Catopsbaatar catopsaloides ( Kielan−Jaworowska, 1974 )

Kielan-Jaworowska, Zofia, Hurum, Jørn H. & Lopatin, Alexey V., 2005, Skull structure in Catopsbaatar and the zygomatic ridges in multituberculate mammals, Acta Palaeontologica Polonica 50 (3), pp. 487-512 : 490-506

Catopsbaatar catopsaloides ( Kielan−Jaworowska, 1974)

Figs. 1–10, 11C .

Djadochtatherium catopsaloides sp. nov.; Kielan−Jaworowska (1974: 40; text−fig. 6; pl. V: 9; pl. XVII: 2; pls. XVIII–XXI).

Catopsalis catopsaloides (Kielan−Jaworowska) ; Kielan−Jaworowska and Sloan (1979: 188, fig. 2B).

Catopsalis catopsaloides (Kielan−Jaworowska) ; Kielan−Jaworowska et al. (1986: pl. 1: 2, pl. 2: 3).

Catopsbaatar catopsaloides (Kielan−Jaworowska) ; Kielan−Jaworowska (1994: 134).

Catopsbaatar catopsaloides (Kielan−Jaworowska) ; Gambaryan and Kielan−Jaworowska (1995: 62, figs. 2A, 7B, 17B).

Catopsbaatar catopsaloides (Kielan−Jaworowska) ; Kielan−Jaworowska and Hurum (1997: figs. 10I, 11I, and 12K).

Catopsbaatar catopsaloides (Kielan−Jaworowska) ; Kielan−Jaworowska et al. (2000: 592, fig. 29.11).

Catopsbaatar catopsaloides (Kielan−Jaworowska) ; Kielan−Jaworowska et al. (2002: 557, figs. 1, 2).

Catopsbaatar catopsaloides (Kielan−Jaworowska) ; Kielan−Jaworowska et al. (2004: 324, figs. 8.9, 8.38I, 8.39F, 8.40K).

Holotype: ZPAL MgM−I/78. Nearly complete skull, associated with both dentaries of a juvenile individual, figured by Kielan−Jaworowska (1974: text−fig. 6, pl. V: 9, pl. XVII: 2, pls. XVIII–XX), by Kielan−Jaworowska et al. (1986: fig. 2), re−figured by Kielan−Jaworowska et al. (2000: fig. 29.11); we figure herein ( Fig. 10A ) the SEM micrographs of the dentition of the holotype specimen. Originally the right zygomatic arch was broken and has been glued back. A comparison with PIN 4537/5 and PM 120/107, in which both zygomatic arches have been preserved, shows that apparently it has been glued too far laterally, with the glenoid fossa retaining an antero−posterior position, rather than anterolateral–postero−medial, as characteristic of PIN 4537/5 and PM

120/107. Type horizon and locality: Red beds of Hermiin Tsav (stratigraphic equivalent of the?late Campanian Baruungoyot Formation), locality Hermiin Tsav I, south−western part of the Gobi Desert, Mongolia .

Material.—In addition to the holotype specimen, there are:

ZPAL MgM−I/79, Hermiin Tsav I, incomplete skull of an adult individual, figured by Kielan−Jaworowska (1974: pl. XXI).

ZPAL MgM−I/80, Hermiin Tsav II, incomplete damaged skull, associated with anterior parts of damaged dentaries; the skull is figured herein ( Fig. 5 ), and the left dentary (Fig. 9A).

ZPAL MgM−I/159, Khulsan, Nemegt Valley, right, slightly worn m2, figured herein ( Fig. 10B ).

PIN 4537/4, Hermiin Tsav (no further data), skull without zygomatic arches and with damaged basicranial region of a juvenile individual, associated with both fragmentary dentaries, figured herein ( Figs. 1A and 4 ).

PIN 4537/5, Hermiin Tsav (no further data), nearly complete skull of a juvenile individual, associated with both dentaries, figured by Gambaryan and Kielan−Jaworowska (1995: figs. 2A, 7B, and 17B), and herein ( Figs. 1B and 3).

PM 120/107, Hermiin Tsav I, almost complete, but somewhat depressed skull of an adult individual, with damaged basicranial region and partly damaged occipital region, associated with both dentaries and a large part of the postcranial skeleton, figured by Kielan−Jaworowska et al. (2002: figs. 1, 2) and herein ( Figs. 1C, 2 , and 9B).

Diagnosis and distribution.—As for the genus.

Description

We describe the skull not by regions, as often done (e.g., Simpson 1937; Kielan−Jaworowska 1971), but bone by bone (as e.g., Kielan−Jaworowska et al. 1986; Miao 1988; Wible and Rougier 2000). We follow the description sequence of Wible and Rougier (2000), in order to facilitate a comparison with Kryptobaatar , which is the best known djadochtatheriid.

Skull as a whole.—The skull is massive, with a wide transverse anterior margin; it is shorter in the middle than laterally, due to the middle incurvature of the nuchal crest, best seen in PM 120/107 (see reconstructions in Figs. 6–8). In PIN 4537/5 ( Figs. 1B and 3), the posterior margin of the skull appears rounded, which is due to its incompleteness, the lateral parts of the nuchal crests being missing; however, the middle part apparently corresponds to the real skull length. The zygomatic arches are strongly expanded laterally—the skull width across the zygomatic arches in PIN 4537/5 is about 85 percent of its length (measured in dorsal view),

associated with both dentaries of adult individual in left lateral (C 1), right lateral (C 2), occipital (C 3), ventral (C 4), and anterior (C 5) views; isolated skull of the same in dorsal view (C 6); left dentary of the same in medial (C 7) and lateral (C 8) views. All in the same scale. Note the tripartite infraorbital foramen (right side of PIN 4537/4, A 3); double (right side of PIN 4537/5, B 1, B 2); and single (both sides of PM 120/107, C 1, C 2, C 5). The structures seen in C 3 are explained. For explanation of structures in other figures, see Figs. 2–4 , 9, and reconstructions in Figs. 6–8.

while in PM 120/107 it is about 90 percent of the length along the middle line ( Table 1). It cannot be excluded that the unusually high width/length ratio of PM 120/107 has in part been caused by the slight dorso−ventral flattening of the specimen. In dorsal view the anterior parts of the orbits are situated more posteriorly than in other djadochtatheriids, which results in a very long snout and a small orbit. The occipital plate, as fragmentarily preserved in PM 120/107, was arranged almost vertically, but it appears slightly concave and it is not seen in dorsal view of the skull (being obscured by the prominent nuchal crest). Although the most complete skull of Catopsbaatar is depressed and the occipital plate has been badly damaged, we assume ( Figs. 7 and 11C ) that the position of the occipital plate with respect to the rest of the skull was apparently as in e.g., Kamptobaatar , Sloanbaatar , and Nemegtbaatar , in which the occipital plate slopes

KIELAN−JAWOROWSKA ET AL.—SKULL STRUCTURE IN CATOPSBAATAR 493

double infraorbital foramen nasal foramina 10 mm depression postorbital process double infraorbital foramen break break I2 premaxillary ridge? groove for ethmoidal artery anterior lamina

postero−dorsally from the condyles (see e.g., Kielan−Jaworowska 1970, 1971, and Kielan−Jaworowska et al. 1986).

Premaxilla.—As usual, the premaxilla consists of facial and palatal parts, at the boundary of which in Djadochtatherioidea there is a distinct ridge designated herein the premaxillary ridge (see e.g., Kielan−Jaworowska 1970: pl. 1: 2b, and Kielan−Jaworowska 1974: pl. XXI: 1b, and Figs. 2C , 3C–E, 4E , and 8 herein). In palatal view the premaxillary ridge is not situated at the margin of the skull, but extends longitudinally and leaves a triangular ventral portion of the facial part of the premaxilla lateral to it. In ZPAL MgM−I/79 the premaxillary ridges on both sides are especially prominent, and have a narrow, flat ventral surface, covered with longitudinally arranged rows of minute foramina ( Kielan−Jaworowska 1974: pl. XXI: 1b).

Wible and Rougier (2000: 18) stated that in K. dashzevegi : “[...] extending posteriorly from the anterior part of the premaxillary−maxillary suture is a small horizontal furrow, perhaps representing an attachment for facial musculature (‘muf' in fig. 14). The most probable occupant of such a furrow was the musculus incisivus superioris, which raised the upper lip [...]”. Examination of fig. 14 in their paper shows that the furrow in question was arranged longitudinally and was situated anterior to the premaxillary−maxillary suture, rather than posterior as stated by the authors. In most Catopsbaatar specimens this region is damaged; however, on the left side of ZPAL MgM−I/79 ( Kielan−Jaworowska 1974: pl. XXI: 1b), on the right side of PIN 4537/4 ( Fig. 4E ), and right side of PIN 4537/5 ( Fig. 3E) in front of the premaxillary−maxillary suture, there is a distinct longitudinally directed facet, which may correspond to muscle attachment, similar to that recognized by Wible and Rougier (2000) for Kryptobaatar . This facet is situated opposite the anterior part of the premaxillary ridge.

The premaxillary−maxillary suture in lateral view extends in the dorsal part roughly horizontally, along the narrow posterodorsal process, which increases in width anteriorly, then the suture turns in a bow downwards, extending almost vertically, and then near the ventral margin of the snout it turns posteriorly again ( Figs. 1B 1, C 1 –C 2, 2B ; 3C, D, 4C, D ). The palatal part of the premaxillary−maxillary suture is perhaps best preserved in PIN 4537/4 ( Fig. 4E ); it extends a short distance lateral to the premaxillary ridge, then turns, crosses the ridge and extends postero−medially. It turns around I3 (or the alveolus for it), some distance to the rear of it, reaches the incisive foramen, extends transversely along its posterior boundary and reaches its counterpart from the other side The palatal part of the premaxilla is concave, but the distinct thickenings, characteristic of Kryptobaatar ( Wible and Rougier 2000) and Tombaatar ( Rougier et al. 1997) are not recognizable. There are depressions on the palatal part, situated between I2 and I3; in PIN 4537/4 on the left side the depression houses a large vascular foramen. In ZPAL MgM−I/79 ( Kielan−Jaworowska 1974: pl. XX: 1b) and PIN 4537/5 ( Fig. 3E), the palatal part is pierced by numerous nutrient foramina distributed at random; these are present but less distinct in other specimens (see our reconstruction in Fig. 8 ). The premaxilla is relatively smaller than in Kryptobaatar , extending for less than one third of the preorbital part of the snout ( Figs. 5 and 7 ).

Nasal.—The nasal in Catopsbaatar is a relatively wide bone, moderately widening posteriorly in the posterior half of its length. Along the lateral border the nasal contacts the premaxilla and maxilla, and posteriorly the lacrimal and frontal. The naso−frontal suture is not pointed anteriorly in the middle, as tentatively reconstructed by Kielan−Jaworowska and Sloan (1979: fig. 1A), Kielan−Jaworowska and Hurum (1997: fig. 10I), and Kielan−Jaworowska et al. (2004: fig. 8.39F), but rather rounded. Reconstruction of the postero−lateral margin of the nasals, as given in Fig. 6, is only tentative, due to the uncertainty of the size and shape of the lacrimal (see below). On the anterior two−thirds of the nasals’ length in PM 120/107, there are four irregularly arranged vascular nasal foramina on the right side and apparently only three on the left side ( Fig. 6), the second from anterior being larger than the others on both sides. In other specimens there may be only two pairs of vascular foramina (see section “Variation” below for discussion on distribution of nasal foramina in various specimens).

Lacrimal.—Reconstruction of the size and shape of the facial wing of the lacrimal bone in C. catopsaloides poses difficulties, as the relevant region in three skulls ZPAL MgMI/79, /80 and PM 120/107 is missing or devoid of the bones, while in the three remaining skulls ZPAL MgM−I/78, PIN 4537/4, and PIN 4537/5 it has been preserved, but damaged. Kielan−Jaworowska and Sloan (1979: fig. 1) reconstructed the facial wing of the lacrimal bone in Kryptobaatar dashzevegi (1979: fig. 1A) as a very large, rectangular bone, extending from the anterior end of the orbit up to the contact with nasal; this reconstruction was accepted by Kielan−Jaworowska and Hurum (1997: fig. 10I) and Kielan−Jaworowska et al. (2004: fig. 8.39F). Wible and Rougier (2000: fig. 32) reconstructed a smaller facial wing in K. dashzevegi from Ukhaa Tolgod, where it invades the maxilla laterally, which we regard as correct. In Djadochtatherium ( Webster 1996: photo on p. 84), however, which is very close to Catopsbaatar , the facial wing of the lacrimal is relatively larger, and apparently sub−rectangular. In Catopsbaatar , in ZPAL MgM−I/78 on the right side, the bone is missing around the orbit, but on the matrix preserved there is an imprint of a presumed lacrimal, which invades the maxilla and has a rounded anterior margin. Similar structure has been preserved on the right side of PIN 4537/4 and the right side of PIN 4537/5. On the basis of these three specimens, and a comparison with Kryptobaatar and Djadochtatherium we tentatively reconstruct the facial wing of the lacrimal as a relatively elongated bone, with a rounded anterior margin ( Fig. 6). The orbital wing of the lacrimal in all the skulls available is not sufficiently well preserved to allow a reconstruction of its shape and size, although in PIN 4537/4, on the right side of the skull a very large lacrimal foramen appears to be present. In Fig. 11C we reconstruct tentatively the orbital wing of the lacrimal, based on a comparison with Kryptobaatar ( Wible and Rougier 2000) .

Frontal.—The frontal is a large bone consisting of a horizontal component, building the large part of the cranial roof, and a vertical component, contributing to the medial wall of the orbit. The horizontal part (e.g., Fig. 2A , and reconstruction in Fig. 6) is very wide, roughly trapezoidal in shape; the left and right parts of the frontals meet one another along the midline in an irregular suture. Anteriorly the frontals contact the nasals, with a suture that is slightly convex anteriorly; laterally they contact the lacrimals and form the margin of part of the orbit, and posteriorly the parietals. The frontal−parietal suture is U−shaped as in almost all djadochtatherioid genera (see Kielan−Jaworowska and Hurum 1997: fig. 10, and the description frontal nasal squamosal

premaxilla maxilla postorbital process parietal ridge

vascular foramina lacrimal parietal anterior lamina

20 mm nuchal crest below). The suture in Catopsbaatar apparently resembles most that in Kryptobaatar ( Wible and Rougier 2000: fig. 32), but forms a relatively narrower and shallower arc. This arc is somewhat more convex in the holotype ZPAL MgM−I/78 ( Kielan−Jaworowska 1974: pl. XVIII: 1c) than in the others, but even in the holotype it is less deep and narrower than in Kryptobaatar . Similar to Kryptobaatar , the lateral part of the suture on each side forms a small bow convex posteriorly. These bows in Catopsbaatar are relatively larger, deeper, and wider transversely than in Kryptobaatar . The suture is different from that in Djadochtatherium , where in the specimen from Tögrög provisionally numbered HMNS 94−10−278 (see Acknowledgements for details, personal communication from Mahito Watabe to ZK−J, October, 1996), the middle part is V−shaped rather than U−shaped, and the lateral bows on the suture are wider transversely in Djadochtatherium than in Catopsbaatar . The supraorbital notch, which has been described in Kryptobaatar by Wible and Rougier (2000) at the boundary between the vertical and horizontal components of the parietal, apparently was also present in Catopsbaatar , as may be inferred from examination of the relevant region in ZPAL MgM−I/78, PM 120/107, and the left side of PIN 4537/4 ( Fig. 7 ). The structure of the orbit partly preserved in PIN 4537/4 and PIN 4537/5 suggests that the vertical component of the frontal is relatively longer in Catopsbaatar than Kryptobaatar . In PIN 4537/4 on the right side the foramen for frontal diploic vein (see Wible and Rougier 2000: fig. 36A) is preserved in the upper part of the frontal ( Fig. 4D ). The suture between the vertical portions of the maxilla and frontal as well as the sphenopalatine foramen are visible on left side of PIN 4537/4 ( Fig. 4C ). In PIN 4537/5, the deep, vertically directed groove in the anterior part of the orbit ( Fig. 3D) might be for ethmoidal artery.

Maxilla.—The maxilla ( Figs. 6, 7 , 8 , 11 ) is a very extensive bone; its facial process (lateral view) comprises most of the lateral wall of the snout and is seen at the lateral parts of the snout in dorsal view; the palatal process contributes to a large part of the palate (ventral view); and the alveolar process, lateral to it, bears all the upper teeth except for the incisors. The zygomatic process of the maxilla forms the anterior part of the zygomatic arch, up to the suture with the squamosal, situated at the boundary between the anterior and the intermediate zygomatic ridges. Finally, the orbital process, poorly preserved in all available specimens, constitutes part of the vertical wall, forming the medial wall of the orbit. The facial process contacts the premaxilla anteriorly (with the suture described above) and the nasal and lacrimal dorsally. The infraorbital foramen as preserved in ZPAL MgM−I/78 ( Kielan−Jaworowska 1974: pl. XVIII: 1a, d), and in PM 120/107 ( Figs. 1C 1, C 2 , C 5 , and 2B) is single, relatively small, slit−like (dorso−ventrally compressed), and visible in anterior, ventral and lateral views; its posterior margin is situated opposite P 3 in ZPAL MgM−I/78, but more anteriorly (opposite of the missing P1, or even in front of it) in PM 120/107. It is not preserved in ZPAL MgM−I/79 and /80. In PIN 4537/5, the infraorbital foramen, preserved on the right side of the skull, is double ( Figs. 1B 1, B 2 , 3A, D), while in PIN 4537/4 ( Figs. 1A 3, 4A, D ), preserved on the right side, it is tripartite, the anteriormost foramen being the largest, the others decreasing in size posteriorly. In contrast to ZPAL Mg−I/78 and PM 120/107, where the infraorbital foramina are slit−like, in two PIN specimens the double or tripartite foramina are rounded.

The most characteristic feature of the facial and zygomatic processes of the maxilla in Catopsbaatar is the enormous anterior zygomatic ridge ( Gambaryan and Kielan−Jaworowska 1995; and Fig. 7 herein), which is relatively much higher than the corresponding ridges in other djadochtatherioidean taxa, except for Djadochtatherium ( Fig. 11 ). In Catopsbaatar the anterior zygomatic ridge encircles almost half a circle, rather than half of an ellipse as in most genera, e.g., Nemegtbaatar , Chulsanbaatar (see Gambaryan and Kielan−Jaworowska 1995: fig. 2B, C) and Kryptobaatar ( Fig. 11A ). The anterior margin of the anterior zygomatic ridge in Catopsbaatar is strongly thickened, producing a bulge on the lateral margin of the snout seen in both ventral and dorsal views (as characteristic also of Kryptobaatar , see e.g., PSS−MAE 113, figured by Wible and Rougier 2000: figs. 4—two upper photographs, and 15). See description of the intermediate zygomatic ridge and the maxilla−squamosal suture under “Squamosal” below. The lower part of the maxilla−squamosal suture, as examined in lateral view, extends along the posterior boundary of the anterior zygomatic ridge and has a sigmoid shape; however, the uppermost part of the suture is poorly recognized (see section “Comparisons” and Fig. 11C below).

Gambaryan and Kielan−Jaworowska (1995: fig. 7B 1) reconstructed the masseter superficialis pars anterior as originating from the anterior zygomatic ridge. The upper margin of the zygomatic arch forms a bow around the lower margin of the orbit. The anterior part of the zygomatic arch (built by the maxilla) is high, but relatively thin latero−medially, and because of this (e.g., in PM 120/107) it is undulating, deformed by preservation.

The palatal processes of the maxilla are extensive, forming most of the palate. Right and left components form together a concave surface. The transverse suture with the horizontal part of the palatine bone is best preserved in ZPAL MgM−I/80 ( Fig. 5A ), where it extends opposite the anterior half of M1. The major palatine foramina are well−preserved; shallow grooves extend from them anteriorly, along the palatal part of the maxilla ( Figs. 5 and 8 ). In ZPAL MgM−I/78 the horizontal part of the palatine bone has not been preserved, but still the suture shows the presence of the major palatine foramina ( Kielan−Jaworowska 1974: pl. XVIII: 1d). In PM 120/107 on the palatine process of the maxilla there are two relatively large foramina created by damage. In PIN 4537/4 the right major palatine foramen is preserved, situated opposite the anterior part of M1 ( Fig. 4E ). The palatine has not been preserved, but the posterior margins of both palatal processes of the maxilla have the shape of an overturned letter U, and apparently correspond to the maxillary−palatine suture. The suture between the palatal part of the maxilla and the alisphenoid is seen on the left side of PIN 4537/4 ( Fig. 4E ); it is placed relatively more close to M2 than in Kryptobaatar ( Wible and Rougier 2000: fig. 34), see also the reconstruction in Fig. 8 .

Palatine.—The palatine bone in multituberculates, as in other mammals, consists of two parts: the horizontal lamina and the perpendicular lamina ( Evans 1995). The horizontal lamina in multituberculates forms a roughly rectangular plate surrounded as usually from anterior and lateral sides by the maxilla, its posterior border (often with a thickened postpalatine torus) forming the anterior margin of the choanae. Among the Late Cretaceous Asian multituberculates the horizontal lamina is perhaps best preserved in the juvenile skull of Kamptobaatar kuczynskii ZPAL MgM−I/33 ( Kielan−Jaworowska 1971: fig. 4 and pl. I: 2b) and a juvenile specimen of Chulsanbaatar vulgaris, ZPAL MgM−I/168 ( Kielan−Jaworowska et al. 1986: fig. 14A). In Kamptobaatar there is a pair of major palatine foramina situated slightly in front of the maxilla−palatine suture, and two pairs of minor palatine foramina within the horizontal lamina (accessory palatine foramina of Wible and Rougier 2000); there is a moderately prominent postpalatine torus and the palatonasal notches at the lateral sides of it. In Chulsanbaatar the horizontal lamina is similarly built as in Kamptobaatar , the difference concerning presence of only one, rather than two pairs of minor palatine foramina.

In Catopsbaatar the horizontal lamina has not been preserved in ZPAL MgM−I/78, PIN 4537/4, and PIN 4537/5. It is almost completely preserved in ZPAL MgM−I/80 ( Fig. 5A ), in which the postpalatine torus is present but very weak. In ZPAL MgM−I/79 ( Kielan−Jaworowska 1974: pl. XXI, 1b) a small part of the horizontal lamina has been preserved on the right side of the skull, including an indistinct postpalatine notch and a tiny fragment of the postpalatine torus (see section “Terminology” above). In PM 120/107 ( Fig. 2C ) the horizontal part of the palatine bone has been preserved, but the suture between it and the palatine process of the maxilla is poorly seen, situated opposite the second and the third cusps of the middle row of M1, as in ZPAL MgM−I/80 (see also Fig. 8 ).

There is controversy concerning the exposure of the perpendicular lamina of the palatine bone in the orbit in multituberculates. Kielan−Jaworowska (1971: fig. 3) recognized the exposure of the palatine within the ventral part of the orbit of Kamptobaatar . Miao (1988) demonstrated that the palatine is wholly excluded from the orbit in Lambdopsalis bulla , while Wible and Rougier (2000) demonstrated the same for Kryptobaatar dashzevegi . Miao (1988) argued that the lack of the palatine exposure in the orbit should be regarded as the autapomorphy of Multituberculata . On the other hand, Hurum (1994, 1998a) demonstrated its presence in the sectioned skull of Nemegtbaatar (see Kielan−Jaworowska et al. 1986), while Sloan (1979: fig. 1) reconstructed it in the orbit of Ectypodus .

In Catopsbaatar , as far as can be judged from the poorly preserved orbital walls (best seen in ZPAL MgM−I/78), the palatine bone was apparently not exposed in the orbit.

Wible and Rougier (2000: 79) concluded: “If Hurum’s (1994, 1998a) interpretations of the orbital sutures in Nemegtbaatar are correct, then the Mongolian Late Cretaceous taxa would be polymorphic regarding the presence of the palatine in the orbit.”

Pterygoid and vomer.— Kielan−Jaworowska (1971) argued that multituberculates differ from all other mammals in the position of the pterygoids, which do not form parts of the lateral walls of the choanae, but are situated between the vomer and the lateral walls of the choanae, the latter apparently built by the alisphenoids. The sutures between the pterygoids and other bones, however, have not been recognized. Since that time similar structure of the choanal region has been recognized in numerous multituberculate taxa, especially in representatives of the exquisitely preserved Late Cretaceous djadochtatherioideans. Wible and Rougier (2000) described this region in Kryptobaatar in detail. They referred to the two channels on each side of the choanal region as the pterygopalatine troughs; the medial one, situated between the vomer and the pterygopalatine ridge (pterygoid of Kielan−Jaworowska 1971: fig. 4), as the medial pterygopalatine trough; and the lateral one (?scaphoid fossa of Kielan−Jaworowska 1971) as the lateral pterygopalatine trough. The term pterygopalatine ridge was introduced by Barghusen (1986: fig. 1B) for the ridge of the pterygoid bone in Thrinaxodon , and used by him in interpreting the choanal structure in Kamptobaatar (fig. 6B in his paper). Barghusen also identified a notch in the ridge of the pterygoid bone in Kamptobaatar , marked by Kielan−Jaworowska (1971: fig. 4) with a question mark, as an incisure connecting the lateral nasopharyngeal passage with the internal naris. In the skulls of Catopsbaatar housed in ZPAL, PIN, and PM the choanal region has been damaged and does not provide information on the structure of the pterygoids and the vomer in this taxon.

Kielan−Jaworowska (1971) suggested that in multituberculates the pterygoids do not form the lateral walls of the choanae, but are situated in the middle of the choanal channel on each side. In light of Barghusen (1986) and Wible and Rougier (2000) papers it is now possible to speculate that in multituberculates the pterygoids apparently form the lateral walls of the choanae, as in all other mammals, being in contact with alisphenoids, and form the roof of the lateral parts of the choanal channels. They differ from those in other mammals in producing a medial longitudinal crest, designated by Wible and Rougier (2000) the pterygopalatine ridge (pterygoid of Kielan−Jaworowska 1971). Our reconstruction of the pterygoids and the whole choanal region in Catopsbaatar , presented in Fig. 8 , is entirely tentative, based on comparisons with Kamptobaatar ( Kielan−Jaworowska 1971) , Nemegtbaatar ( Kielan−Jaworowska et al. 1986) , and Kryptobaatar ( Wible and Rougier 2000) .

Sphenoid complex.—The details of the sphenoid complex are difficult to recognize in multituberculates because of the early fusion of the presphenoid, basisphenoid, orbitosphenoid, and alisphenoid. In the juvenile skull of Kamptobaatar ZPAL MgM−I/33, Kielan−Jaworowska (1971: fig. 3) recognized a very large, fan−shaped orbitosphenoid, less certainly a small alisphenoid, and the anterior lamina of the petrosal. The extent of the latter bone has been subsequently corrected by Kielan−Jaworowska et al. (1986: figs. 20b, 14B, C) on the basis of better preserved braincases in Nemegtbaatar and Chulsanbaatar , showing a very extensive anterior lamina, apparently characteristic for multituberculates as a whole. The large, fan−shaped orbitosphenoid, small alisphenoid, and extensive anterior lamina of the petrosal with a large antero−dorsal process have subsequently been confirmed in Nemegtbaatar and Chulsanbaatar by Hurum (1994, 1998a), who, on the basis of the sectioned skulls, reconstructed their reliable lateral aspects in both genera ( Hurum 1998a: fig. 14).

In Kryptobaatar Wible and Rougier (2000) reconstructed an orbitosphenoid, relatively smaller than in Kamptobaatar , with the ethmoid foramen situated at the boundary with frontal, rather than within the orbitosphenoid, as in Kamptobaatar ( Kielan−Jaworowska 1971) , and the anterior lamina of the petrosal without an antero−dorsal process, present in Nemegtbaatar and Chulsanbaatar ( Hurum 1998a: fig. 14). We were unable to recognize the extent of these bones in Catopsbaatar .

The mesocranial region has been damaged in all the available skulls of Catopsbaatar , and no traces of the presphenoid and basisphenoid have been preserved. The orbital wall has been fragmentarily, but apparently best preserved in PIN 4537 /4 and PIN 4537 /5, and has been discussed above under the description of the frontal bone. It has been partly preserved also on the left side of the holotype ZPAL MgM−I/78 and on the right side of PM 120 /107, but in both specimens no sutures or foramina are recognized. Our reconstruction of this region, presented in Fig. 8 , is entirely tentative, based on comparisons with the genera discussed above .

Petrosal.—The petrosal is one of the most complicated bones in the mammalian skull, housing the organs of hearing and balance. It has been studied preserved in situ in multituberculate skulls by Simpson (1937), followed by Kielan−Jaworowska (1971, 1974), Miao (1988), Rougier, Wible, and Hopson (1992), Rougier, Wible, and Novacek (1996), Hurum (1998a, b), and Wible and Rougier (2000). Because of its robustness and resistance, the multituberculate petrosal is rather well−known, as it has often been preserved not only in skulls, but also in isolation and described in detail (e.g., Kielan−Jaworowska et al. 1986; Wible and Hopson 1993, 1995; Fox and Meng 1997). The meticulous works by Rougier, Wible, and Hopson (1992) and Wible and Hopson (1993, 1995) resulted not only in the description of the osteology of the basicranial region and the lateral wall of braincase, but also in reconstruction of the cranial vasculature. The multituberculate cranial vasculature was earlier studied by Kielan−Jaworowska et al. (1986) on the basis of sectioned skulls belonging to the genera Chulsanbaatar and Nemegtbaatar , but Wible and Hopson (1995) challenged part of the conclusions of these authors and offered a somewhat different interpretation.

Wible and Rougier (2000) provided a detailed description of the petrosal in Kryptobaatar dashzevegi , accompanied by reconstruction of the cranial vasculature.

The material of Catopsbaatar studied by us is much less complete and not as well preserved as that of K. dashzevegi studied by Wible and Rougier. Fragments of the petrosal preserved in ZPAL MgM−I/78 and /79 have been figured and described by Kielan−Jaworowska et al. (1986). In ZPAL MgM−I/80 and PIN 4537/4 the petrosal region has not been preserved; in PIN 4537/5, the hardly discernible right epitympanic recess, promontorium, and jugular fossa may be tentatively identified. In PM 120/107 the basicranial region is poorly preserved, with the middle part (basisphenoid and basioccipital) almost entirely missing (see reconstruction in Fig. 8 , largely based on a comparison with better preserved basicranial region in other genera).

Kielan−Jaworowska et al. (1986: 534) described the promontorium in ZPAL MgM−I/78 as follows: “In front of the fenestra cochleae the promontorium is roughly rectangular with a swollen, rounded and slightly pointed anterior margin.” The left promontorium in PM 120/107 ( Figs. 2C , 8 ) has a similar structure (ventral view), being associated on the medial side with a fragment of basisphenoid−basioccipital. At the postero−lateral part the promontorium strongly narrows, its medial part being occupied by a jugular fossa, which appears relatively smaller than in other djadochtatheriid genera, while on the lateral part there is a distinct fenestra vestibuli, better seen in a ventro−lateral view of the skull. The fenestra cochleae is not discernible, possibly being confluent with the jugular fossa (as tentatively reconstructed in Fig. 8 ). Lateral to the promontorium a relatively wide horizontal part of the anterior lamina of the petrosal has been preserved (see Kielan−Jaworowska 1971: 12 for discussion concerning the notion of the “anterior lamina of the petrosal” in multituberculates, and the section “Terminology” above). The horizontal part of the anterior lamina is divided by a ridge of the lateral flange (which extends roughly parallel to the promontorium turning laterally in the posterior part) into a narrow medial trough, housing the foramina and the wider lateral part. The foramina situated in the trough (see reconstruction of Nemegtbaatar skull by Kielan−Jaworowska et al. 2004: fig. 8.4B) cannot be recognized in Catopsbaatar . Lateral to the lateral flange, there is a very deep and wide epitympanic recess (which houses the ear ossicles) and anterior to it at the suture with the alisphenoid and at the edge between the horizontal and vertical parts of the anterior lamina, traces of two V 3 foramina are tentatively recognized. The carotid foramina have not been preserved.

The vertical wall of the anterior lamina of the petrosal is almost completely preserved on the left side of PIN 4537/4 ( Fig. 4C ), where the suture with the squamosal is recognizable; the concave anterior margin of the anterior lamina apparently corresponds to its original contact with the orbitosphenoid, as in other multituberculate genera the anterior margin of the anterior lamina is concave (see e.g. Kielan−Jaworowska 1971; Kielan−Jaworowska et al. 1986; Wible and Rougier 2000). The anterior lamina has been also partly preserved on the right side of ZPAL MgM−I/78 and less completely on the left side of the same specimen ( Kielan−Jaworowska 1974: pl. XIX: 1d, 1e); it has been preserved on the left side of ZPAL MgM−I/79 ( Kielan−Jaworowska 1974: pl. XXI: 1d) and partly on the right side of the same specimen. Finally the anterior lamina has been preserved, but strongly cracked on both sides of PM 120/107 ( Fig. 2B ); in this specimen on the left side, below the postorbital process, a foramen has been preserved ( Fig. 1C 1). It is probably the opening of the orbitotemporal canal ( Wible and Rougier 2000: figs. 12 and 36), referred to by Kielan−Jaworowska et al. (1986: fig. 20b) as the postorbital foramen. This foramen in Catopsbaatar is situated more posteriorly than the opening of the orbitotemporal canal, identified by Wible and Rougier in Kryptobaatar . On the right side the sutures with the parietal and squamosal are tentatively recognized ( Fig. 2B , see also Fig. 7 ), but the orbitosphenoid has not been preserved and the contacts with that bone (and alisphenoid ventrally) are not discernible.

The occipital exposure of the petrosal (preserved only in PM 120/107, where it is strongly damaged) is similar to that in Kryptobaatar , the petrosal (mastoid) being pierced by a posttemporal canal of moderate size ( Fig. 1C 3). In Fig. 8 we tentatively reconstruct a relatively large paroccipital process, with a?rounded anterior margin, similarly placed as in other djadochtatherioidean genera.

Jugal.—The jugal in multituberculates was reported for the first time by Hopson et al. (1989), who recognized its presence in Ptilodus , Ectypodus , Nemegtbaatar , and Chulsanbaatar , and suggested (based on scars) its presence also in the Late Jurassic paulchoffatiids. In all cases the jugal is placed on the medial side of the zygomatic arch. The largest is the well−preserved jugal in Ptilodus , which extends between the floor of the orbit and the glenoid. Wible and Rougier (2000) described a well−preserved jugal in Kryptobaatar , which is a very delicate, regularly oval bone, distinctly smaller than in Ptilodus .

The jugal has not been preserved in studied specimens of Catopsbaatar , but in PIN 4537/5 ( Figs. 1B 3, 3B) on the medial side of the right zygomatic arch, and of both sides in PM 120/107 ( Figs. 1C 6, 2A ), there is a depression on the squamosal, in the posterior part of zygomatic arch, which possibly could have accommodated the jugal. This depression is, however, situated more posteriorly than in all other multituberculate genera in which the jugal has been preserved.

Squamosal.— Wible and Rougier (2000) divided the squamosal bone into two parts for descriptive purposes: the zygomatic process and the dorsal flange. In Catopsbaatar the dorsal flange is similarly built as in Kryptobaatar . The zygomatic part differs from that in Kryptobaatar in having the zygomatic arch relatively higher and the zygomatic ridges differently arranged. These are best preserved in PIN 4537/5 (see Gambaryan and Kielan−Jaworowska 1995: fig. 2A, and our Figs. 1B 1, 3C, D, 11C ), but also well seen in PM 120/107 ( Figs. 1C 1, C 2, 2B ). The intermediate zygomatic ridge, which contacts the anterior zygomatic ridge anteriorly, is well−preserved in PIN 4537/5 and PM 120/107 ( Figs. 1B 1, 3C, D and 1C 1, C 2, 2B respectively). It is much smaller and lower than the anterior one. The maxilla−squamosal suture extends along the posterior boundary of the anterior zygomatic ridge, but its upper part has not been preserved. We reconstruct the dorsal part of the maxilla−squamosal suture ( Figs. 7 , 11C ), based mostly on PIN 4537/5 ( Fig. 1B 1, 3D) and a comparison with Djadochtatherium , as extending antero−dorsally and reaching the lacrimal. On both sides of PIN 4537/5 the zygomatic arch was slightly displaced along the posterior margin of the anterior zygomatic ridge, the lower margin of the arch to the rear of the anterior zygomatic ridge being placed higher than it originally was.

The posterior zygomatic “ridge” in Catopsbaatar is the weakest of the three ridges and the muscle scar related to it is as a rule marked only as a depression (arrow in Fig. 7 ). This depression is recognizable in PIN 4537/ 5 in lateral view ( Figs. 1B 1, 2D ) and in PM 120/107 ( Fig. 1C 1, C 3, on the left side), above the posterior part of the glenoid fossa.

The contact of the squamosal with the petrosal (mastoid) ( Fig. 1C 3) seen in occipital view appears as in Kryptobaatar ( Wible and Rougier 2000: figs. 16, 17, 35). The glenoid fossa preserved in ZPAL MgM−I/78, PM 120/107, and PIN 4537/5, when examined in lateral view slopes posteriorly (as characteristic of multituberculates, see Gambaryan and Kielan−Jaworowska 1995). The glenoid fossa ( Figs. 1B 2, 2C , 3E; see also Kielan−Jaworowska 1974: pl. XVIII: 1d, where it has been wrongly glued, as discussed under “ Holotype ” above), may be characterized as roughly semicircular, with its base arranged antero−medially (towards the orbit), and the margins of the semicircle developed as rounded ridges, protruding downwards ( Fig. 8 ). The greatest downward protrusion is at the antero−medial corner of the glenoid, where the surrounding ridge forms a bulge.

The glenoid fossa resembles that in Kryptobaatar dashzevegi , from which it differs in having an almost straight antero−medial margin, rather than more convex, which allowed Wible and Rougier to describe it in Kryptobaatar as tear−drop shaped.

Parietal.—The parietal bones, as with most bones of the cranial roof, consist of dorsal and lateral components. The dorsal components, which form the whole cranial roof posteriorly, have been preserved partly in the holotype (ZPAL MgM−I/78, see Kielan−Jaworowska 1974: pl. XVIII: 1c), PIN 4537/4 ( Figs. 1A 2 , 4b ) and PIN 4537/5 ( Figs. 1B 3, 3B). In these three specimens the posteriormost part of the parietal is missing. The parietal is completely preserved only in PM 120/107 ( Figs. 1C 6, 2A , see also Fig. 6). The fronto−parietal suture has been described above under “Frontal”. According to the reconstruction of the dorsal component of the parietal bone in Kryptobaatar dashzevegi by Wible and Rougier (2000: 46): “Anteriorly, they [the parietals] contact the frontals as a broad U−shaped suture; lateral to the arms of the U, a narrow anterior process of the parietals extends forwards, nearly to the lacrimal on the orbital rim, to form the posterior part of the supraorbital notch and supraorbital crest.” The pattern of bone arrangement in Catopsbaatar resembles very closely that in Kryptobaatar dashzevegi ; the state of preservation of all the skulls of Catopsbaatar does not allow us, however, to state with any certainty whether the similar anterior process of the parietal was also present in Catopsbaatar catopsaloides . Exceptions are the two juvenile skulls PIN 4537/4 and 4537/5, in which the small anterior processes are tentatively recognized.

The postorbital processes are very long. The parietal ridges extend from the posterior margin of the postorbital processes postero−medially, not reaching one another, but leaving a broad middorsal ridge between them. In this respect, the parietals of Catopsbaatar resemble those in Kryptobaatar dashzevegi ( Wible and Rougier 2000: figs. 9 and 32), but not in K. mandahuensis ( Smith et al. 2001: pl. 3: 1 and text−fig. 4), in which the parietal ridges meet at the end of the parietal. Lateral to the parietal ridges the parietals slope steeply latero−ventrally, giving the impression that the skull strongly narrows in the posterior part, which is not the case. Such structure of the parietals is shared with both species of Kryptobaatar ( Wible and Rougier 2000; Smith et al. 2001) and with Djadochtatherium , as observed on the specimen HMNS 94−10−278.

Along the posterior margin the parietals contact the supraoccipital medially and the petrosals (mastoids) laterally, forming together a very prominent nuchal crest ( Fig. 6). The nuchal crest extends transversely a short distance in the middle and then slopes postero−latero−ventrally, producing the extensive wings of the nuchal crest, which are convex posteriorly. Within the lateral parts of the nuchal crest the parietal contacts the squamosal.

The lateral component of the parietal contacts the frontal anteriorly; in the middle it has a long contact with the anterior lamina. The suture between the bones is generally straight, directed obliquely antero−dorsally/postero−ventrally. In the most posterior part the parietal contacts the squamosal.

Supraoccipital.—The supraoccipital of Catopsbaatar has been preserved only in PM 120/107, where it is partly distorted ( Fig. 1C 3). It is limited to the occipital plate and, as in Kryptobaatar dashzevegi ( Wible and Rougier 2000: figs. 7, 17, 35) and K. mandahuensis ( Smith et al. 2001: pl. 2: 3 and pl. 6: 3), it is hexagonal and contacts the parietal dorsally, the petrosals laterally, and the exoccipitals ventro−laterally. It contributes to the middle part of the upper margin of the foramen magnum ventrally.

Exoccipital.— Wible and Rougier (2000) divided the exoccipital in K. dashzevegi for descriptive purposes into: the occipital plate, the condyle, and contribution to the jugular fossa. In PM 120/107 and in ZPAL and PIN specimens all these elements are badly damaged, except perhaps for parts of the jugular fossae preserved in ZPAL MgM−I/78, described above under “Petrosal”.

Basioccipital.—The basioccipital is badly damaged in all ZPAL, PIN specimens and in PM 120/107; its reconstruction in Fig. 8 is based on a comparison with other djadochtatherioidean genera and is entirely tentative.

Endocranium.—The studied skull material of Catopsbaatar catopsaloides does not provide information on the endocranial structure in this taxon, except for some fragments preserved in ZPAL MgM−I/79, discussed and figured by Kielan−Jaworowska et al. (1986: 535 and pl. 2: 3).

Mandible.— Both dentaries have been preserved in the holotype ( ZPAL MgM−I/78, figured by Kielan−Jaworowska 1974: pl. V: 9, pl. XVII: 2, and pl. XX). Here we figure the fragmentary left mandible of ZPAL MgM−I/80 (Fig. 9A), fragmentary mandibles of juvenile PIN 4537 /4 ( Fig. 1A 4 , A 5 ), the right mandible of juvenile PIN 4537 /5 ( Fig. 1B 4 ), and both right and left complete mandibles of adult PM 120 /107 (Fig. 9B) .

In Catopsbaatar the mandible is robust and strongly elongated. The angle between the ventral border and the horizontal line (measured along the upper margin of p4 and the molars) varies around 17 degrees. In lateral view the diastema, which is distinctly concave dorsally, extends for 20 percent of the dentary length (measured without the incisor). The small mental foramen, plainly visible in ZPAL MgM−I/78, PM 120/107, PIN 4537/5, and PIN 4537/4, is situated close to the upper margin of the diastema, about the middle of the diastema length. The coronoid process has been most completely preserved in the right dentary of ZPAL MgM−I/78 ( Kielan−Jaworowska 1974: pl. V: 9, pl. XVII: 2, pl. XX: 1) and the right dentary of PIN 4537/5. In the latter specimen ( Fig. 1B 4 ) it apparently shows its complete height, but its posterior part is damaged, giving the process a needle−like appearance. It appears very long, possibly relatively the longest among all djadochtatherioideans and narrower than in other djadochtatherioidean taxa (compare Kielan−Jaworowska et al. 2004: fig. 8.40 and our reconstructions, Figs. 7 and 11C ), based on all available specimens. The coronoid process starts opposite m1 and rises at an angle of around 45 degrees to the horizontal, less steeply than in Kryptobaatar dashzevegi ( Wible and Rougier 2000: 60, and figs. 29 and 30, where it is about 60 degrees). It is distinctly separated from the alveolar process, which contains the teeth, by a wide temporal groove.

The condyle is situated only slightly above the upper level of the molars. It is devoid of a neck as in most djadochtatherioidean genera; it is oval in dorsal view and slopes posteriorly and slightly medially. The masseteric fossa is roughly triangular, extending anteriorly to below the m1. In the anterior part it is very sharply delimited ventrally by the masseteric crest, which is very prominent in its anterior course, where it forms a bulge, designated herein the masseteric protuberance (especially well−seen in the left dentary of PM 120/107, Fig. 9B 1). In poorly preserved dentaries (worn) the masseteric protuberance is hardly discernible (see e.g., Fig. 1A 4 , A 5 , and B 4). The masseteric crest is relatively more robust than in closely related Kryptobaatar , which does

Fig. 9. Catopsbaatar catopsaloides, Gobi Desert , Mongolia,?upper Campanian. A. Stereo−photographs of the partial left dentary of ZPAL MgM−I/80 from Ą Hermiin Tsav II, in medial (A 1), lateral (A 2), and dorsal (A 3) views. B. PM 120/107, Hermiin Tsav I. Stereo−photographs of the left dentary in lateral view (B 1), in medial view (B 2) and in occlusal view (B 3); stereo−photographs of the right dentary of the same specimen in dorsal (B 4) medial (B 5), and lateral (B 6) views. Note large masseteric protuberance at the anterior end of the masseteric crest in B 1, damaged but discernible in B 6 and A 2.

not develop the masseteric protuberance ( Kielan−Jaworowska 1971; Gambaryan and Kielan−Jaworowska 1995; Wible and Rougier 2000). A robust masseteric crest apparently ending with the masseteric protuberance occurs in Djadochtatherium matthewi (observed by ZK−J in the holotype specimen AMNH 20440) and in the fragmentary dentary from Ukhaa Tolgod (GI 5/301) figured by Kielan−Jaworowska and Hurum (1997: fig. 5). The masseteric protuberance apparently occurs also in Taeniolabis taoensis ( Granger and Simpson 1929: fig. 4; Simmons 1987: fig. 4.4).

The masseteric fovea is situated in front of the masseteric fossa ( Gambaryan and Kielan−Jaworowska 1995: fig. 1); the fovea is well−visible in both fragmentary dentaries of PIN 4537/4 ( Fig. 1A 4 , A 5 ) and in PIN 4537/5 ( Fig. 1B 4 ); it is probably more distinct in Catopsbaatar than in other djadochtatherioidean genera (see also Fig. 9B 1, B 6).

On the medial side the symphysis is best seen on the left dentary of ZPAL MgM−I/80 (Fig. 9A 1) and on the right dentary of PM 120/107 (Fig. 9B 2, B 5) where it forms a tear dropshaped structure, vertically directed and narrowing ventrally.

The pterygoid fossa is very large, occupying the whole posterior surface of the middle part of the dentary to the rear of m2. In both dentaries of PM 120 / 107 (Fig. 9B 2, B 5) the fossa appears to be divided by a vertical ridge into two parts, the smaller of which is the anterior one. The division of the fossa is not discernible in the holotype specimen figured by Kielan−Jaworowska (1974: pl. XVII: 2c and pl. XX: 1b), nor in PIN 4537 / 5 (not figured). On the other hand, Gambaryan and Kielan−Jaworowska (1995) divided the pterygoid fossa in Nemegtbaatar into two parts, the anterior one deeper, designated the pterygoid fossa (for m. pterygoideus medialis), and the shallower and smaller posterior one, which they designated the pterygoid fovea (for m. pterygoideus lateralis), which would correspond to the division observed by us in PM 120 /107.

In Catopsbaatar the pterygoid fossa is bounded ventrally by an extensive pterygoideus shelf, which narrows posteriorly. The shelf was referred to by Simpson (1926) as the pterygoid crest, but we follow Miao (1988), who designated it the pterygoideus shelf.

In dorsal view (Fig. 9B 3, B 4) the area of the diastema forms a wide shelf, which gently slopes downwards medially. On both sides a ridge protrudes; laterally it is the masseteric protuberance at the anterior part of the masseteric crest, and on the medial side the pterygoideus shelf.

Dentition.—The dental formula of Catopsbaatar catopsaloides is 2032/1022.

Upper teeth: In the holotype ZPAL MgM−I/78 the I2 are broken, only the roots are preserved; in ZPAL MgM−I/80 I2 and I3 are missing; in ZPAL MgM−I/79, the left I2 has been preserved ( Kielan−Jaworowska 1974: pl. XXI), but it is now broken, only the roots are exposed dorsally, the right root measured on this specimen is 10.3 mm long and extends further posteriorly, hidden in the alveolus. Both I 2 have been preserved in PM 120 /107; are very robust, with a sharply limited enamel band, and converge slightly medially, touching one another ( Figs. 1C 4 , C 5 , 2C ). In PIN 4537 /4 there are the roots of both I2, while in PIN 4537 /5 the root of right I2 and complete left I2 ( Figs. 1B 2 , 3E) are preserved .

The I3 are poorly known. The right I3 has been preserved in the holotype ZPAL MgM−I/78 ( Kielan−Jaworowska 1974: pl. XVIII: 1d) and is cone−like with a broken tip; on the left side it is broken. The right I3 is also present in PIN 4537/4 ( Figs. 1A 1, 4E ). The broken I3 is well seen on the right side of ZPAL MgM−I/79 ( Kielan−Jaworowska 1974: pl. XXI: 1b), while on the left side only the alveolus has been preserved. In ZPAL MgM−I/80 and PM 120/107 both I3 are missing. In PIN 4537/5 this region is badly damaged and both I3 are missing.

The anterior upper premolars P1 and P3 are present only in juvenile individuals, in the older ones they are lost and even the alveoli disappear. In the holotype ZPAL MgM−I/78, which belongs to a young individual, all the upper premolars and molars have been preserved, the M2s are still erupting ( Kielan−Jaworowska 1974: text−fig. 6; pl. XVIII: 1b, d; Kielan−Jaworowska et al. 1986: pl. 1: 2a; and our Fig. 10A 5 ). On the left side the P1 appears to have two uncertain cusps; no cusps have been preserved on P3. In PIN 4537/4 ( Figs. 1A 1, 4E ) the right DP1 is single−rooted, with a cone−like blunt crown. On the left side there is the single root of DP 1 in its alveolus. The roots of the right and left DP3 are preserved. Judging from them, DP3 is single−rooted and smaller than DP1. In PIN 4537/5 this region is badly damaged, and P1 and P3 are not preserved. In ZPAL MgM−I/79, which belongs to an adult individual, roots of P1 and P3 are preserved on both sides ( Kielan−Jaworowska 1974: pl. XXI: 1b.) In ZPAL MgM−I/80 ( Fig. 5A ), which belongs to a relatively young individual (the m2 on the left side is still erupting, and cusps on the molars are moderately worn), only two roots (P1 and P3) have been preserved on the left side (though less certain on the right side). For interpretation of these roots see caption to Fig. 5 . In PM 120/107 ( Fig. 2C ) no traces of the roots of P1 or P3 are preserved.

The P4 cusp formula is 5−4:1. The P4 has been preserved on both sides of ZPAL MgM−I/78 ( Fig. 10A 5 ) and has been schematically figured by Kielan−Jaworowska (1974: fig. 6). In PM 120/107 the P4 is preserved on both sides ( Fig. 2C ). In PIN 4537/5 it is preserved on the right side and with cusp formula 4:1, the central cusp being distinctly larger than the others ( Fig. 3E). In PIN 4537/4 the P4 is damaged on both sides. In ZPAL MgM−I/79 the P4 is missing on the left side and badly damaged on the right; in ZPAL MgM−I/80 it is preserved only on the left side. Our reconstruction in Fig. 8 has been based on all the specimens.

Both M1 have been preserved (at least partially) in all ZPAL, PM and PIN skulls. The cusp formula of M1 is 5−6:5−6:4. The inner ridge extends for about 75 percent of the tooth length. In juvenile individuals (ZPAL MgM−I/78 and PIN 4537/5) the cusps are sharp and not worn, moderately worn in ZPAL MgM−I/80 and PIN 4537/4, while in ZPAL MgM−I/79 and PM 120/107 they are strongly worn and concave. The same concerns M2s.

The M2 has also been preserved in all the specimens. The cusp formula is 2:2−3:2−3. In ZPAL MgM−I/78 both M2 are erupting.

Lower teeth: The single pair of the lower incisors, characteristic of all Multituberculata , is very strong and compressed laterally in Catopsbaatar . It has a sharply limited enamel band and was ever−growing. It is best preserved in the right dentary of the holotype ZPAL MgM−I/78 ( Kielan−Jaworowska 1974: pl. XVII: 2a, pl. XX: 1a, b) and PM 120/107 (Fig. 9B), in which in dorsal view the large, oval, worn surface is clearly seen on its posterior surface.

Of the two premolars, the p3 is very small entirely adhering to the posterior wall of the diastema and hidden under p4. It has been preserved in the right dentary of ZPAL MgM−I/78 ( Fig. 10A 1) and partly in the left dentary ( Fig. 10A 3). In the badly damaged anterior portions of the dentaries of ZPAL MgM−I/80 (Fig. 9A) the p3s are poorly preserved; while in PM 120/107 only the upper part of p3 has been preserved in the right dentary, but is missing from the left (Fig. 9B). In PIN 4537/4 the single root of p3 is preserved on both sides ( Fig. 1A 4 , A 5 ). The p4 has been well preserved in both dentaries of ZPAL MgM−I/78 ( Kielan−Jaworowska 1974: pl. XVII: 2, pl. XX; and our Fig. 10A, B ), in left ZPAL MgM−I/80 (Fig. 9A), in left PIN 4537/4 ( Fig. 1A 4 ), in left PIN 4537/5 (not figured), and in both PM 120/107 (Fig. 9B). The p4 is roughly trapezoidal in lateral view, blade−like, with three cusps along the straight, horizontal upper margin and a single postero−labial cusp. The labial and lingual ridges, characteristic of p4s of most multituberculates, including Kryptobaatar and Djadochtatherium , are absent in Catopsbaatar .

The m1 ( Fig. 10A 1 –A 4 ) is almost symmetrical with cusp formula 4:4. The size of the cusps varies even between right and left teeth of the same specimen; e.g., in the right dentary of ZPAL MgM−I/78 the cusps in the outer row are larger than in the inner row, while in the left dentary those of the inner row are larger. In both rows the size of cusps decreases anteriorly.

The cusp formula of m2 is 2−3:2; however, in most specimens it is 2:2, except for the right dentary of PM 120/107, in which it is 3:2. The cusps of the medial row are distinctly wider than in the lateral row, the medial row of cusps is shorter than the lateral one, and the posterior margin of the tooth is obliquely arranged. The single right m2 from Khulsan (ZPAL MgM−I/159) is more worn than all the m2s preserved in dentaries, and because of wear, may appear on first sight different, e.g., from that in the holotype specimen ( Fig. 10A 4 and 10B respectively). It is insignificantly wider than the m2s preserved in dentaries associated with skulls ( Table 2); however, we believe that there is no reason to assign it to species other than Catopsbaatar catopsaloides . We figure the Khulsan specimen herein ( Fig. 10B ) in order to demonstrate that C. catopsaloides was not restricted to Hermiin Tsav (which yielded most of its specimens), but occurred also in the Baruungoyot Formation at an other Gobi Desert locality, being very rare there.

Figures

Fig. 11. Comparison of three djadochtatheriid skulls in lateral view, rendered to approximately the same length. The muscle scars surrounded by the anterior and intermediate zygomatic ridges are shaded.The details of the structure of the orbit have been omitted. A. Kryptobaatar dashzevegi Kielan−Jaworowska, 1969, based on photographs and reconstruction of the specimens from Ukhaa Tolgod, figured by Wible and Rougier (2000), and skulls from Bayan Zag housed in ZPAL. The vascular nasal foramina as preserved in the holotype ZPAL MgM−I/21. B. Djadochtatherium matthewi Simpson, 1925, tentative reconstruction based on: the holotype AMNH 20440 (rostral part of the skull and both dentaries from Bayan Zag); HMNS 94−10−278 (fairly complete skull with incomplete dentaries from Tögrög, examined on the photographs); GI 5/301 (partial dentary from Ukhaa Tolgod); and the photograph of the complete skull with both dentaries from Ukhaa Tolgod, figured by Webster (1996, the number not available). C. Catopsbaatar catopsaloides (Kielan−Jaworowska, 1994), reconstruction of the skull in lateral view, based on all known specimens. Orbital wings of the lacrimal, not preserved in Djadochtatherium (B) and Catopsbaatar (C) have been reconstructed on the basis of comparisons with Kryptobaatar (A).

Fig. 10. Catopsbaatar catopsaloides, Gobi Desert, Mongolia,?upper Campanian. A. ZPAL MgM−I/78, holotype (juvenile individual) from Hermiin Tsav I. SEM micrographs of the dentition. Fragment of the right dentary in lateral view, showing p3 (peg−like adhering the wall of the dentary), p4 and m1 (A1). Stereo−micrographs of p4, m1, and m2 of the same in dorsal view (A2). Fragment of the left dentary of the same specimen, showing p4 and m1, the p3 is broken, only its lower part has been preserved (A3). Stereo−photograph of the complete right side of the upper dentition, showing P1, P3, P4, M1, and erupting M2 (A5); P1 and P3 are double−rooted. B. ZPAL MgM−I/159, Khulsan, Nemegt Valley, right m2 in occlusal view.

Fig. 5. Catopsbaatar catopsaloides, ZPAL MgM−I/80, Hermiin Tsav II, Gobi Desert, Mongolia,?upper Campanian, severely damaged skull of a juvenile individual; stereo−photograph in ventral view (A), dorsal view of the same (B). The palatine bone, damaged in other specimens, has been preserved, showing major palatine foramina and palatonasal notches. The roots (in A) designated P1 and P3 might either belong to deciduous premolars, or might represent the posterior root of P1 and anterior root of P3 respectively. The latter interpretation appears more probable as the skull MgM−I/78 figured here is slightly bigger (see Table 1) than that of the holotype MgM−I/78, in which there are double−rooted P1 and P3.

Fig. 4. Catopsbaatar catopsaloides, PIN 4537/4, Hermiin Tsav, Gobi Desert, Mongolia,?upper Campanian, stereo−photographs of an incomplete, severely damaged skull of a juvenile individual in anterior (A), dorsal (B), left lateral (C), right lateral (D), and ventral (E) views.Anterior upper premolars preserved on both sides of this specimen are recognized as deciduous, as they are single−rooted, rather than double−rooted as in specimens of more mature individuals. The tripartite infraorbital foramen is seen on the right side in A and the premaxillary ridge on the right side in E; single−rooted DP1 and DP3 (replaced in ontogeny by double−rooted P1, and P3, which disappear in adult individuals), are preserved in this specimen (E).

Fig. 2. Catopsbaatar catopsaloides, PM 120/107, Hermiin Tsav I, Gobi Desert, Mongolia,?upper Campanian, stereo−photographs of the skull of an adult individual, with missing choanal and damaged basicranial regions, in dorsal (A), right lateral (B), and ventral (C) views. In A on the left side of the specimen a depression is seen medially, at the posterior part of zygomatic arch. In comparison with Ptilodus, Nemegtbaatar, and Kryptobaatar, it is situated too far posteriorly to house the jugal bone (not preserved). The premaxillary ridge is seen in C; the P1 and P3, present in juvenile individuals (best seen in Figs. 4E and 10A) have been resorbed.In B the suture between the maxilla and squamosal is discernible along the posterior margin of the anterior zygomatic ridge.

Fig. 7. Catopsbaatar catopsaloides, reconstruction of the skull, based on all available specimens, in lateral view. The reconstruction is of an adult individual, but the upper premolars P1 and P3, which disappear during ontogeny, are reconstructed. The bones of the orbit, not preserved in ZPAL and PM specimens and very fragmentarily preserved only in PIN 4537/4 and /5, have not been reconstructed. The dorsal part of the maxilla/squamosal suture is tentatively reconstructed on the basis of PIN 4537/5 (Fig. 3C, D) and a comparison with Djadochtatherium (Fig. 11B). The arrow points to the posterior zygomatic “ridge” (muscle scar), preserved on the squamosal above the glenoid fossa and discernible in occipital view. The teeth not marked on the dentary are p4 and m1, on maxilla P4 and M1.

Fig. 8. Catopsbaatar catopsaloides, reconstruction of the skull, based on all available specimens, in ventral view. The reconstruction is of an adult individual, but the upper premolars P1 and P3, which disappear during ontogeny, have been reconstructed. The teeth in maxilla are P1, P3, P4, M1, and M2. The choanal region and the middle part of the basicranial region (diagrammatical), including recognition of most of the foramina not preserved or poorly preserved in Catopsbaatar (see text for details), have been reconstructed on the basis of other djadochtatherioid genera, especially Kamptobaatar (Kielan−Jaworowska 1971), Nemegtbaatar (Kielan−Jaworowska et al. 1976), and Kryptobaatar (Wible and Rougier 2000).