Osteology, phylogenetic affinities, and palaeobiogeographic significance of the bizarre ornithischian dinosaur Ajkaceratops kozmai from the Late Cretaceous European archipelago Author Czepiński, Łukasz Institute of Paleobiology, Polish Academy of Sciences, Twarda 51 / 55, 00 - 818 Warsaw, Poland & Institute of Evolutionary Biology, Biological and Chemical Research Centre, Faculty of Biology, University of Warsaw, ul. Żwirki i Wigury 101, 02 - 089 Warsaw, Poland lczepinski@twarda.pan.pl Author Madzia, Daniel Institute of Paleobiology, Polish Academy of Sciences, Twarda 51 / 55, 00 - 818 Warsaw, Poland text Zoological Journal of the Linnean Society 2024 2024-04-29 202 4 1 25 https://doi.org/10.1093/zoolinnean/zlae048 journal article 308176 10.1093/zoolinnean/zlae048 d69ea230-64ae-45d4-b846-5236ae1410d0 0024-4082 14893326 Ajkaceratops kozmai Ősi et al . 2010 Holotype : MTM V2009.192.1 , a proposed rostral, premaxillae, and the partial anteromedial processes of maxillae ( Figs 2–7 ). Referred material: MTM V2009.195.1 , an isolated predentary with the anteriormost portion of the right dentary ( Fig. 8 ) ; MTM V2009.193.1 ( Fig. 9 ) , MTM V2009.194.1 ( Fig. 10A–D ) , and MTM V2009.196.1 ( Fig. 10E–H ), three isolated predentary bones . Locality and horizon: Iharkút, Bakony Mountains, western Hungary ( Fig. 1 ); Csehbánya Formation (Santonian, Upper Cretaceous; see: Szalai 2005 , Botfalvai et al. 2015 , 2021 ). Revised diagnosis: An ornithischian dinosaur with the following unique combination of features (*previously identified autapomorphy; **newly identified autapomorphy): (i) Edentulous premaxilla (shared with derived Neoceratopsia, Iguanodontia, derived Ankylosauria, and Stegosauridae ). (ii) A premaxilla ventral to the external naris and accessory fenestra is shallow relative to its anteroposterior length (derived within Coronosauria, shared only with ceratopsids and protoceratopsids among adults; modified after Ősi et al. 2010 and Morschhauser 2012 ). (iii) Presence of a large oval fenestra between the premaxilla and maxilla, similar in size to the external nares (shared with Bagaceratops rozhdestvenskyi ; in Diabloceratops eatoni , Sinoceratops zhuchengensis , Zuniceratops christopheri , and some chasmosaurine specimens the fenestration is much smaller than the external nares; modified after Ősi et al. 2010 and Morschhauser 2012 ). (iv) The posterior portion of the posterolateral process of premaxilla is directed posterodorsally (shared with Iani smithi , Jeholosaurus shangyuanensis , Lesothosaurus diagnosticus , Thescelosaurus neglectus , and some later-diverging iguanodontians and stegosaurs, but unlike the condition observed in ceratopsians; modified after Ősi et al. 2010 and Morschhauser 2012 ). (v) **A very deep concavity of the premaxillary palate, penetrating more than 60% of the subnarial height of the premaxilla (in all known ceratopsians the concavity is up to ~25% of the subnarial premaxillary height). (vi) **The premaxillary palate is vaulted posteriorly to the posterior border of the external nares (in all known ceratopsians the premaxillary palate is vaulted anteriorly to the external nares). (vii) The internarial bar formed by posterodorsal processes of premaxillae is transversely wider than dorsoventrally high (in contrast to non-ceratopsoid ceratopsians). (viii) **The anteromedial processes of maxillae are placed more ventrally than the posteroventral process of the premaxilla (in all known ornithischians the anteromedial processes of maxillae are roughly in the same plane as the posterior portion of the premaxillary palate). (ix) *A high buccal margin of predentary limited to the posterior portion of the bone, sharp and not bevelled (modified after Ősi et al. 2010 and Morschhauser 2012 ). (x) The ventral process of the predentary bone is not bifurcated posteriorly (in contrast to early-diverging neornithischians and ceratopsians, with the exception of Psittacosaurus spp. ). (xi) **The anterior part of the dorsal margin of the dentary is elevated, forming a tubercle placed significantly dorsally than the posterolateral process of the predentary bone. Additional features can be added to the diagnostic characters providing that the interpretation of the anteriormost portion of the upper jaw as a rostral is correct: (i) Unkeeled rostral bone (shared with Liaoceratops yanzigouensis , Mosaiceratops azumai , Psittacosaurus spp. , Yamaceratops dorngobiensis , and Zuniceratops christopheri ). (ii) **Rostral and predentary bones with flat occlusal surfaces anteriorly, deprived of the cutting lateral edges in its anterior portion, resulting in a subtriangular transverse cross-section of the bones (in contrast to all known ceratopsians, where the rostral bone is excavated anteriorly creating space for the predentary bone). Figure 2. Ajkaceratops kozmai MTM V 2009.192.1, holotype specimen in left lateral view. (iii) **Rostral and predentary bones ornamented with pits, which are absent in their anteroventral tips (in all ceratopsians the anteroventral tip is usually the most rugose part of the rostral bone, and the rugosity is mostly manifested through channels and ridges rather than pits). Figure 3. Ajkaceratops kozmai MTM V 2009.192.1, holotype specimen in right lateral view. (iv) **Rostral extensively covering the premaxilla in the palatal direction, reaching nearly as posteriorly as the buccal process (in all known ceratopsians the buccal process terminates significantly more posteriorly than the palatal extension of the rostral bone). Figure 4. Ajkaceratops kozmai MTM V 2009.192.1, holotype specimen in ventral view. The posterodorsal processes of premaxillae were digitally removed for clarity. (v) The posterior margin of the rostral bone is deeply concave anteriorly in lateral view (shared with Zuniceratops christopheri and Chasmosaurinae). Remarks According to the original diagnosis by Ősi et al. (2010) , Ajkaceratops kozmai represented a coronosaur ceratopsian with the following character combination: ‘(1) large oval accessory fenestra are present between the premaxilla and maxilla, with the nasal excluded from its margin; (2) the part of the premaxilla ventral to the external naris and the accessory fenestra is dorsoventrally shallow relative to its rostrocaudal length; (3) the posterolateral process of the premaxilla is curved along its length becoming nearly horizontal caudally (autapomorphy); (4) the buccal margins of the predentary are sharp and not beveled (autapomorphy)’. With respect to character ‘1’, however, it is uncertain whether the nasal was excluded from the dorsal margin of the fenestra. In turn, the dorsoventrally shallow body of the premaxilla (diagnostic character ‘2’) has quite a wide distribution among ornithischian dinosaurs, as pointed out by Morschhauser (2012) . According to Morschhauser (2012) , Ajkaceratops kozmai can bediagnosedas'aceratopsianwiththefollowingautapomorphies and unique combination of characters: (1) large oval fenestra between the premaxilla and maxilla, with the nasal excluded from their margin (shared with Bagaceratops ); (2) the premaxilla ventral to the external naris and accessory fenestra is shallow relative to its rostrocaudal length (derived within Coronosauria, shared only with ceratopsids and protoceratopsids among adults); (3) posterolateral process of the premaxilla is curved, becoming nearly horizontal at its caudal end (autapomorphy); (4) the buccal margin of the predentary is sharp, not beveled (plesiomorphic for all taxa crownwards of Liaoceratops )'. In fact, with respect to character ‘1’, in Bagaceratops and other ceratopsians, the nasal seems to always contribute to the margin of the fenestra ( Czepiński 2020 ), and it is unclear whether it was excluded in Ajkaceratops . Figure 5. Ajkaceratops kozmai MTM V 2009.192.1, holotype specimen in dorsal view. Description and comparisons The type specimen of Ajkaceratops kozmai (MTM V2009 . 192.1) includes both premaxillae and the anteriormost portion of the anteromedial processes of both maxillae. The rugose anterior region of the snout has been originally interpreted as a rostral bone that would be fully and indistinguishably fused to the premaxillae ( Figs 2 , 3 , 7 , 11 ). The maximum length of the specimen, as measured from the anteroventral tip of the beak to the posteriormost tip of the right posterolateral process, equals 87 mm . In turn, the maximum dorsoventral height, as measured from the anteroventral tip of the beak to the dorsal surface of the posterodorsal process, equals 70 mm . The specimen shows a partial taphonomic deformation, with the posterior half of the left premaxilla being broken and slightly displaced posteroventrally and a bit medially ( Fig. 2 ), and the posteroventral part of the right premaxilla being displaced posteromedially. The right side of the specimen is slightly ‘elevated’ ( Fig. 3 ), resulting in the posterolateral process of the right premaxilla being displaced posterodorsally. Figure 6. Ajkaceratops kozmai MTM V 2009.192.1, holotype specimen in posterior view. Proposed rostral The anteroventral part of the snout of Ajkaceratops kozmai has been interpreted by Ősi et al. (2010) to represent a rostral bone. Although the overall shape seems to be similar to the rostral bone of a ceratopsian dinosaur, the structure differs from that in a number of features, including the texture and detailed morphology of the element. Additionally, due to its suggested complete fusion with premaxillae, the proposed margins of this structure are obscured. Below, we provide a detailed osteological redescription of the anteriormost section of the upper jaw of Ajkaceratops kozmai , highlighting especially the peculiarities of its anatomy. The main body: The anterior surface of the purported rostral in Ajkaceratops is smooth, lacking the anterior keel typical for the rostral bones in most neoceratopsians (except Mosaiceratops , Yamaceratops , and the early-diverging ceratopsoid Zuniceratops ; Makovicky and Norell 2006 , Zheng et al. 2015 , Son et al. 2022 ). In the earliest ceratopsian dinosaurs, such as Yinlong downsi , there is a weak medial keel developed ventrally; however, there is no indication of a keel dorsally ( Han et al. 2016 ); a weakly-developed keel was also observed in some specimens of Liaoceratops ( Yang et al. 2017 ) . Sharp keel is present in later-diverging neoceratopsians ( Archaeoceratops , Auroraceratops , Aquilops , and the majority of euceratopsians). In Psittacosaurus the rostral bone lacks any keel ( Sereno 2010 ), and the element itself is rather thin and U-shaped in ventral view. The beak terminates anteroventrally to form a pointed tip, which is 16 mm below the ventral edge of the premaxilla. Ősi et al. (2010) described this element as being relatively short and curved ventrally, terminating as a sharp point and noted that this condition resembled that observable in Archaeoceratops oshimai , Protoceratops andrewsi , and Bagaceratops rozhdestvenskyi . However, the anteroventral tip of the Ajkaceratops kozmai snout has a relatively flat, smooth, ventral surface, lacking the developed lateral edges ( Figs 4 , 7C ), resulting in the subtriangular cross-section of the structure ( Fig. 12A ). As such, the appearance of the bone is more robust. In all ceratopsians the ventral surface of the rostral bone is vaulted for accommodating the predentary bone, and the lateral cutting edges of the rostral are evident in the anterior portion, forming the typically V-shaped cross-section of the bone ( Fig. 12F–K ). Such a condition is observable even in the earliest-diverging ceratopsians [e.g. Yinlong downsi (IVPP V14530) and Psittacosaurus spp. ]. Figure 7. Ajkaceratops kozmai MTM V 2009.192.1. A, B, close-up on the broken part of the internarial bar in left posterolateral (A) and posterior (B) views, with no visible osteological boundaries between the premaxilla and the purported rostral. C, ventral surface of the snout in the right lateroventral view. The pits cover a significant area posteriorly, with no traces of the margin between the purported rostral bone and the premaxilla. D, anterior tip of the purported rostral bone in the left anterolateral view. In ventral view, the proposed rostral covers the premaxillary palate extensively ( Fig. 4 ). The pitted surface of the bone extends far posteriorly to the point where the anteromedial premaxillary wings are not visible ( Figs 4 , 7C ). The part of the vaulted premaxillary palate is not preserved in its midpoint and it is not clear how far posteriorly the proposed rostral bone develops. In any case, such condition contrasts with that seen in all ceratopsians, where the premaxilla is clearly visible in the ventral view and anteriorly wedged between the rostral buccal processes, with the medial walls of the rostral bone steeply inclined dorsomedially. No rugose surface covering the palate area, as it is in Ajkaceratops , is visible in any ceratopsian. Texture: The anteriormost portion of the snout of Ajkaceratops kozmai is rugous and covered with pits, indicating the presence of a keratinous beak during the animal’s life. Two distinctly-sized types of pits can be distinguished. The larger ones (approximately 3 mm in diameter) are relatively rare and shallow, concentrated mainly in the anterior region, although the anteroventral tip is deprived of any larger pits ( Figs 2 , 3 , 7D ). The smaller pits (less than 0.5 mm in diameter) are present around the whole snout and are not limited only to the rostral bone. In the early-diverging ceratopsian Yinlong downsi , the rostral bone was most likely texturized, with shallow, longitudinal grooves, mostly oriented dorsoventrally ( Han et al. 2016 ; Fig. 11B ). In some specimens of Psittacosaurus spp. , the ornamentation of the rostral bone is slightly more reminiscent of that observed in the Ajkaceratops snout, showing a number of shallow pits. Most of the Psittacosaurus specimens do not exhibit any longitudinal grooves (e.g. Psittacosaurus sp. MPC-D 100/610; see Fig.11C ; Psittacosaurus lujiatunensis CAGS-IG-VD-004), while in some the grooves are very short [e.g. Psittacosaurus lujiatunensis (IVPP V12617, ZMNH M8137)]. Still, in Psittacosaurus the pits are mostly distributed adjacent to the anteroventral tip, in contrast to Ajkaceratops . In neoceratopsian dinosaurs, such as Archaeoceratops oshimai (IVPP V11114), Aquilops americanus (OMNH 34557), Yamaceratops dorngobiensis (MPC-D 100/1303), and Protoceratops andrewsi ( Fig. 11E ), the rostral bone is mostly covered in longitudinal grooves, with pits being comparatively larger than those in Ajkaceratops , and limited solely to the posterior portion of the bone. In leptoceratopsid dinosaurs, such as Leptoceratops gracilis (CMN 8889) and Udanoceratops tschizhovi (PIN 3907/11), the pits are more densely distributed, and most of them fuse together dorsally, forming the longitudinal channels ( Fig. 11D ). There are no longitudinal channels or ridges in Ajkaceratops . The only longitudinal marks observable in Ajkaceratops kozmai are shallow and present along the posterior contact between the rugose region and the main body of the premaxilla ( Fig. 2 ). If that element is indeed a rostral bone, these structures may derive from the fusion between the rostral and the premaxillae. In centrosaurine ceratopsids, such as Centrosaurus apertus (USNM 8897; see Fig. 11F ), Pachyrhinosaurus lakustai (TMP 1986.55.258), and Styracosaurus albertensis (USNM 14765), the pits are relatively large and deep, and are concentrated solely in the anteroventral portion of the snout. A very similar condition is visible in the chasmosaurine ceratopsid Chasmosaurus russelli (e.g. TMP 1981.19.175), while in some specimens of the chasmosaurine Triceratops spp. the ornamentation is somewhat similar to that seen in Ajkaceratops , showing a scarcity of the longitudinal grooves (see MOR 2574 and MOR 1625; Horner and Goodwin 2008 ). In some specimens of Triceratops spp. there are also longitudinal channels, mostly in the posterior and the dorsalmost region of the rostral bone (e.g. LACM 45807, USNM 1201, USNM 8081). Figure 8. Ajkaceratops kozmai MTM V 2009.195.1, referred predentary and a dentary fragment in left lateral (A, B), right lateral (C, D), anterior (E, F), dorsal (G, H), and ventral (I, J) views. Additionally, in Ajkaceratops the large pits cover most of the anterior portion of the snout, also in the ventral view ( Figs 4 , 7C ). In ceratopsian dinosaurs, however, the grooves and pits are limited solely to the anteriormost portions of the ventral aspects of the rostrals, and their density is significantly lower than on the lateral sides [see, e.g. Psittacosaurus sp. (MPC-D 100/610); Anchiceratops ornatus (CMN 8535); Centrosaurus apertus (CMN 348, CMN 8795); Styracosaurus albertensis (USNM 14765)]. The dorsal process: The proposed rostral bone apparently develops posterodorsally along the anterior surface of the premaxillary internarial bar, as indicated by the rugose texture. The dorsal process of the rostral bone would, therefore, be very long, reaching to at least the midpoint of the external nares ( Figs 2 , 3 , 5 ). It is not certain whether the proposed rostral bone would contact the nasals; however, the texture on the internarial bar indicates that it terminated just anteriorly to the purported suture with the nasals ( Figs 5 , 7B ). Among ceratopsians the rostral bone typically does not contact with the nasals, with the exception of Psittacosaurus spp. , in which the elongated anteroventral nasal processes meet the rostral below the external nares ( Sereno 2010 ). Figure 9. Ajkaceratops kozmai MTM V 2009.193.1, referred predentary in right lateral (A), anterior (B), dorsal (C), left lateral (D), posterior (E) and ventral (F) views. In Ajkaceratops the dorsal process is as long as, or longer than, the buccal processes of the proposed rostral bone, and the posterior margin of the bone is deeply excavated. Elongation of the dorsal process is not typical for ceratopsians, and can be observed only in some specimens of non-ceratopsid euceratopsians [ Protoceratops andrewsi (AMNH FARB 6637, MPC-D 100/522); Udanoceratops tschizhovi ( Kurzanov 1992 , PIN 3907/11)]; though in these specimens the posterior margin of the rostral bone is straight, not as deeply concave as in Ajkaceratops . In turn, similar morphology is visible in the ceratopsoid Zuniceratops (MSM P4185; Wolfe et al. 2010 ; Fig. 12J ), and the chasmosaurine ceratopsids, where the rostral bone is slender, elongated, and deeply excavated posteriorly in lateral view ( Dodson et al. 2004 ). The buccal process: The buccal process of the beak is 22 mm long, covering nearly half of the ventral margin of the premaxilla. Such a condition is reminiscent of the well-developed lateral process of the rostral bone observable mostly in neoceratopsian dinosaurs ( Makovicky and Norell 2006 ; however, see: Han et al. 2016 ). Protoceratops was reported to lack the buccal process of the rostral bone due to a sheet of bone extending between the buccal process and the rostral, and covering a large area of the premaxilla ( Lambert et al. 2001 , Morschhauser et al. 2018a ). Yet, in some specimens of Protoceratops the buccal process is visible (e.g. AMNH FARB 6429, 6637, PIN 614/63). The ventral margin of the buccal process of Ajkaceratops is arched in lateral view, as in neoceratopsians. Premaxillae The premaxillae in the type specimen of Ajkaceratops kozmai (MTM V2009.192.1) are almost completely preserved. Only the posteriormost portions of the posterodorsal processes and the ventral surface of the premaxillary symphysis (just anterior to the anteromedial processes of maxillae) are not preserved ( Figs 4 , 6 ). The length of the premaxilla (from the anterior tip of the subnarial fossa to the posterior tip of the posterolateral process) is 67 mm . The length along the ventral edge (between the anterior margin of the subnarial fossa and the posterior tip of the posteroventral process) is 42 mm . The premaxillae of Ajkaceratops kozmai are edentulous, with no traces of alveoli. In the anterior half, their ventral margins are covered by a rugose buccal process of the purported rostral bone, being visible only in the posterior part. The ventral margin of the posterior part of premaxilla forms a rather sharp cutting edge. Among the ornithischian dinosaurs edentulous premaxilla is present in later-diverging stegosaurs ( Raven and Maidment 2017 ), later-diverging ankylosaurs ( Arbour and Currie 2016 ), in later-diverging iguanodontians ( Madzia et al. 2020 , Poole 2022 ), and in some ceratopsians, such as Psittacosaurus spp. , Mosaiceratops azumai , later-diverging leptoceratopsids, some protoceratopsids ( Bagaceratops rozhdestvenskyi and Protoceratops hellenikorhinus ), and all Ceratopsoidea ( Maryańska and Osmólska 1975 , Chinnery 2004 , Sereno 2010 , Wolfe et al. 2010 , Zheng et al. 2015 , Czepiński 2020 ). The premaxilla is dorsoventrally shallow, as noted by Ősi et al. (2010) . The dorsoventral height of the main body of premaxilla (measured at the level of the anterior margin of the external nares) is 18 mm , while the height of the whole premaxilla (with the posterodorsal process) is 32 mm . Thus, the main body of the premaxilla is twice as long as it is high. The premaxilla is usually longer than high in ceratopsians (except Psittacosaurus spp. ), but only in some neoceratopsians [ Aquilops americanus ( Farke et al. 2014 , OMNH 34557)] its ratio is similar to that obtained for Ajkaceratops . It differs strikingly from the rather high premaxillae seen in protoceratopsids and leptoceratopsids. Figure 10. Ajkaceratops kozmai MTM V 2009.194.1 (A–D) and MTM V2009.196.1 (E–H), referred predentaries. MTM V2009.194.1 in dorsal (A), ventral (B), left lateral (C), and right lateral (D) views. MTM V2009.196.1 in posterior (E), left lateral (F), right lateral (G), and dorsal (H) views. The visible parts of the premaxillae are not fused to each other in Ajkaceratops . The posterodorsal processes preserved the suture, visible in the ventral view ( Fig. 6 ), and on the anterior and dorsal margins of the external nares ( Fig. 7A, B ). The posterodorsal processes: The posterodorsal processes of the premaxillae are broken just anteriorly to the contact with the nasals ( Figs 5 , 7B ). The rugose texture of the beak continues to be present on the anterior surface posterodorsally; however, in the posteriormost portion of the bar, the pits are absent and there are longitudinal striations present, suggesting that the contact with dorsally overlapping nasals was developing nearby. Each of the processes measures 6 mm in transverse width posteriorly. Dorsally to the external nares, the internarial bar formed by both posterodorsal processes is transversely wider ( 12.5 mm ) than it is dorsoventrally high ( 4.5 mm ), due to the laterally pronounced contact with the rugose anterior part. Such a condition differs from that observable in the internarial bars of ceratopsian dinosaurs, where the posterodorsal processes of premaxillae are laterally flattened (in, e.g. Yinlong , Archaeoceratops , Protoceratops , and Bagaceratops ). This is true even for material that shows dorsoventral deformation, such as the Leptoceratops gracilis specimen CMN 8889. In Psittacosaurus spp. the internarial bar is wider than high, but it is formed by the nasal bones instead of the premaxillae. Only in some ceratopsoids, the premaxilla becomes transversely wider than high, which is due to the development of the rugosities related to the presence of the nasal horn [e.g. Centrosaurus (UALVP 11735)], although transverse widening of the internarial bar is visible also in Zuniceratops (MSM P2225) . The laterally wide lateral ridge on the posterodorsal processes of premaxillae was also observed in the rhabdodontomorphs Iani smithi ( Zanno et al. 2023 ; NCSM 29373) and Rhabdodon sp. (MC-CY.QR.3; Chanthasit 2010 ). In Ajkaceratops both processes lean ventromedially, forming a slight ridge along the suture on the ventral side of the internarial bar that anteriorly develops into the medial wall of the subnarial fossa. The subnarial fossa: A lateral excavation is present on both premaxillae of Ajkaceratops anteroventrally to the external nares. To some point the excavation is shallower than the truly distinct subnarial fossa visible in some late-diverging ornithischians (hadrosauroids, ceratopsids). The fossa in Ajkaceratops is placed more ventrally than that in non-ceratopsoid ceratopsians, as noted by Ősi et al. (2010) . In the majority of ceratopsian dinosaurs a much more subtle concavity is present ventral to the external nares (e.g. Yinlong , Auroraceratops , and Archaeoceratops ). However, in some taxa ( Bagaceratops , most specimens of Protoceratops ) the fossa is not visible, as the lateral wall of the premaxilla and its posterodorsal and posterolateral processes are placed in about the same plane. In Ajkaceratops , the posterolateral processes are placed more laterally. In turn, in Leptoceratops gracilis (CMN 8887 and CMN 8880) the narial fossa is formed by a concavity on the posteroventral corner of the external nares. Figure 11. Simplified comparison of the ornamentation of the anterior portion of the snout of Ajkaceratops kozmai (A), and the rostral bone of the ceratopsian dinosaurs (B–F). B, Yinlong downsi (based on IVPP V18637); C, Psittacosaurus sp. (based on MPC-D 100/610); D, Udanoceratops tschizhovi (based on PIN 3907/11); E, Protoceratops andrewsi (based on AMNH FARB 6637); F, Centrosaurus apertus (based on USNM 8897). Not to scale. Figure 12. Comparison of the snouts of selected ornithischian dinosaurs, with emphasis on the shape of the vaulted premaxillary palate and the cross-section of the anterior part (near the posterior border of the rostral bone, as indicated by the dashed line). A, Ajkaceratops kozmai MTM V 2009.192.1; B, Stegouros elengassen CPAP-3165; C, Lesothosaurus diagnosticus NHMUK PV RU B 17; D, Thescelosaurus neglectus NCSM 15728; E, Yinlong downsi IVPP V 14530; F, Psittacosaurus sp. MPC-D 100/610; G, Leptoceratops gracilis CMN 8889; H, Bagaceratops rozhdestvenskyi (based on of ZPAL MgD-I/129 and MPC-D 100/535); I, Protoceratops andrewsi AMNH FARB 6433; J, Zuniceratops christopheri (based on MSM P2225, MSM P3197 and MSM P4185); K, Chasmosaurus sp. CMN 8801. The anterior part of the premaxillary palate is convex even in the early-diverging ceratopsians, while in Ajkaceratops the palatal convexity is displaced posteriorly. Not to scale. In Ajkaceratops the fossa develops anteroventrally, reaching the highest concavity in the anterior portion of the premaxillae, just posteriorly to the contact with the proposed rostral. More posteriorly, it is subtly sloping ventrolaterally. The anterior and ventral borders of the fossa are further pronounced due to a transverse overgrowth of the beak, resulting in a much more lateral placement of the rugose area in the dorsal view, similar to that in Mosaiceratops (ZMNH M8856) and Leptoceratops (CMN 8889). In contrast, in ceratopsid dinosaurs the subnarial fossa is bordered anteroventrally by the thickened oral margin of the premaxilla, not just the rostral bone. Ventrally, both posterodorsal processes form a small, thin crest in the anteroventral corner of the narial fossa ( Fig.7A ). The posterodorsal processes are directly appressed to each other, but with a visible suture between the left and the right premaxilla. In early-diverging ceratopsians and Zuniceratops , the two premaxillary processes meet but remain separate entities, regardless of the ontogenetic stage of the individual. The thin crest in the anteroventral corner of the narial fossa is somewhat similar to the premaxillary septum of ceratopsid dinosaurs ( Figs 2 , 3 , 7A ). However, in ceratopsids the septum is much more developed and occupies the majority of the narial fossa (see: Lund et al. 2016 ). No narial strut, known in ceratopsids, is visible in Ajkaceratops . In the dorsal view, just posterior to the ‘premaxillary septum’, the specimen is weathered, and it is not clear whether the posterolateral processes met medially ( Fig. 6 ). There are at least four foramina visible on the left premaxilla,just behind the rugose part of the snout ( Figs 2 , 3 ). The anterodorsal one is placed immediately posterior to the border of the rugose part, in the deepest part of the subnarial fossa.Posterior to it, there is a large foramen near the ‘premaxillary septum’ that is nearly twice the size of the anterodorsal foramen. The placement of these two foramina is similar to, respectively, the anterior premaxillary foramen and premaxillary foramen ( sensu Sereno 1991 ) reported in some ornithischian dinosaurs (i.e. Thescelosaurus ; Boyd 2014 ), including ceratopsians ( Protoceratops and Auroraceratops ; Morschhauser et al. 2018a ). The morphologies of these two foramina are obscured by the presence of sediment. More ventrally there are two other foramina lying nearly in line with the dorsal ones. All are significantly larger than the pits present on the rugose anterior surface of the snout. The first one is situated just posteriorly to the proposed rostral bone. The second one is placed just anterior to the base of the posterolateral process. On the right premaxilla there are also two prominent premaxillary foramina, although these are placed more closely to each other. More ventrally there is only one foramen present. It is placed posteriorly to the rugose area. Several smaller nutrient foramina are present along the bones. The posterolateral processes: The posterolateral processes of the premaxillae project much more posteriorly than in ceratopsian dinosaurs, resembling more the morphology observed in Thescelosaurus neglectus ( Boyd 2014 ) . That condition is most apparent on the left premaxilla and is slightly pronounced by taphonomic deformation of the specimen in this region. The posterior tip of the posterolateral processes bears an articular surface ( Fig. 6 ). On the laterodorsal surface there are ridges that indicate a contact, most likely with the slightly overlapping nasal bone. That surface of the contact is orientated laterodorsally along the dorsal margin of the process. A similar contact surface is visible also on the medial side, covering the area just posterior to the natural dorsal border of the premaxillo-maxillary fenestra. Posteriorly, there is another ridged surface, orientated posteriorly, that may indicate contact with the posteroventral process of the nasal or with the maxilla. In the majority of ceratopsian dinosaurs, the posterolateral process of the premaxilla overlaps the nasal bone laterally. However, in protoceratopsids ( Protoceratops andrewsi and Bagaceratops rozhdestvenskyi ) the nasal overlaps the posterolateral process of the premaxilla anteriorly, and is overlapped by the premaxilla dorsally and posteriorly. Some striations on the anterior portion of the lateral surface of the posterolateral premaxillary process are visible in the Udanoceratops specimen PIN 3907/11 (Ł.C. pers. obs.), although it is uncertain whether it was related to the lateral overlapping of the bone. In ceratopsians with the premaxillo-maxillary fenestration, such as Bagaceratops rozhdestvenskyi and Diabloceratops eatoni , the posterolateral premaxillary process contacts only the nasal bone, being excluded from contact with the maxilla. In Zuniceratops the posterolateral process is deeply bifurcated and its dorsal tip forms the anterior margin of the premaxillo-maxillary fenestra, while the posterior tip forms the anteroventral one, resulting in contact with the maxilla ( Fig. 12H ). The contact with the nasals seems to be even more extensive in Ajkaceratops , than in protoceratopsids and Diabloceratops . It contrasts with the seemingly limited contact between the premaxilla and nasals in Zuniceratops , although it can be only estimated as the known isolated bones of Zuniceratops most likely belong to multiple individuals ( Wolfe et al. 2010 ). In Zuniceratops the premaxillo-maxillary fenestration occurs on the distalmost portion of the posterolateral process, in contrast to Ajkaceratops , Bagaceratops , and Diabloceratops , in which the margin of the fenestra is formed by the posteroventral part of the process. The premaxillo-maxillary fenestration: Posteriorly, the posterior ramus of the premaxilla develops almost horizontally, forming the edge of the fenestra. Ősi et al. (2010) suggested that this opening is homologous with that observable in the protoceratopsid Bagaceratops rozhdestvenskyi , as it is formed solely by the premaxilla and maxilla. However, as observed in numerous specimens of Bagaceratops , the nasal narrowly contributes to the dorsal edge of the fenestra as well ( Czepiński 2020 ), similar to the condition in the ceratopsoid dinosaurs, such as Zuniceratops and Diabloceratops . Given that it is unclear whether in Ajkaceratops the nasal or maxilla was in contact with the posterior end of the posterolateral premaxillary process, it cannot be confirmed that the nasal bone was excluded from the margin of the premaxillo-maxillary fenestra. The posterolateral processes of the premaxillary bone expand laterally, resulting in the fenestra being located more laterally in relation to the longitudinal axis of the skull than the external nares. This contrasts with the condition present in Bagaceratops , Zuniceratops , and Diabloceratops , in which the fenestra lies roughly in the same plane as the narial openings. On the lateral surface of the base of the posterolateral processes, just anterior to the premaxillo-maxillary fenestra, there is a foramen present in a shallow pocket. Such a foramen has not been traced in any ceratopsian with the premaxillo-maxillary fenestration. The additional foramen is present on the anteromedial wall of the premaxillo-maxillary fenestra, at the base of the posterolateral process of premaxilla ( Fig. 6A ). The premaxillary palate: The premaxillary palate is ventrally vaulted to a high degree and strongly arched. However, the arching is present much more posteriorly (posterior to the external nares; Fig. 12A ) than in ceratopsian dinosaurs, making the anteriormost portion of the snout solid in the cross-section. The dorsalmost part of the premaxillary palate is not preserved and it is not certain how deeply concave it was; however, due to the rugose surface of the proposed rostral bone preserved on the medial walls of the palate, it can be concluded that it reached up to 75% of the subnarial premaxillary height. In ceratopsian dinosaurs, the premaxillary palate is usually vaulted to the level of the lower quarter or third of the premaxillary body ( Fig. 12 ). In Ajkaceratops the vaulted premaxillary palate is restricted posteriorly by the strongly anterodorsally oblique anteroventral maxillary processes, a condition not seen in any ceratopsian dinosaur. The anteriorly concave palatal surface of the premaxilla was also observed in Thescelosaurus neglectus ( Boyd 2014 ) . In Lesothosaurus diagnosticus (NHMUK PV RU B17) and Laquintasaura venezuelae (MBLUZ P-488) the premaxillae are vaulted posteriorly to the level of the anterior margin of the external nares, similar to those in Ajkaceratops kozmai , but in the former the premaxillae are relatively flat anteriorly ( Porro et al. 2015 , Herrera-Castillo et al. 2021 ). In the rhabdodontomorph Iani smithi (NSCM 29373), the ventral shelf of the premaxilla is vaulted posteriorly, at the level of the external nares; however, Zanno et al. (2023) suggested that this condition is caused by the post-mortem deformation of the specimen. No traces of the rostral palatal foramina ( sensu Sereno 1991 ) are visible in Ajkaceratops . The posteroventral processes: The medial projection of the posteroventral part of the premaxilla, which contacts the anteromedial processes of the maxilla, develops medially, resulting in the ventral margin of the extremity being placed more ventrally than the margin of the premaxilla. It may be only slightly pronounced due to the deformation of the specimen. In ceratopsian dinosaurs the contact between the premaxilla and the anterior processes of the maxilla are developed dorsomedially, so the contact with the maxilla is placed much more dorsally than the edge of the premaxilla. In neoceratopsians (such as Leptoceratops and Bagaceratops ) the premaxillae meet together medially in ventral view, just in front of the anteroventral processes of the maxillae. Then they form a roughly horizontal shelf, anteriorly developing into a vaulted concavity. In Ajkaceratops this region is weathered off, therefore, the condition cannot be assessed. The lateral surface of the posteroventral process of the premaxilla of Ajkaceratops further shows a foramen that is located at the bottom of a shallow fossa, just anterior to the contact with the maxilla. In some protoceratopsid specimens [e.g. Protoceratops andrewsi (AMNH FARB 6429), protoceratopsid MPC-D 100/1246], there is a nutrient foramen present near the posteroventral margin of the premaxilla, roughly in the same place, although without any fossa developed. Maxilla Only the anteriormost portions of the anteromedial processes of both maxillae are preserved, measuring 14 mm each, and are wedged between the posteroventral processes of the premaxillae ( Figs 3 , 4 , 6 ). Because of the taphonomic deformation of the specimen, they are slightly displaced. Specifically, the anteromedial process of the right maxilla overlaps the left one dorsally. The intermaxillary and premaxillo-maxillary sutures are clearly visible in the ventral view. They seem to be almost completely preserved anteriorly, although the dorsal surface of the vaulted palate is eroded just in front of them. On the ventral surfaces of both maxillae there are small foramina present medially. Such foramina are sometimes present on the anteroventral processes of maxillae in some ceratopsians, as in Bagaceratops rozhdestvenskyi (ZPAL MgD-I/129). In Ajkaceratops the anteromedial processes develop only to the posterior quarter of the premaxilla. Such a condition differs from that observed in ceratopsian dinosaurs [e.g. Bagaceratops rozhdestvenskyi (ZPAL MgD-I/129), Centrosaurus apertus (USNM 8897), Liaoceratops yanzigouensis (IVPP V12738), Protoceratops andrewsi (AMNH FARB 6637), and Psittacosaurus major (LHPV 1)], in which the anteromedial processes extend up to the midlength of the premaxilla. In Ajkaceratops the preserved portions of the maxillae direct anterodorsally, with an angle of nearly 45° in relation to the longer axis of the ventral margin of the premaxilla. They are placed more ventrally than the posteroventral processes of the premaxillae, resulting in being almost visible in the lateral view ( Fig. 3 ). In all ceratopsian dinosaurs the anteroventral processes of the maxillae are placed much more dorsally, nearly in the same plane as the premaxillary palate, and are deeply hidden in the lateral view by the lateroventral edge of the premaxillae and maxillae. In the holotype of Yamaceratops dorngobiensis (MPC-D 100/1315), the rostral maxillary processes are directed slightly anterodorsally ( Makovicky and Norell 2006 ); however, they are still located deeply dorsally when compared to the ventral edge of the maxilla. In Ajkaceratops the contact between the maxilla and premaxilla seems to be limited to the posterior surface of the posteroventral process of the premaxilla only, with no lateral overlapping of the premaxilla by the anterolateral maxillary process ( Fig. 6 ). It suggests that the anterolateral process of the maxilla was not well-developed in Ajkaceratops . It contrasts with the condition present in most ceratopsians [e.g. in Archaeoceratops oshimai (IVPP V11114), Auroraceratops (GSGM-FV-00505), Bagaceratops , Prenoceratops , Protoceratops andrewsi , Udanoceratops (PIN 3907/11), Yamaceratops (MPC-D 100/1315), and ceratopsoids]. In Yinlong (IVPPV8637),andintheearly-divergingneoceratopsian Liaoceratops , the lateral overlapping of the premaxilla seems to be only slightly visible. Only in Psittacosaurus spp. does the posteroventral contact between the premaxilla and maxilla seem to be straight in ventral view. Predentary Four isolated predentary bones of various sizes were referred by Ősi et al. (2010) to Ajkaceratops kozmai . Only the largest specimen, MTM V2009.195.1, preserves the anterodorsal tip of the bone ( Fig. 8 ). The anterior end of the predentary is sharply pointed in dorsal view, and it curves anterodorsally in lateral view, as in all ceratopsians (with the exception of Psittacosaurus spp. ), a number of basal neornithischians (e.g. Changchunsaurus parvus , Haya griva , Jeholosaurus shangyuanensis , Talenkauen santacrucensis , and Thescelosaurus neglectus ; Barrett and Han 2009 , Butler et al. 2011 , Boyd 2014 , Rozadilla et al. 2019 , Barta and Norell 2021 ), and in Jakapil kaniukura ( Riguetti et al. 2022 ) . Similarly to the proposed rostral bone of the type specimen (MTM V2009.192.1) and other referred predentaries (MTM V2009.193.1; Fig. 9 ), the anterior surface is unkeeled, and the tip of the predentary is subtriangular in transverse cross-section. The pattern of ornamentation accurately matches that of the holotype , such as the tip being deprived of any rugosities, with the two ventralmost pits placed in the longitudinal grooves some distance from the anterodorsal end of the bone. On the dorsal surface of the predentary MTM V2009.195.1, the pits are less common and there are three longitudinal grooves visible, running from just behind the anterior tip of the bone. The smaller specimen referred to Ajkaceratops , MTM V 2009.194.1, exhibits distinct morphology, with relatively densely distributed, large, and deep pits ( Fig. 10A–D ). There is a quite deep medial longitudinal groove on the dorsal surface of MTM V2009.194.1, and in the smallest specimen, MTM V2009.196.1 ( Fig. 10E–H ). The posterolateral processes are not well-developed in these two specimens . As noted by Ősi et al. (2010) , the predentaries MTM V2009.195.1 and MTM V2009.193.1, and partly also MTM V2009.196.1, have relatively sharp lateral edges posteriorly, which contrasts with the bevelled edges in the earliest-diverging ceratopsians ( Makovicky and Norell 2006 ). In MTM V2009.194.1, the oral margins of the predentary are comparatively smoother. The posterolateral process: In all specimens referred to Ajkaceratops , the posterolateral process of the predentary is relatively short, as in the majority of early-diverging ornithischians and non-ceratopsid ceratopsians, but unlike the condition observed in Zuniceratops and ceratopsids, in which the posterolateral process is more elongated. There is no visible slit into the posterolateral process that would enclose the dentary, as in euceratopsian dinosaurs (leptoceratopsids, protoceratopsids, and ceratopsoids), which is described as the morphotype III of the predentary–dentary articulation by Nabavizadeh and Weishampel (2016) . A clear articulation surface is visible on the right posterolateral process of MTM V2009.194.1, matching the Morphotype I by Nabavizadeh and Weishampel (2016) . However, in specimen MTM V2009.193.1, the articular surface for the dentaries on the posterior surface of the predentary seems to be relatively deeply excavated ( Fig. 9 ). The posteroventral process: A posteroventral process is completely preserved in MTM V2009.194.1 and partially in MTM V2009.193.1. Originally, it was also present in MTM V2009.196.1, as illustrated by Ősi et al. (2010 :SI), but it is no longer preserved in that specimen ( Fig. 10E–H ). In Ajkaceratops , the process is relatively wide and much longer than the posterolateral one, resembling that observable in several groups of ornithischians ( Changchunsaurus , Jeholosaurus , Lesothosaurus , and ceratopsians; see: Barrett and Han 2009 , Jin et al. 2010 , Porro et al. 2015 , Nabavizadeh and Weishampel 2016 ). It shows a very low interdental process reaching the posterior tip of the posteroventral process. The interdental process is much thinner in Ajkaceratops (MTM V2009.195.1) than in neoceratopsians [e.g. Archaeoceratops , Auroraceratops , and Yamaceratops (MPC-D 100/1867); Makovicky and Norell 2006 , Morschhauser et al. 2018a ]. It would mean that the dentaries were very close to each other anteriorly, and perhaps articulated along the symphysis. As observed byMakovicky (2012), the posteroventral process of the predentary does not bifurcate posteriorly. Its most posterior point is placed along its longitudinal axis. It differs from the condition observable in all ceratopsian dinosaurs, including chaoyangsaurids ( Hualianceratops and Yinlong ; see: Han et al. 2015 , 2016 ), some specimens of Psittacosaurus spp. (e.g. IVPP V12617; Tanoue et al. 2010 ), early-diverging neoceratopsians Archaeoceratops , Auroraceratops , Liaoceratops , euceratopsians Leptoceratops , Protoceratops , and ceratopsoids, in which the posteroventral process of the predentary bifurcates posteriorly to articulate with the posteriorly diverging dentaries ( Tanoue et al. 2010 ). The posterior bifurcation of the predentary is also present in the majority of early-diverging neornithischians, such as Changchunsaurus , Dryosaurus , Haya , Thescelosaurus , and Zalmoxes ( Weishampel et al. 2003 , Jin et al. 2010 , Makovicky et al. 2011 , Boyd 2014 , Nabavizadeh and Weishampel 2016 , Barta and Norell 2021 ). It does not seem to be an ontogenetically dependent feature because the posterior bifurcation is present even in near-embryonic specimens of protoceratopsids (MPC-D 100/1021). Dentary One of the specimens referred to Ajkaceratops kozmai (MTM V2009.195.1) preserves a fused, 17 mm-long anteriormost portion of the right dentary bone ( Fig. 8 ). The merging of the predentary with the dentary is rarely reported for ornithischians, although the bones fuse late in ontogeny in Psittacosurus ( Sereno 2010 ). In ceratopsians, the two bones are in firm contact with each other, restricting the mobility at this junction ( Bell et al. 2009 , Nabavizadeh and Weishampel 2016 ). The left side of the specimen is preserved up to the region where the predentary contacts the dentary, suggesting that the fusion in this individual might not have been fully developed. Posterior to the fusion with the predentary, there is a very shallow fossa, bordered by a swollen posterior margin of the predentary. It is similar to the structure observable in the premaxilla posterior to the assumed contact with the proposed rostral bone in the type specimen. It is most likely caused by the transverse development of the fused contact with the predentary. Several small foramina on the lateral surface of the dentary are present, most likely being a neurovascular foraminasupplying a keratinous covering of the predentary (see: Nabavizadeh and Weishampel 2016 ). There is a small, rugose tubercle on the anterodorsal corner of the dentary, just posterior to the contact with the predentary. The tubercle is directed dorsolaterally, and although it is difficult to trace the contact between the two bones, it seems that it does not represent a mere extension of the posterolateral predentary process. Similar tubercles are present in some euceratopsians [ Leptoceratops (CMN 8889) and Protoceratops (AMNH FARB 6429)]; in those, however, the tubercle contacts with the posterolateral process of the predentary. Additionally, at the base of the posterolateral process of Leptoceratops , there is a notch that accommodates a short anterior process of the dentary. In Ajkaceratops , the contact between the predentary and the dentary is much more simple and the tubercle is placed significantly more dorsally to the dorsal margin of the posterolateral process of the predentary. In the dorsal view, the dentary develops posteriorly into a rather mediolaterally thin structure. The preserved dorsal margin of the dentary is edentulous. No traces of the alveoli are preserved, indicating that the first dentary tooth might have been separated by a relatively long diastema (at least as long as a third of the predentary). The presence of the anterior dentary diastema was described in later-diverging ceratopsians and iguanodontians ( Bell et al. 2009 , Kubota and Kobayashi 2009 , Nabavizadeh and Weishampel 2016 ). Among ceratopsians, the teeth are present almost immediately behind the predentary–dentary contact in Archaeoceratops oshimai (IVPP V1114) , Auroraceratops ( Morschhauser et al. 2018a ) , and in some early-diverging neoceratopsians ( Makovicky and Norell 2006 ). Phylogeny reconstruction Owing to the difficulties in interpreting the osteology of Ajkaceratops kozmai and the substantial differences between the ornithischian tree topologies reconstructed in recently published studies (e.g. Han et al. 2018 , Madzia et al. 2018 , Herne et al. 2019 , Dieudonné et al. 2021 , Černý et al. 2022 ), the inference of the phylogenetic placement of Ajkaceratops represents a particularly challenging task. Therefore, rather than assessing the affinities of Ajkaceratops using a single ‘preferred’ dataset and following a certain interpretation of its anatomical peculiarities, we explore the effect of applying differing approaches to scoring the Ajkaceratops ’ operational taxonomic unit (OTU) and of the use of multiple datasets. We have selected three ornithischian-wide datasets that represent modified versions of those first published by Han et al. (2018) , Herne et al. (2019) , and Dieudonné et al. (2021) ; hereafter referred to as ‘HA18’, ‘HE19’, and ‘D21’, respectively. Each of these datasets was originally constructed to tackle different research tasks. Considering the current uncertainties associated with the material of Ajkaceratops , it is impossible to treat any of them to be more appropriate for exploring the placement of the taxon or to provide more reliable results in general. Indeed, a thorough phylogenetic assessment of Ajkaceratops , and also of other taxa whose phylogenetic affinities are obscured, may actually require that a new character–taxon matrix is drafted that would include a more ‘complete’ sampling of characters present in various ornithischian lineages. In order to test the impact of different interpretations of the Ajkaceratops osteology on its phylogenetic placement, we have scored five separate OTUs for each of the selected datasets: (1) the holotype and referred material;with the anterior section interpreted as the rostral bone (scored in the matrices as ‘ Ajkaceratops _whole_Rostral’; hereafter, ‘OTU 1’); (2) the holotype only; with the anterior section interpreted as the rostral bone (‘ Ajkaceratops _Rostral’; ‘OTU 2’); (3) the holotype and referred material; with the anterior section interpreted as the premaxillary bone (‘ Ajkaceratops _whole_PMX’, ‘OTU 3’); (4) the holotype only; with the anterior section interpreted as the premaxillary bone (‘ Ajkaceratops _PMX’; ‘OTU 4’); (5) the referred predentaries only (‘ Ajkaceratops _ Predentary’; ‘OTU 5’). Particular Ajkaceratops OTUs are provided in the Supporting Information, S1 and need to be added to the matrices separately. Results For full tree topologies and extended numerical results, see the Supporting Information, S1. As could be expected, OTUs 1 and 2 were reconstructed as part of Ceratopsia using all three datasets. The datasets HE19 and D21 do not comprise an adequate sample of ceratopsians and their characters to provide a reliable reconstruction of their interrelationships. However, the placement of Ajkaceratops within the clade is well-supported in both these runs (see Fig. 13 ). In HA18, the dataset that was originally constructed to focus especially on early-diverging ceratopsians and that comprises the highest number of ceratopsian taxa from among the ornithischian-wide datasets used in the present study, OTUs 1 and 2 are nested within Euceratopsia, as the sister-taxon to Zuniceratops . When the anterior section of the snout was interpreted as the premaxilla (OTUs 3 and 4), the results differed depending on whether the OTU also comprised data from the predentaries (OTU 3) or not (OTU 4). When analysed together (OTU 3), HA18 and D21 again inferred Ajkaceratops as a ceratopsian, though in D21 the support for such placement was poorer ( Fig. 13 ). The analysis of HE19 inferred Ajkaceratops in a basal polytomy with other marginocephalians. Perhaps interestingly, using HA18, Ajkaceratops was inferred as the sister-taxon to Zuniceratops , even when the anterior section of the snout was interpreted as the premaxilla and was analysed without the predentaries (OTU 4). In contrast, HE19 and D21 could not place OTU 4 beyond Ornithischia and Neornithischia , respectively ( Fig. 14 ). When assessing the majority rule consensus trees (that, however, should be treated with considerably greater caution), HE19 tends to place OTU 4 at an early-diverging position among ornithopods, while D21 infers it at the basal polytomy of Neornithischia (Supplementary Material 1: Figs S18 and S28, respectively). Finally, when assessing the referred predentaries only (OTU 5), the analyses place the material either in a large polytomy with other ornithischians (HA18), with other ceratopsians (D21) or possibly within Neoceratopsia (HE19) ( Fig. 15 ).