus of four MPTs from an analysis in which Asiaceratops was excluded. Numbers next to nodes indicate branch support ( =  decay index) values.

Support for the phylogenetic position of Yamaceratops was assessed using branch support values (Bremer, 1988, 1994) (fig. 20). Despite the relatively large number of synapomorphies supporting the position of Yamaceratops and separating it from more derived neoceratopsians, branch support values for the most exclusive node that includes Yamaceratops, as well nodes immediately above and below it, are remarkably weak. Taxon deletion experiments indicate that the low support is caused by inclusion of the poorly known and questionably diagnosed Asiaceratops salsopaludalis. Following removal of this contentious taxon, support for the nodes at the base of Neoceratopsia below the position of Asiaceratops increases noticeably (fig. 20B) and the 13-step increase in tree support (the sum of all branch supports; Bremer, 1994) markedly exceeds the two-step decrease in tree length achieved through exclusion of Asiaceratops. This demonstrates that the low support for basal nodes in the tree is strongly linked to homoplasy and especially missing data in Asiaceratops, rather than being caused by more broadly distributed homoplasy across the lower nodes of the tree. In other words, instability at the Asiaceratops–leptoceratopsid node also adversely affects several nodes basal to it, although it has no apparent effect on more derived clades (i.e., Coronosauria and Leptoceratopsidae). Codings for Asiaceratops were observed on individual, isolated bones and teeth, and the problems regarding validity of this taxon and referral of isolated elements to it were discussed above. It is noteworthy that the removal of Asiaceratops affects neither of the diagnoses for the Yamaceratops + higher neoceratopsian clade and the next grouping up the tree that excludes Yamaceratops. This again indicates that many of the poor support values were related to missing data in Asiaceratops and the ambiguous character optimizations it caused, rather than homoplasy.