Upper Premolars (figs. 1, 3)

There are five upper premolars, which we designate P1–P5. The P1 is a small, mediolaterally compressed, digitiform, single-rooted tooth that tapers to a blunt tip. The crowns of P2 and P3 are broken. P2 has two subequal roots, is more than twice the length of P1, but is similar in width. At its alveolus, the single-rooted P3 is similar in dimensions to P1, but slightly smaller. There is little space between the canine alveolus, P1, P2, and P3, and none of these is imbricated.

P4 is a tall, trenchant, mediolaterally compressed tooth, much longer than wide (figs. 3, 4A, B). In occlusal view (fig. 4B), it is triangular with a large main cusp, the paracone, situated along the labial margin and a small posterolingual (protoconal) swelling. At the mesial end of the preparacrista are a small parastyle and parastylar lobe, and at the distal end of the postparacrista are a small metastyle and metastylar lobe. The metastyle is better developed than the parastyle, whereas the metastylar lobe is smaller than the parastylar lobe. There is no metacone, although there is a very faint indication of a metaconal swelling on the postparacrista visible in labial view (fig. 4A). On the lingual aspect of the preparacrista is a broad, concave, fusiform wear facet formed by the distal crest of the main cusp of p4 (fig. 4B). The postparacrista has a much narrower wear facet formed by the mesial crest of the main cusp of p5. The CT scans show two roots on P4 with the distal root much wider than the mesial one. P4 is separated by narrow diastemata from the P3 and P5 (fig. 3). Distal to the protoconal swelling of P4 is a small pit in the maxilla for the main cusp of p5 (fig. 3).

Compared to P4, P5 is a more molariform tooth (fig. 4A, B); it is not as tall or long as P4, but is considerably wider and has three roots. The tallest cusp is the paracone, which is centrally located along the mesiodistal axis, close to the labial margin. The preparacrista is steeper and straighter than the postparacrista. The postparacrista curves distolabially to the well-developed metastyle, without any indication of a metacone but with a trace metaconal swelling visible in labial view (fig. 4A); a parastylar groove separates the preparacrista from the parastyle, which is distinct, but lower than the metastyle. A narrow stylar shelf connects the parastyle and metastyle; this cingulum is widest mesially and distally, but is very narrow in between, opposite the paracone tip. The second tallest cusp is the protocone, which is narrow, procumbent, and at the lingual margin. The preprotocrista extends to the parastyle; the postparacrista does not extend beyond the paracone. Neither crest shows any indication of conules. A short crest runs from the paracone base toward the protocone, dividing the trigon into smaller mesial and larger distal basins, probably a by-product of wear. A narrow precingulum extends from a level lingual to the paracone toward the lingual margin, and a wider postcingulum is situated at the distolingual margin. The P5 is separated from M1 by a narrow gap (fig. 3). Distal to the postcingulum of P5 is a small pit in the maxilla for the metaconid of m1 (fig. 3).

Upper Molars (figs. 2A, B, 3, 4A, B)

There are three upper molars, all much wider than long. M1 is the longest, M2 the widest, and M3 the shortest but intermediate in width. On M1, the paracone is the tallest cusp, with the metacone much shorter but slightly taller than the protocone. The paracone and metacone bases are adjoined, and the metacone base is approximately at the same level as the paracone base (i.e., immediately distal to the paracone). The centrocrista is straight. The preparacrista is short and weak, and extends to a low stylocone (stylar cusp B of Reig et al., 1987) on the parastylar lobe. The low parastyle (stylar cusp A of Reig et al., 1987) is connected to the stylocone by a short crest of comparable height to the cusps. The parastylar lobe has a weak anterolabial cingulum of Reig et al. (1987) or preparacingulum of Rougier et al. (1998) that does not extend lingually beyond the paracone. The postmetacrista is longer and more prominent than the preparacrista; it extends to the metastyle (stylar cusp E of Reig et al., 1987). The parastylar and metastylar lobes are similar in size, but the latter extends slightly more labially. The metastylar lobe contacts the parastylar lobe of M2. A low, narrow stylar shelf devoid of cusps connects the parastylar and metastylar lobes of M1. The ectoflexus is very shallow. The protocone is slightly procumbent, narrow, and lingually placed. The preprotocrista has a very small elevation just lingual to the paracone that represents a tiny paraconule. Between the paraconule and paracone is a small triangular wear surface, the mesial and distal edges of which represent very short, low cristae. The preparaconular crista is worn and appears to extend labially beyond the paracone base. The postprotocrista also has a slight elevation just lingual to the metacone that is a small or worn metaconule. The postmetaconular crista extends distal to the middle of the metacone base; a premetaconular crista is lacking. A narrow precingulum is positioned at the level of the middle third of the distance between the paracone and protocone. A wider postcingulum of Nessov et al. (1998) extends nearly the total distance between the protocone and metacone. During study, the tip of the metacone broke off, but fortunately was not lost.

The M2 (figs. 2A, B, 4A, B) generally resembles M1, but there are some significant differences. The M2 is shorter than M1, with the parastylar and metastylar lobes positioned more labially, producing a deeper ectoflexus and increasing the length of the preparacrista and postmetacrista. The parastylar lobe bears a parastyle, which is more prominent than the metastyle, and a stylocone that is shorter compared to that on M1. The metastylar lobe is separated from the parastylar lobe of M3 by a narrow gap. Compared to the paracone, the metacone is shorter on M2 and is subequal in height to the protocone. The protocone is shifted labially to a slight degree and is more procumbent than that on M1. The paraconule is less prominent, whereas the metaconule is more prominent with a weak premetaconule crista. The preparaconular crista is worn but does not appear to extend labially beyond the paracone base. The precingulum is shifted slightly labially, and the postcingulum broadens at the distolingual corner of the tooth to form a low, but distinct hypocone.

The M3 (fig. 4A, B) lacks a metastylar lobe. The parastylar lobe bears a parastyle and a shorter stylocone. Compared to M1 and M2, the metacone is even shorter compared to the paracone, with the protocone the second highest cusp. The preprotocrista widens as it approaches the paracone into a low paraconule that has a small triangular wear facet labial to it. The postprotocrista has an elevation just behind the metacone that represents a metaconule. The labial face of the metaconule has a small wear facet set at a right angle to a larger, vertical wear facet on the distolingual face of the metacone. The pre- and postcingulum are much reduced compared to those on M1 and M2.

Lower Incisors (figs. 5–9)

The left lower jaw has been removed from the skull and preserves the full dentition. There are three procumbent lower incisors decreasing slightly in size and procumbency posteriorly. We designate these as i1, i2, and i3. The i3 is separated from the lower canine by a narrow diastema. Because of this separation, our working hypothesis is that the incisors are homologs of the first three incisors of Ukhaatherium, which has four.

The i1 is the most damaged incisor, represented only by its root (fig. 8). The root is exposed by breakage and follows along the contour of the ventromedial border of the dentary from a level opposite the distal edge of the lower canine to the preserved tip of the dentary. Consequently, posteriorly the root is nearly horizontal, whereas it is curved slightly dorsally at the tip. The anterior three-fourths of the root is damaged along its ventromedial edge, exposing the pulp cavity. The anterior end of the root is broken, leaving a jagged end. Based on the CT sections, the i1 root is the longest and the largest of the incisors in cross section; it extends more posteriorly than the roots of i2 and i3 (fig. 9). Also, the root is oval in cross section, procumbent (with the long axis tilted up 60° from the horizontal), uniform in size throughout most of its length (tapering slightly at its end), and is closed posteriorly. As preserved, it is uncertain whether there is enamel on the root.

The i2 is a short cylindrical tooth, subcircular in cross section, that tapers slightly toward it tip, which has a vertical break (fig. 7). It is unknown how much of the tooth is missing, but what is preserved of the crown is procumbent. In anterior view, the i2 is dorsolateral to the anterior tip of the i1 root. The i2 alveolus is dorsal to that of i1, and i2 at its alveolus is slightly smaller in diameter than the anterior i1 root. The incidence of enamel on i2 is uncertain, but there is no evidence of restricted enamel. Based on the CT sections, the i2 root is oval in cross section anteriorly (with the long axis tilted up 70° from the horizontal) and tapers to a small circle posteriorly. The root has no large apical opening as would be expected in an ever-growing tooth and extends nearly as far back as the root of i3, just in front of the root of the lower canine (fig. 9).

In anterior view, i3 is positioned slightly dorsolateral to i2; the i3 alveolus is posterior to that of i2. At its alveolus, i3 is comparable in size to i2, but it tapers more to its tip, which is broken (fig. 7). It is unknown how much of the tooth is missing; the preserved crown is procumbent, although less so that i2. From the tip to the end of its root in front of the lower canine root, i3 is shorter than i2. The proximal two-thirds of the extra-alveolar part of i3 is oval in cross section (with the long axis tilted up 80° from the horizontal), tapering from proximal to distal. The distal one-third bears an oval-shaped broken surface on its dorsomedial aspect. Exposed along the edge of the broken surface is enamel, which covers the remainder of the tip. The enamel appears restricted to the tip and does not extend to the alveolus. Based on the CT sections, the i3 root is closed posteriorly and resembles that of i2 in cross-sectional shape but is smaller (fig. 9).

Lower Canine (figs. 5–9)

The lower canine is single rooted, large, pointed, trenchant, and arcs in a mesiodorsal direction. Its labial surface is rounded, its lingual surface flattened; consequently, a horizontal cross section is D-shaped with the straight side facing lingually. Enamel is broken off the middle of the lingual surface. Based on the labial surface, it is clear that enamel continues into the alveolus. The canine is separated by small diastemata from i3 in front and p1 behind. However, the canine alveolus is only narrowly separated from the alveoli of these teeth, which means that the canine does not entirely fill its alveolus. The canine root is open, as is visible in the CT scans (fig. 9B) and in medial view of the dentary (fig. 8), because of bone loss along the ventromedial aspect below p1 and p2. Given that the canine root is open and the alveolus not filled, we interpret that this tooth is not fully erupted.

Lower Premolars (figs. 5–9)

There are five lower premolars, which we designate p1–p5. The p1 is a tiny, digitiform tooth that tapers to a blunt tip and is slightly labiolingually compressed (fig. 5). It is procumbent, but less so than i3. It is separated from p2 by a short diastema and from the canine by an even shorter diastema. The CT scans show a single short root that is directed distoventrolingually and ends just in front of the mesiolingual surface of the mesial root of p2 (fig. 9B).

The p2 is a trenchant tooth with a main anterior cusp, labiolingually compressed, a prominent, low distal basal cusp, and no mesial basal cusp (fig. 5). It has two roots (figs. 7, 8), with the distal root larger than the mesial, as visible in the CT scans (fig. 9B). The p2 is much larger than p1 and p3, but considerably smaller than p4 and p5. The p2 is separated from p1 by a short diastema, as noted above, and from p3 by an even shorter diastema.

The p3 is broken off near its base; what is preserved is small, erect, and labiolingually compressed (figs. 7, 8). At its alveolus, p3 is larger than p1. The p3 is separated by short diastemata from p2 and p4, although the size of the latter may have been affected by a crack through the dentary. Based on the CT scans, the p3 root is short and straight and ends dorsal to the mandibular canal (fig. 9A).

The p4 is a tall, trenchant tooth with a labiolingually compressed main cusp and a prominent, low distal basal cusp (figs. 4C–E, 5). Mesially, where the crown meets the mesial root is a very weak swelling that may be the remains of a heavily worn mesial basal cusp. The anterior surface of the main cusp is set back from the anterior root, further suggesting heavy wear. The distolabial surface of the main cusp and the mesiolabial surface of the distal basal cusp show a flat wear facet formed by the tall, trenchant P4. The CT scans show the posterior root of p4 to be slightly larger than the anterior, with the roots extending nearly to the ventral border of the dentary (fig. 9B).

The p5 is a taller, more robust version of p4 (figs. 4C–E, 5). The main cusp is taller and thicker and the distal basal cusp higher and wider than those on p4. The distal surface of the main cusp and the mesial face of the distal cusp have flat wear facets for the pre- and postparacrista of P5; these wear surfaces are larger than those on p4. Mesially, where the crown meets the mesial root is a worn mesial basal cusp that is higher and more prominent than that on p4. Also, the anterior surface of the main cusp is set back from the anterior root, but not to the extent of p4. The CT scans show the mesial root of p5 to be slightly larger than the distal, both extending to the mandibular canal (fig. 9B).

Lower Molars (figs. 2C, D, 4C–E, 5)

There are three lower molars, with m3 the longest and m2 slightly shorter than m1. On m1 (figs. 2C, D, 4C–E), the trigonid is wider and shorter than the talonid. The highest and largest cusp is the metaconid, which is only slightly taller than the protoconid; the much smaller paraconid is in a mesiolingual position just above the distal basal cusp of p5. The protocristid is slightly oblique to the long axis of the tooth; that is, the metaconid is slightly distal to the protoconid (fig. 4C). A shelflike precingulid extends from the labial base of the protoconid anterodorsally to just below the level of the distal basal cusp of p5. The mesial end of the precingulid forms a distinct, low cingular (mesiolabial) cuspule f (Crompton, 1974); a cingular (mesiolingual) cuspule e is lacking. The m1 talonid is slightly wider than long. The highest cusp on the talonid is the entoconid, which is located at the distolingual corner. A high postcristid extends labially and slightly distally from the entoconid to the hypoconulid, which is only slightly lower than the entoconid and nearer the entoconid than hypoconid. The distal surface of the hypoconulid contacts the m2 lingual to its cuspule f and ventral to its broken paraconid. A low crest curves mesiolabially from the hypoconulid to the hypoconid, the lowest cusp on the talonid. A steep entocristid extends mesially from the entoconid and is separated from the metaconid base by the talonid notch, which represents the deepest part of the talonid. The cristid obliqua extends mesially from the hypoconid to contact the back wall of the trigonid just labial to the bottom of the V of the protocristid. In labial view, the hypoconid is 50% of the height of the protoconid measured from the base of the crown. The CT scans show both roots of m1 (and only the mesial root of m2). The m1 roots are subequal, with the mesial one slightly bowed anteriorly, and extend ventrally to the mandibular canal (fig. 9).

The m2 (fig. 4C–E) is damaged, with the anterior part of the trigonid missing. This includes the paraconid, the anterolabial face of the metaconid, and the anterolingual face of the protoconid. The m2 resembles m1 in most features. Differences include a more transverse protocristid, a protoconid nearer in height to the metaconid (though still slightly shorter), a more lingually placed cristid obliqua (in the bottom of the V of the protocristid), a more worn entoconid, and a narrower talonid compared to the trigonid. In labial view, the hypoconid is 49% of the height of the protoconid measured from the base of the crown.

The m3 (fig. 4C–E) is the least worn molar. Its trigonid is comparable in size to that of the other molars, but its talonid is longer and narrower. The disposition of the trigonid cusps and precingulid resembles the other molars. However, the protoconid is the same height as the metaconid, although the metaconid is still larger at its base; the paraconid is still a low cusp, but it is more substantial in size; and the protocristid is transverse, as in m2. The m3 talonid is different from that on m1 and m2. The hypoconulid is more distally positioned and is the tallest and largest cusp; it is erect and reaches nearly to the same level as the paraconid. The entoconid is the next highest cusp and is more anteriorly positioned, in light of the more distally placed hypoconulid. The hypoconid is also more anteriorly positioned with the crest between the hypoconid and hypoconulid more oblique than the postcristid. Lastly, the cristid obliqua is more lingually placed (just lingual to the bottom of the V of the protocristid). In labial view, the hypoconid is 46% the height of the protoconid measured from the base of the crown.

Rostrum

In the skull, little bone is preserved on the dorsal aspect of the rostrum and the dorsal and lateral aspects of the braincase (figs. 10, 11, 13). On the lateral aspects of the rostrum, much of the facial processes of the both maxillae are preserved (figs. 11, 13). Compared to the rest of the skull, much of the bone that is preserved on the rostrum is worn and the difference between bone and impression of bone on matrix is not always clear.

On the dorsum, a segment of the paired nasal (“na” in fig. 10) that extends from the level of the anterior margin of the canine alveolus to the P4 is indicated by scraps of preserved bone and by the impression of these bones on the underlying matrix. In this segment, the nasals are narrow, with parallel sides. Between the orbits, a flat, roughly quadrangular bone fragment includes part of the paired frontal (“fr” in fig. 10). On the midline, remnants of a suture between the left and right frontals are visible. In the right anteromedial aspect of the quadrangle is a small quadrangular piece of a second overlying bone, part of the right nasal. This indicates that the nasal extended at least to the level of the anterior orbital rim near the midline. However, whether the nasal was expanded posteriorly is uncertain. Posterolateral to the quadrangle of frontal on the right side is a small, raised digitiform piece of worn bone in the position where a postorbital process of the frontal (“pop” in fig. 10) would be expected; the left side here is flat, but the bone is even more worn. Further posteriorly, near what may have been the interorbital constriction are more bone fragments. The frontal is positioned centrally, and lateral to that preserved on both sides is a layer of overlying parietal (“pa” in fig. 10). This represents the anterior margin of the parietal, indicating that bone reached anteriorly to the level of the interorbital constriction. The anterior margin of the parietals appears to be either arcuate or the shape of an inverted V, but the shape of the intervening frontal-parietal suture is not known. On the right side, the anteriormost part of the arc is missing, exposing a facet on the frontal for the parietal (“paf” in fig. 10).

On the face, the facial process of the left maxilla (“mx” in fig. 11) is preserved in proximity to the postcanine dentition. Missing are the parts forming the lateral wall of the canine alveolus and the roof of the nasal cavity. Anterior to the canine alveolus is an oblique line of bone, angling posterodorsally, which dorsally overlies the lateral margin of the nasal and ventromedially is continuous with the maxilla. About halfway up this line of bone is a faint suture indicating two layers, the outer ventral one being the maxilla and the inner one being the facial process of the premaxilla (“pmx” in fig. 11), which extends dorsally to overlie the nasal. The preserved anatomy suggests that the facial process of the premaxilla only extended to the level of the anterior canine alveolus. The most prominent feature of the facial process of the maxilla is the infraorbital foramen (“iof” in fig. 11). It is positioned at the level of the paracone of P4. It is oval in anterior view, higher than wide, with a sizeable sulcus extending anteriorly from it to the level of P3. Within the orbit, the maxillary foramen (“mxf” in fig. 12), the posterior opening of the infraorbital canal, is positioned dorsal to the P5–M1 embrasure, which means the infraorbital canal is the length of P5 and the posterior root of P4. On the right side (fig. 13), the facial process of the maxilla proximal to the postcanine dentition has been worn down, exposing the posterior root of P2, the single root of P3, and the labial roots of P4-M2. Dorsal to the posterolabial root of P4 and the labial roots of P5 is a narrow, longitudinal bar of matrix representing the infraorbital canal (fig. 13). Dorsal to the anterior premolars is a larger area of more opaque matrix presumably representing the nasal cavity.

At the anterior root of the left zygoma, the jugal (“ju” in fig. 11) overlies the facial processes of the maxilla and lacrimal (“lac” in fig. 11) forming an elongate, oblique contact. The bulk of the preserved contact is with the maxilla, with the lacrimal underlying only the anterodorsal tip of the jugal. In ventral view (fig. 3), the jugal approximates the labial margin of the molars, and only a small sliver of maxilla is visible posterior to the jugal opposite the parastylar lobe of M3, serving as the zygomatic process (“zmx” in fig. 14). Posteriorly (fig. 11), the inferior edge of the jugal lies just dorsal to the embrasure between M2 and M3. Anteriorly, the inferior edge of the jugal lies dorsal to the embrasure between M1 and M2 at the level of the dorsal margin of the infraorbital foramen. The anterior margin of the jugal is broken and continued forward at least to the level of the anterior root of P5, as evidenced by a facet partly on the maxilla and partly on the lacrimal. The anterodorsal margin of this facet is missing, so that the full anterior extent of the jugal is unknown. The superior margin of the jugal at the anterior zygoma is gently curved, with a sharp edge forming the infraorbital margin. The inferior margin of the jugal bears a distinct ventrally directed process, a masseteric spine (Krause, 1884), dorsal to the M2–M3 embrasure (“mas” in figs. 3, 11).

The damaged facial process of the left lacrimal is interposed between the jugal inferiorly and frontal superiorly, and forms the anterior orbital rim (figs. 11, 12). Of the anterior rim, only a central segment is preserved on the lacrimal. It is smooth and bears no tubercle, but the incidence of this process cannot be denied or confirmed. The missing rim on the ventrolateral aspect of the lacrimal uncovers the presence of two subequal lacrimal foramina within the lacrimal's orbital process, one ventrolateral and the other dorsomedial (“lacf” in figs. 11, 12). As preserved, the lacrimal's facial process is small and narrow (fig. 11). However, the disposition of the portion of the lacrimal covered by the jugal and the impression of the rostral continuation of the maxillary-jugal suture shows that the lacrimal's facial process was larger than what is preserved.

Palate (fig. 3)

On the palate, the palatal portions of the premaxillae are wholly lacking. The palatal processes of the maxilla and palatine (“pal” in fig. 3) are fully exposed on the left side and only the part nearest the midline is exposed on the right. The anterior ends of the left and right maxillae are broken and the lateral part of the left canine alveolus is missing.

From the level of P3 forward, the palatal process of the maxilla is essentially flat and featureless (fig. 3). A midline palatal vacuity (“pv” in fig. 3) lies between the palatal processes of the maxillae and palatines, between the levels of the P5 parastyle and M1 metastyle. The maxillae border the concave anterior half of the vacuity and the palatines border the convex posterior half. The lateral margins of the vacuity seem well preserved, but the anterior and posterior borders at the midline probably are not complete. Consequently, we anticipate the presence of a midline septum dividing the vacuity into left and right sides. There are no separate major palatine foramina. The palatal vacuity transmitted the major palatine nerve and vessels as evidenced by a short, shallow bilateral groove (“gmpn” in fig. 3) running anterolaterally from the palatal vacuity to the level of P4, which transmitted the major palatine nerves and vessels. Medial to the molars, only a narrow strip of maxilla is visible on the palate.

The palatines complete the palate medial to the molars and posterior to the palatal vacuity and border the inverted U-shaped choanae at the level of the M2–M3 embrasure (fig. 3). The posterior edge of the palatine bears a low postpalatine torus (“tor” in fig. 3). Posterolateral to the torus is the elongate, obliquely oriented minor palatine foramen (“mipf” in fig. 3) for the minor palatine nerve and vessels. Forming the anterior and medial borders of the minor palatine foramen is the palatine; forming the posterior border is the pterygoid (“pt” in fig. 3); and forming the lateral border are the maxilla and pterygoid.

Orbit (figs. 11–13)

Aspects of both orbits are preserved. Between the two sides, a fairly complete reconstruction of the anterior orbit is possible, but only a few features of the posterior orbit are forthcoming.

The flat floor of the orbit, from the weak maxillary tuberosity posteriorly to the maxillary foramen anteriorly, is formed by the alveolar process of the maxilla (figs. 10, 12). Bilaterally preserved are sizeable openings that expose the lingual roots of M2 and M3 (“M2rt” and “M3rt” in fig. 12). The opening for the latter is triangular and roughly symmetric between the two sides, whereas that for the former is more irregular (smaller on the left). Similar openings have been reported in other Late Cretaceous eutherians (e.g., Zalambdalestes, Wible et al., 2004) and extant placentals (e.g., Solenodon, Wible, 2008), and we believe those present in Maelestes are natural. In the floor dorsal to M1 are a half-dozen small foramina alveolaria on both sides. Dorsal to the P5–M1 embrasure is the subcircular, anteriorly directed maxillary foramen (fig. 12), the borders of which are well preserved on the right side and include the lacrimal dorsally and the maxilla ventrally, medially, and laterally; the palatine approximates but falls short of the dorsomedial border.

The lateral part of the orbital process of the lacrimal is preserved on the left side (figs. 11, 12) and the medial part on the right (fig. 13). The orbital process is triangular with a transverse ventral edge contributing to the roof of the maxillary foramen and contacting the maxilla, an oblique medial edge in contact with the palatine and frontal, and an oblique lateral edge forming the orbital margin. It is near the lateral edge on the left side that the two lacrimal foramina occur (figs. 11, 12). The remaining surface of the orbital process is smooth with no indication of a pit or fenestra for the inferior oblique muscle. The left side preserves a small zygomatic process of the lacrimal buttressing the anterior jugal (“zlac” in fig. 12).

The medial wall of the orbit dorsal to the alveolar process of the maxilla is composed of two elements, the orbital processes of the palatine inferiorly and the frontal superiorly, with the former overlapping the latter. The left side (figs. 11, 12) preserves much of the palatine's orbital process, save the anteriormost part, which fortunately is present on the right side (fig. 13). Anteriorly, behind the maxillary foramen, the palatine extends more than halfway up the medial wall (figs. 11, 12). The frontopalatine suture runs posteroventrally such that at the level of M3 the palatine forms only the ventral quarter of the medial wall. On the right side, a broad longitudinal sulcus (“gr” in fig. 13) runs on the ventral aspect of the palatine, just above the maxilla, between the maxillary foramen anteriorly and the sphenopalatine foramen posteriorly; the sulcus is less evident on the left side (“gr” in fig. 12), chiefly as a result of preservation. Behind the palatomaxillary suture on the left side is a depression that includes two foramina: in front, the anteromedially directed sphenopalatine foramen (“spf” in fig. 12) and behind, the ventrally directed minor palatine foramen described with the palate. The sphenopalatine foramen is oval, longer than high, with the maxilla forming its ventrolateral wall and the palatine the remainder. As on the palate, the palatine forms the border for the anterior half of the minor palatine foramen on its dorsal surface; the composition of the posterior half on the dorsal surface is unknown because matrix has been left to buttress the fragile, broken entopterygoid process. Posterior to the minor palatine foramen is an elongate, narrow process of palatine, which extends posterior to the level of the ethmoidal foramen (“ef” in fig. 11) and overlaps what is likely part of the orbitosphenoid (“os” in fig. 11) (see below).

Although the orbital process of the frontal is incompletely known, it is certainly the largest element in the medial orbital wall. Much of the anterior part of the orbital process of the frontal is preserved between the two sides, with only a small, incomplete portion of the posterior part preserved on the left side (figs. 11–13). The frontal contacts the orbital process of the lacrimal anteriorly and of the palatine ventrally, with both bones overlapping the frontal. The frontal's contribution to the supraorbital margin is damaged on both sides. The left frontal preserves two large foramina (fig. 11). The first (“fdv” in figs. 11, 12) lies just below the supraorbital margin at the level of the sphenopalatine foramen, is dorsally directed into the frontal, and has a groove leading into it from below. This opening is in the position of the foramen for the frontal diploic vein of some extant therians (Thewissen, 1989; Wible, 2003). The second, the ethmoidal foramen (“ef” in figs. 11, 12), lies ventral and slightly posterior to the first, is ventrally directed, and has a groove leading into it from below. The borders of the ethmoidal foramen are imperfectly preserved. It appears that the ventral border is formed by a separate bone, an isolated piece of the damaged orbitosphenoid. Behind the frontal diploic foramen, the left frontal extends posteriorly with a roughly boot-shaped form, with the toe pointing ventrally and the back of the boot in the supraorbital margin (fig. 11). The wall of the orbit ventral and posterior to the boot is virtually devoid of bone. If the preserved shape of the frontal is natural, then the presphenoid + orbitosphenoid likely filled the ventral gap and the parietal, squamosal, and alisphenoid lay posterior to the boot. On the toe of the boot is a short, near vertical groove, which tapers slightly ventrally (“vg” in fig. 11). This groove likely contained the ramus supraorbitalis of the stapedial artery and accompanying veins and is somehow associated with the anterior opening of the orbitotemporal canal (e.g., either extending ventrally from that opening or covered by other bones to form that opening).

Different isolated fragments of the presphenoid + orbitosphenoid bone are preserved on each side, with the presphenoid being the ventral base on the midline that continues seamlessly into the orbital wall as the orbitosphenoid. On the left side (fig. 11), an isolated piece of orbitosphenoid forms the ventral border of the ethmoidal foramen and more posteriorly, the orbitosphenoid contacts the palatine and the toe boot of frontal described above. This bar of orbitosphenoid may have served to separate the optic foramen from the sphenorbital fissure, as in other eutherians, but there is no direct indication of either opening. Part of this bar is also preserved on the right side (fig. 13), and ventral to it is a deep concavity in the bone that is open posteriorly. This concavity is likely the anteromedial margin of the sphenorbital fissure (“sof?” in fig. 13), which would have been closed laterally by the alisphenoid. In light of the preserved isolated pieces, the presphenoid + orbitosphenoid is a fairly substantial bone.

Mesocranium (figs. 14, 15)

The mesocranium, the portion of the skull base between the palate and ear region, is imperfectly preserved. At the choanae (“ch” in fig. 14), the palatines form the lateral wall, but the choanal roof is missing. Posterior to the choanae is the basipharyngeal canal (“bpc” in fig. 14), the midline passage containing the nasopharynx that is open ventrally and walled laterally by the base of the ento- and ectopterygoid processes. Most of the roof and part of the lateral wall of the basipharyngeal canal are preserved on the left side, although matrix left to support the delicate entopterygoid process (“enp” in fig. 14) obscures part of the posterior canal. In contrast, most of the right side is either missing or has been distorted dorsomedially and overlies the left side.

Anteriorly in the basipharyngeal canal is a quadrangular segment of the roof (fig. 14). The left side, which includes a small segment of the downturned lip of the lateral wall, appears undistorted; the right side is distorted and has been shifted medially over the left side. The left roof and wall appear to be formed by a single element, presumably the pterygoid, and include small segments of a midline suture near the anterior and posterior borders. The right side is damaged, missing most of the lateral wall and appears to be continuous with the presphenoid in the orbit and to preserve a suture anterolaterally with a small bone fragment that might be the palatine. Posterior to this, the basipharyngeal canal roof on the left side has a sizeable gap, which contains a bone fragment that is probably more pterygoid, and then the left roof is continuous posteriorly to the ear region (fig. 14). Unfortunately, the left side is not completely visible for study, because, as noted above, matrix has been left anterolateral to the carotid foramen (“cf” in fig. 14) to buttress the fragile entopterygoid process. Anterior to the matrix support, the left basipharyngeal canal roof is formed by a single element, which we interpret as the pterygoid, because at its lateral margin it curves ventrally to form the lateral wall of the canal and the base of the ento- and ectopterygoid processes. The posterolateral edge of the pterygoid is concave and presumably underlay the basisphenoid (“bs” in fig. 14), although matrix obscures any overlying bone. Only a narrow segment of the right side of the basispharyngeal canal roof is visible overlying the posterior part of the left pterygoid at the midline. This segment reveals the presence of a sizeable midline crest (“mc” in fig. 14) extending the length of this part of the pterygoid back to the level of the carotid foramen. The crest is damaged, but was likely sharp in life, and has no obvious sutures separating it from the bone behind, the basisphenoid. Zalambdalestes has essentially the same morphology on its basisphenoid (fig. 36), and it was speculated that the crest might be a fused parasphenoid (Wible et al., 2004), which we echo here.

As noted above, preserved on the left side are parts of the lateral wall of the basipharyngeal canal, presumably pterygoid bone, which also serve as the base of the left ento- and ectopterygoid processes (fig. 14). The preserved wall segments are very thin and a longitudinal mass of matrix has been left lateral to the bone for support. At the posteroventromedial aspect of this matrix is a small, oblong piece of bone that is separated from the skull base, ventral to the carotid foramen, and angled from anterolateral to posteromedial. Thin wisps of bone connect this oblong piece anteriorly with the preserved ventral margin of the basipharyngeal canal's lateral wall. We interpret this as the pterygoid hamulus at the posterior end of the entopterygoid process (“ham” in fig. 14). The ectopterygoid process is treated with the basicranium below.

Basicranium (figs. 14–16)

The skull base between the left pterygoid and ear region is formed by a large midline element, the basisphenoid (figs. 14, 15). The left side of the basisphenoid appears fully preserved, but the anterolateral aspect is hidden by matrix supporting the left pterygoid hamulus; the right side is preserved posteriorly and near the midline anteriorly. The only sutural contacts of the basisphenoid that are visible are with the basioccipital posteriorly and with the petrosal posterolaterally (“bo” and “pr” in fig. 15), the former being a straight suture and the latter curved with the basisphenoid the concave member. In the concavity of the petrosal suture, the basisphenoid is at its narrowest. It expands slightly posteriorly at its contact with the basioccipital and expands considerably anteriorly to encompass the area lateral to the carotid foramen. The basisphenoid presumably continues lateral to the carotid foramen as the alisphenoid, but any continuity is hidden by the matrix supporting the pterygoid hamulus.

A midline eminence runs the length of the basisphenoid (fig. 15). The anterior half of the eminence is developed as the broken crest described under Mesocranium above. The posterior half, from the anterior margin of the carotid foramen back, is a low, rounded eminence. Located near the middle of the length of the basisphenoid are the paired carotid foramina. The left foramen is fully preserved; the right preserves only the medial margin. The carotid foramen is oval, longer than wide, and has a broad sulcus on the basisphenoid leading into it from posterolateral. Running posteriorly from the medial side of each carotid foramen is another rounded eminence similar in size to that on the midline. Therefore, the posterior half of the basisphenoid has three longitudinal, rounded eminences, the middle one on the midline; Zalambdalestes (fig. 36; Wible et al., 2004) and Barunlestes (Kielan-Jaworowska and Trofimov, 1980: fig. 2) have the same arrangement.

Lateral to the matrix mass supporting the hamulus is a small exposure of the left alisphenoid, part in the braincase wall (“as” in fig. 15) and part below the braincase as the ectopterygoid process (“ecp” in figs. 11, 16). Only the lateral surface of the posterior base of the ectopterygoid process is preserved (figs. 11, 16). It is vertical, angled somewhat from posterolateral to anteromedial, and dominated by a deep fossa for the lateral pterygoid muscle. Lateral to the ectopterygoid process is a narrow piece of alisphenoid that with the squamosal (“sq” in fig. 15) forms the braincase wall medial to the glenoid fossa (“gf” in fig. 15) and anterior to the ear region. The alisphenoid's contribution to the braincase is smaller than that of the squamosal, except at the piriform fenestra (“pf” in fig. 15) (the long, narrow opening in the skull base between the alisphenoid, squamosal, and petrosal) where a digitiform process of alisphenoid reaches toward the postglenoid foramen. This process is fairly thick, and its posterior surface is rounded and forms the bulk of the anterolateral border of the piriform fenestra, which is completed posterolaterally by the squamosal and petrosal. The preserved alisphenoid provides no evidence for any foramina or canals.

The left squamosal is partly preserved, but enough is present to reconstruct its salient features in the skull base (fig. 15) and in the sidewall of the braincase (fig. 16). The principal features of the former are the glenoid fossa, the posterior zygomatic root, and the postglenoid region and of the latter the squama and the nuchal crest.

Dominating the basicranial surface of the squamosal is the glenoid fossa, which is missing the lateralmost aspect and has several cracks anteriorly (fig. 15). The borders of the glenoid are best delimited posteriorly; medially, it grades into the braincase floor and anteriorly there are two low eminences (see below). As preserved, the glenoid is roughly teardrop shaped, wider than long, with the narrow end directed medially. The bulk of the glenoid is on the posterior zygomatic root; only the point of the tear is on the ventral external surface of the braincase proper. Centrally positioned in and forming most of the posterior wall of the glenoid is the sizeable, tongue-shaped postglenoid process (“pgp” in fig. 15). In lateral view (fig. 11), the postglenoid process projects farther ventrally than any other part of the squamosal or petrosal, and in posterior view (fig. 17), its medial end projects farther than the lateral. Serving as a low preglenoid process are two eminences, medial and lateral, on the squamosal. The medial one is very low, on the braincase proper, opposite the medial end of the postglenoid process; the lateral one is the leading edge of the posterior zygomatic root, which is angled ventrolaterally from the braincase (fig. 15). This angulation appears natural, although there is a crack as wide as the carotid foramen separating the anterior aspect of the posterior zygomatic root from the braincase. The squamosal contributes to the braincase floor anterior and medial to the glenoid fossa (fig. 15). The full extent of its contribution anteriorly and anteromedially is not preserved. Medially, the squamosal contacts the alisphenoid. Posteriorly, the suture with the alisphenoid is just medial to the postglenoid process in the posterolateral border of the piriform fenestra. From there, the suture runs forward and then turns anteromedially parallel to the piriform fenestra such that a small rod of alisphenoid forms the anterior border of the fenestra. Lateral to the posterolateral base of the ectopterygoid process, the suture turns forward a short distance before both bones are missing owing to breakage.

The postglenoid region of the left basicranium is not pristinely preserved (fig. 15); cracks and breakage complicate our reconstruction of the squamosal and petrosal (see below). Nevertheless, we are confident with our ultimate interpretation of the morphological arrangement. The postglenoid region of the squamosal is a very narrow space, in both its length and width, posterior to the medial half of the glenoid fossa. The most conspicuous feature is the postglenoid foramen (“pgf” in fig. 15), which lies posteromedial to the medial aspect of the postglenoid process. In fact, a short sulcus on the posteromedial aspect of the postglenoid process accommodated contents of the postglenoid foramen. The postglenoid foramen is ovoid and slightly obliquely set with its long axis directed from posteromedial to anterolateral. The medial and lateral borders of the postglenoid foramen are formed by narrow rods of bone that serve as borders to other spaces; the medial rod forms the lateral border to the epitympanic recess and fossa incudis (“er” and “fi” in fig. 15; described with the petrosal below) and the lateral rod forms the ventromedial border to the suprameatal foramen (“smf” in fig. 15). The medial rod continues dorsally as a complete wall separating the postglenoid foramen and fossa incudis. However, the lateral rod has a gap dorsal to it and the postglenoid and suprameatal foramina are continuous through that gap; extant didelphids have the same relationship between the two foramina (Wible, 2003). Posterior to the postglenoid foramen, the medial and lateral rods meet to form a short trunk that posteriorly abuts the petrosal bone.

In left lateral view (figs. 11, 16), the squamosal as preserved is not a large element, extending only from the nuchal crest (“nc” in fig. 16) at the occipital margin to just in front of the posterior zygomatic root. However, the small size is the result of breakage, because the entire dorsal and anterior margins are missing; these might have doubled the size of the bone. Also missing is the central section of the preserved element that would have been exposed on the braincase wall. This damage conveniently divides the squamosal into two parts for descriptive purposes, the squama anteriorly and the caudal process posteriorly, which meet at the suprameatal foramen.

The squama of the squamosal is the flattened portion of the bone forming the sidewall of the braincase. In left lateral view (fig. 11), the preserved squama is small, only extending slightly anterior and dorsal to the posterior zygomatic root, but it was likely much larger in life. In addition to probable damage to the periphery of the squama, there are two large, artificial gaps where that element meets the dorsal aspect of the posterior zygomatic root. The surface of the squama is featureless except at the preserved posterodorsal margin, where there is a short, low oblique crest directed posterodorsally. This crest represents the anterior part of the suprameatal bridge; the posterior part is on the caudal process (see below), and the central part is missing.

The caudal process of the left squamosal is a roughly J-shaped element situated on the posteroventral margin of the sidewall of the braincase (figs. 11, 16), but only its posterior and ventral borders are natural ones. The posterior border forms the ventrolateral margin of the nuchal crest (fig. 16), abuts the mastoid exposure of the petrosal (“me” in fig. 16), and is visible on the occiput (fig. 17), and the ventral border is slightly irregular and contacts the petrosal (fig. 16). The posterodorsal corner of the posterior border appears to be natural (figs. 11, 16), which requires the element completing the braincase above the squamosal (the parietal and perhaps interparietal) to be exceptionally large. Situated in the middle of the ventral third of the caudal process is an oval foramen for the ramus temporalis (“rt” in figs. 11, 16), longer than high, that is directed laterally and slightly dorsally. Forming the ventral margin of the foramen is the posterior part of the low suprameatal bridge. Between the caudal process and the underlying pars canalicularis of the petrosal bone is a large gap filled with matrix that represents the posttemporal canal (see below).

As noted, the part of the left squamosal connecting the squama and caudal process is missing. The ventralmost part of this missing piece would form the dorsolateral border of the suprameatal foramen and contacts the anterodorsal border of the petrosal (fig. 16). It would also serve as the lateral wall of a short, oblique canal (called the postglenoid canal here) leading to the suprameatal foramen from above and behind, transmitting vessels from the endocranium to the suprameatal and postglenoid foramina. The medial wall of this canal is preserved and appears to be formed by petrosal. The anterodorsal border of the petrosal may have made a minor contribution to the posterior wall of the postglenoid canal or it may have been excluded by the missing squamosal.

On the central basicranium anterior to the foramen magnum in mammals are the unpaired basioccipital anteriorly and paired exoccipital bones posterolaterally (“bo” and “eo” in figs. 14, 15). There is no indication of a suture between the basioccipital and exoccipitals in Maelestes, but we describe them individually based on their usual positions in mammals. The usual position for the suture between the basioccipital and exoccipital is from the jugular foramen (“jf” in fig. 15) to the intercondyloid (odontoid) notch (“icn” in fig. 15), with the hypoglossal foramina (“hf” in fig. 15) mostly or entirely within the exoccipitals (see Wible, 2003, 2007; Wible and Gaudin, 2004; Giannini et al., 2006).

As preserved in ventral view (fig. 15), the basioccipital and exoccipitals have the shape of a funnel with the constricted end pointing anteriorly. However, the extreme constriction of the anterior part is artificial, caused by the ventromedial displacement of both the right and left petrosals and dorsal displacement of the basioccipital. Based on the CT scans, these two bones are of similar width at the basisphenoid-basioccipital suture, but as they are preserved on the skull base, the latter is roughly half the width of the former. The petrosals are covering half the basioccipital in roughly the anterior third of the basioccipital bone's midline length. On the left side, only the anterior half of the suture between the basioccipital and petrosal is deformed, whereas on the right side the entire suture is deformed. The exoccipital-petrosal suture is intact on the left side, posterior to the jugular foramen.

The basioccipital is flat anteriorly and has a low, inverted V-shaped crest occupying the middle third of the midline (fig. 15). On either side of the midline crest are very shallow depressions, which house the rectus capitis ventralis muscles in the dog (Evans, 1993). The posterior third of the midline is flat, but on either side the exoccipital bones are gently curved posteroventrally ending in the occipital condyles (“oc” in figs. 14, 15), such that the ventral surface of the condyles lies ventral to the midline; this is most easily seen in occipital view (fig. 17). Additionally, the condyles extend farther posteriorly than the midline, being separated by a deep intercondyloid notch (fig. 15). The condyles are roughly teardrop shaped with the narrow end anteroventromedially and the bulk of the tear obliquely set posterodorsolaterally; this is also most easily seen in occipital view with the left condyle (fig. 17), because the right is slightly damaged. The surface anterior to the condyles is smooth, with no ventral condyloid fossa (fig. 15). Anterior to the central part of the condyle is a large, round opening, the hypoglossal foramen. The right foramen is significantly larger than the left, and this may have been a natural condition because the borders do not appear to be damaged or deformed. Within the right foramen, three much smaller foramina are clearly visible, positioned anteromedially, posteromedially, and anterolaterally. The left foramen, which has not been as fully prepared, only exposes an anteromedial opening (see Endocranium). Lateral to the hypoglossal foramen, the left exoccipital has a low, digitiform process (“pcp” in fig. 15) that abuts the much larger medial end of the caudal tympanic process of the petrosal (“ctpp” in fig. 15). The process on the left exoccipital is a small paracondylar (jugular) process; it is broken on the right side. Anteromedial to the paracondylar process, the exoccipital (or possibly basioccipital) forms the posteromedial border of the jugular foramen, the remaining borders of which are formed by the petrosal. The jugular foramen, well preserved on the left side, is fusiform with the long axis slightly oblique, anteromedially directed. The jugular foramen is small with a total area that is subequal to that of the left hypoglossal foramen.

The petrosal is the most complex element of the basicranium, housing the organs of hearing and equilibration, as well as structures of the middle and external ear, and being crossed by various major nerves and vessels. The petrosal is divisible into two parts: the pars cochlearis, anteroventrally placed and housing the cochlea, and the pars canalicularis, posterodorsally placed and housing the vestibule and semicircular canals. In addition, the petrosal can be visualized as a tetrahedron with the following four sides (MacIntyre, 1972): tympanic, facing the middle ear; cerebellar, within the cranial cavity; squamosal, covered by the squamosal bone; and mastoid, on the occiput.

In Maelestes, as noted above, both petrosals have been shifted ventromedially such that the anteromedial aspect of each bone underlies the lateral parts of the basioccipital (fig. 15). The left petrosal is nearly complete, only missing a little bone from its posterodorsal apex; missing from the right petrosal is a significant part of the pars canalicularis. All four surfaces of the left petrosal are visible, although part of the squamosal surface is covered by the undamaged portion of the squamosal bone; only the tympanic surface of the right petrosal is visible and then not fully because the anteriormost part is covered by matrix left to support the fragmentary ectotympanic bone (“ec” in fig. 15). Described here are the tympanic and squamosal surfaces; the mastoid surface is described with the occiput and the cerebellar with the endocranium.

In tympanic view, the principal feature of the pars cochlearis is the ovoid promontorium (“pr” in fig. 15), the tympanic surface of which reflects the enclosed coiled cochlea. Based on the CT scans and following the protocol of West (1985) for measuring the number of turns in the cochlear spiral, the cochlea of Maelestes is coiled 360°. There are posterior and posterolateral openings into the promontorium: the former is the aperture of the cochlear fossula, which leads to the fenestra cochleae or round window (“fc” in fig. 15), and the latter the fenestra vestibuli or oval window (“fv” in fig. 15). The aperture of the cochlear fossula is not fully visible in direct ventral view as it is in the vertical posterior wall of the promontorium. The opening is directed posterolaterally and is oval, wider than high, with the narrower end facing medially. The cochlear fossula is a small depression in the roof internal to the aperture and immediately external to the fenestra cochleae, to which the secondary tympanic membrane attached in life. The fenestra vestibuli, accommodating the footplate of the stapes in life, opens in the sloped posterolateral wall of the promontorium and, therefore, is more fully visible in direct ventral view. The opening is directed anteroventrolaterally and is elliptical, with the long axis obliquely set from posterolateral to anteromedial. The well-preserved left fenestra vestibuli has a stapedial ratio (length/width, see Segall, 1970) of 1.8; the right opening was not measured because its lateral rim is damaged. The posterior rim of the left fenestra vestibuli is recessed from the promontorial surface, creating a shallow vestibular fossula.

In addition to reflecting the coiled cochlea, the tympanic surface of the promontorium is crossed by two faint vascular grooves, present on both the left and right sides (fig. 15). The longer of the two (“gica” in fig. 15) begins at the posteromedial corner, in front of the ventromedial edge of the aperture of the cochlear fossula. It runs straight anterolaterally to a point just lateral to the greatest prominence of the promontorium. There it bends anteromedially and continues straight to the anteromedial pole of the promontorium. This groove would accommodate the transpromontorial internal carotid artery (see Wible, 1986); the groove is faintest in its posteriormost segment, and the remainder is of uniform size. The second groove (“gsa” in fig. 15) would accommodate a stapedial artery (see Wible, 1987); the groove is similar in size to that for the internal carotid and diverges from the posterior aspect of the internal carotid groove in front of the aperture of the cochlear fossula. It runs straight laterally and slightly anteriorly to the ventromedial rim of the fenestra vestibuli, which it notches slightly posterior to the midpoint in a position consistent with a bicrurate stapes having an intracrural foramen.

A horizontal shelf extends anteriorly and medially from the promontorium (“ew” and “mfl” in fig. 15). Although the shelf is continuous, for descriptive purposes, it is treated and named in anterior and medial parts. The larger anterior part is equivalent to the epitympanic wing of the petrosal of MacPhee (1981: fig. 3), and the medial part is termed medial flange here. Visible only on the left side, the epitympanic wing is roughly triangular, coming to a point anteriorly. Its anteromedial edge, which underlies the basisphenoid, is not entirely flat, but is curved slightly ventrally. This curvature is caused, at least in part, by the ventromedial displacement of the petrosal. The deepest curvature appears to mark the lateral edge of a very faint groove that would have transmitted the internal carotid artery from the promontorium to the basisphenoid. The shelf lateral to the groove likely served as attachment area for the tensor tympani muscle, which appears to extend posterolaterally into a faint depression on the promontorium. The anterolateral edge of the triangular shelf and promontorium behind it form the medial border of the piriform fenestra. Running near the anterolateral edge of the triangular shelf and extending posteriorly onto the promontorium toward the primary facial foramen (see below) is a narrow groove (“gpn” in fig. 15), which would have transmitted the greater petrosal nerve.

The medial flange is preserved on both sides, but more fully on the left (fig. 15). It extends from the basisphenoid-basioccipital juncture to the jugular foramen. As noted above, it underlies the basioccipital just behind the basisphenoid, because of the ventromedial displacement of the petrosal. It is widest in its anterior half. Posteriorly, it forms the anteromedial border of the jugular foramen. Although the medial flange is continuous with the epitympanic wing, it narrows in the vicinity of the basisphenoid-basioccipital juncture. We believe this narrowest part of the medial flange to be artificial, having been broken by the displacement of the petrosal.

Extending from the posterior and posterolateral margin of the promontorium are the visible aspects of the pars canalicularis, best preserved on the left side (fig. 15). Posterior to the promontorium are two depressions, medial and lateral, bordered posteriorly by a broad, low-lying wall, the caudal tympanic process of the petrosal. The medial depression, the post-promontorial tympanic sinus (“pps” in fig. 15) is smaller and not as deep; it is a flat, ovoid surface (wider than long) behind the aperture of the cochlear fossula, which would have accommodated a diverticulum of the cavum tympani, the air-filled sac occupying the middle ear. The long axis of the sinus is in the same oblique plane as the long axis of the aperture of the cochlear fossula. The medial edge of the sinus forms the lateral border of the jugular foramen. The lateral depression (“sf” in fig. 15), for the stapedius muscle, is similar in shape and orientation to the post-promontorial tympanic sinus, but is roughly twice the size and considerably deeper. Separating the sinus and stapedius fossa is a thin, low, longitudinal crest that connects the crista interfenestralis (“cif” in fig. 15) between the aperture of the cochlear fossula and fenestra vestibuli on the back of the promontorium and the caudal tympanic process.

The caudal tympanic process is most prominent at its medial and lateral ends (fig. 15). At the medial end, behind the post-promontorial tympanic sinus, the process is thickened and appears to be a lower-lying version of the tympanic process of Kielan-Jaworowska (1981), which occurs in asioryctitheres and zalambdalestids (“typ” in fig. 36) as well as in other extinct and extant eutherians. The caudal tympanic process abuts the lower lying paracondylar process of the exoccipital medially (fig. 15). At the lateral end, the caudal tympanic process curves anteriorly as the crista parotica (“cp” in fig. 15; see below). Just lateral to this inflexion is a well-developed process that is rod-shaped in ventral view (“ppr” in fig. 15). The anteromedial end of the rod is continuous with the crista parotica and the posteromedial end is the terminus of the nuchal crest. If this prominence were on the squamosal, it would be a posttympanic process; however, it appears to be on the petrosal, and, therefore, following Wible et al. (2004), it is a paroccipital process ( =  the lateral section of the caudal tympanic process of MacPhee, 1981, or the mastoid process of Kielan-Jaworowska, 1981).

The caudal tympanic process forms the posterior wall of the middle ear, but it does not form the posterior limit of the basicranium in ventral view (fig. 15). Behind the caudal tympanic process is a posterodorsally sloping surface of bone, formed by the exoccipital medially and by the mastoid exposure of the petrosal laterally (“me” in fig. 15). At the midpoint of the posterior edge of this surface is a large protuberance on the petrosal formed by the juncture of the lateral and posterior semicircular canals.

Extending laterally from the posterolateral half of the left promontorium, anterior to the lateral half of the stapedius fossa, is a horizontal, rectangular shelf, longer than wide (largely hidden by “cp” in fig. 15). At the point where this shelf meets the promontorium anteriorly, there is a small, oval, laterally directed foramen in the promontorial wall. This foramen is nearly completely hidden from view on the left side; however, because there is no bone preserved lateral to the promontorium on the right side, this foramen is fully exposed (“pff” in fig. 15). This opening is the primary facial foramen, which would have transmitted the facial nerve from the internal acoustic meatus in the cranial cavity (see below) into the middle ear. Within the middle ear, the facial nerve would have bulged into the geniculate ganglion, out of which arose two nerves: the anteriorly directed greater petrosal nerve and the posteriorly directed continuation of the facial nerve. The former would have occupied the groove on the anterolateral edge of the promontorium and epitympanic wing described above and the latter a shallow groove dorsal to the fenestra vestibuli visible on the right side. This arrangement of the facial nerve is unusual among extant therians, which typically have the primary facial foramen opening into a separate bony space housing the geniculate ganglion, the cavum supracochleare of Voit (1909), which leads to two foramina: the hiatus Fallopii for the greater petrosal nerve and the secondary facial foramen for the facial nerve (Wible, 1990; Wible and Hopson, 1993). There is the possibility that the bone enclosing the cavum supracochleare has not been preserved in Maelestes, but we regard this unlikely because the surfaces around the right primary facial foramen appear unbroken.

At the lateral margin of the horizontal, rectangular shelf, there is a prominent, ventrally directed wall of bone that initially is vertical, but then tilts medially toward, but not achieving, the promontorium (“cp” in fig. 15). (In the anterior part of the vertical component of this wall is a small opening, the prootic canal; see Endocranium.) In lateral view, the ventral edge of this wall (“cp” in fig. 16), which lies between the postglenoid and paroccipital processes, obscures the promontorium. The anterior three-fourths of this ventral edge is a prominent thick crest; the posterior one-fourth is thinner and lower. This ventral edge represents the crista parotica, which, as mentioned above, contacts the caudal tympanic process posteriorly. Opposite the back of the fenestra vestibuli, a small process extends posterodorsomedially from the crista parotica (“th” in fig. 15); this is the tympanohyal, the attachment point of Reichert's cartilage. Immediately behind the tympanohyal, the medial aspect of the crista parotica is notched; this is the stylomastoid notch, by which the facial nerve exits the middle ear.

Lateral to the crista parotica is a flat, quadrangular surface of bone tilted slightly dorsolaterally that contributes to the roof of the external acoustic meatus (“eam” in fig. 15). In Zalambdalestes (Wible et al., 2004) and most extant placentals (Kampen, 1905; Klaauw, 1931), it is the squamosal that forms the roof of the external acoustic meatus (“sq eam” in fig. 36). As noted above, the postglenoid region of Maelestes has several cracks that complicate reconstruction. Nevertheless, it appears that the quadrangular bone in the meatal roof is petrosal. First, the bone in question is continuous with the crista parotica and tympanohyal medially and appears continuous with the caudal tympanic and paroccipital processes posteromedially; second, what we interpret as sutures (and not cracks) separate the anterior and posterolateral aspects of the bone from the squamosal. The lateral edge of the bone appears natural and would have contacted the missing squamosal posterior to the suprameatal foramen. Asioryctes and Kennalestes have a quadrangular piece of petrosal in the same position as in Maelestes, posterior to the postglenoid foramen (“pet eam” in fig. 36; labeled “tympanohyal” in Kielan-Jaworowska, 1981: figs. 3, 6, 7). Petrosal in the same position also occurs in an isolated petrosal referred to the early Cretaceous eutherian Prokennalestes Kielan-Jaworowska and Dashzeveg, 1989 (Wible et al., 2001) and in more basal taxa, such as Early Cretaceous Vincelestes Bonaparte, 1986 (Rougier et al., 1992; Rougier, 1993).

Anterolateral to the crista parotica and medial to the postglenoid foramen is a narrow, almost dumbbell-shaped depression that is obliquely oriented, posterolaterally to anteromedially (“fi” and “er” in fig. 15). This depression is walled laterally by the squamosal, but only the posterior half of the medial side has a wall. The incomplete medial wall and entire floor appear to be petrosal, which represents the tegmen tympani in light of the position anterior to the crista parotica (De Beer, 1929, 1937). We identify the posterior half of this depression as the fossa incudis, which accommodated the crus breve (short process) of the incus, and the anterior half as the epitympanic recess, the space over the mallear-incudal articulation. At the anteromedial aspect of the epitympanic recess, the petrosal has a narrow contact with the alisphenoid. The epitympanic recess forms the posterolateral border of the piriform fenestra.

Preserved on the right side separated from the anterior promontorium by a sizeable amount of matrix is a small piece of bone that represents the anterior part of the right ectotympanic bone (“ec” in fig. 15). The preserved piece is crescentic and flat, with the ends of the crescent representing the broken anterior and posterior crura.

Occiput (figs. 16, 17)

Only the ventral part of the occiput is preserved and more so on the left side; the right squamosal and most of mastoid exposure of the right petrosal are missing (fig. 17). The principal preserved element of the occiput is the centrally positioned paired exoccipital bone. Two small pieces of the midline supraoccipital bone (“so” in fig. 17) that lie dorsal to the exoccipital on either side of the foramen magnum (“fm” in fig. 17) are also preserved, with the piece on the right side more substantial than on the left. Beginning at the foramen magnum, the suture between the right exoccipital and supraoccipital runs laterally with the former element convex and the latter concave and then it turns dorsolaterally. The suture between the exoccipital medially and the other major element of the occiput laterally, the mastoid exposure of the petrosal, is preserved bilaterally, but is more complete on the left side. The ventral two-thirds of the suture roughly follow the shape of the ventrolateral border of the foramen magnum, with the exoccipital convex and the petrosal concave. The dorsal one-third turns dorsolaterally at a near right angle and is straight. Lateral to the left petrosal is a narrow exposure of squamosal contributing to the ventral part of the nuchal crest.

The principal feature of the occiput is the foramen magnum (fig. 17). The ventral and lateral borders are preserved and formed by the exoccipital bone with a small contribution from the supraoccipital to the dorsolateral border; the missing narrow dorsal border would have been completed by the supraoccipital. The foramen appears to be teardrop shaped, but this could be altered by the missing dorsal border. In the ventrolateral margins of the foramen are the occipital condyles, with the left one best preserved. As noted with the basicranium, the condyle has the shape of an oblique teardrop, with the narrow end anteroventromedially directed. The dorsal, and even more so the lateral, margins of the left condyle are well demarcated from the adjacent bone, but there is no true condyloid fossa on the occiput. The exoccipital bone dorsal to the condyle is not flat, but is obliquely set with the medial border more posterior than the lateral.

The mastoid exposure of the left petrosal is well preserved, only missing some bone from its dorsalmost margin (figs. 16, 17). It is roughly triangular, with the medial border contacting the exoccipital, the lateral the squamosal, and the ventral forming the posterior wall of the middle ear. The occipital surface of the mastoid exposure is not flat. Just lateral to the suture with the exoccipital is a sizeable vertical eminence, which reflects the underlying posterior semicircular canal (“psc” in fig. 17). The mastoid surface lateral to this eminence is obliquely set with the lateral margin more anterior than the medial. Near the ventral limit of and at a right angle to the eminence over the posterior semicircular canal is a less prominent horizontal eminence, which reflects the underlying lateral semicircular canal (“lsc” in fig. 17). Dorsal and parallel to this is a deep vascular groove that leads into a round, anteriorly directed posttemporal foramen into the posttemporal canal (“ptc” in fig. 17) entirely within the petrosal, near the squamosal suture. The bulk of the ventral border of the mastoid exposure is formed by the caudal tympanic process of the petrosal. At the ventrolateral corner is a low eminence formed entirely by the petrosal, the paroccipital process.

Endocranium (figs. 10, 18)

Damage to the specimen's braincase combined with preparation has exposed some endocranial surfaces of the basisphenoid, basioccipital, and the entire left petrosal.

Only a small rectangular area of the basisphenoid (wider than long) anterior to the basioccipital bone, but posterior to the carotid foramen, is exposed (fig. 18). It is highest posteriorly and laterally and slopes gently anteriorly and medially. The sloped surface probably is the posterior part of a shallow hypophyseal fossa (“fh” in fig. 18) and the posterior border is the very low dorsum sellae; posterior clinoid processes are not present.

Approximately the left half of the endocranial surface of the basioccipital bone's basicranial portion is exposed (fig. 18). At its anterior end, the basioccipital is higher than the basisphenoid, but this is probably due to distortion, with the anterior petrosal displaced ventromedially and the anterior occipital dorsally. Most of the left endocranial surface of the basioccipital is flat and featureless. Posterolaterally, on the exoccipital is a shallow depression with a single hypoglossal foramen within it (not visible in the figures). Posteriorly from there to the occipital condyle, the exoccipital is gently depressed with no indication of other hypoglossal foramina or a condyloid canal. The surface lateral to the single hypoglossal foramen has not been prepared, but no other openings here are seen on the CT scans. The right side of the exoccipital was prepared in the area dorsal to its hypoglossal foramen, and the three openings observed on the ventral surface were also visible dorsally (not visible in the figures). Consequently, the asymmetry in number of hypoglossal foramina suggested by the ventral surface is confirmed by the dorsal.

The endocranial surface of the left petrosal is well preserved and well exposed (fig. 18). It can be visualized as two subequal parts, each surrounding a large opening, at nearly a right angle to each other. Anteroventromedially is the pars cochlearis around the internal acoustic meatus (“fai” and “fas” in fig. 18), which transmits the facial and vestibulocochlear nerves, and posterodorsolaterally is the pars canalicularis around the subarcuate fossa (“saf” in fig. 18), which houses the paraflocculus of the cerebellum.

The internal acoustic meatus is bow-tie shaped; it is not situated centrally on the pars cochlearis but somewhat posterolaterally. The medial part of the bow is the foramen acusticum inferius and the lateral part the foramen acusticum superius (fig. 18); the knot of the bow tie is the transverse crest, which is lowest at its midpoint (though still higher than the neighboring foramina) and sloping to the periphery from there. The inferior and superior foramen could not be fully prepared because of their depth and the CT scans did not provide additional details. Based on extant therians (Wible, 2003, 2008; Wible and Gaudin, 2004; Giannini et al., 2006), the inferior foramen should contain tiny perforations for the spiral cribriform tract of the cochlear nerve and perhaps posteriorly a separate foramen singulare for part of the vestibular nerve; the superior foramen should contain an anterior opening for the facial nerve into the facial canal, which ends at the primary facial foramen, and a posterior pit, the cribriform dorsal vestibular area, for the remaining part of the vestibular nerve. The lateral wall of the superior foramen is the prefacial commissure (“pfc” in fig. 18), which is mediolaterally thin and vertically shallow. The dorsal edge of the prefacial commissure is the lowest part of a crista petrosa (“crp” in fig. 18), which extends from the anterior pole of the pars cochlearis to the lateral side of the subarcuate fossa. At the anterior pole, the crista petrosa is at its broadest with its anterior face covered with matrix, so that the full extent of this ridge is unclear. Lateral to the subarcuate fossa, the crista petrosa is a sharp, high ridge. The surface of the pars cochlearis anterior and medial to the internal acoustic meatus is smooth. Its medialmost edge, which might have evidence of the position of the inferior petrosal sinus, is not prepared. Anterodorsolateral to the jugular foramen is a small, shallow pit that probably represents the cochlear canaliculus for the perilymphatic duct (not visible in the figures).

The subarcuate fossa is centrally positioned on the endocranial surface of the pars canalicularis; it is large, elliptical (taller than wide), and deep (based on the CT scans as its floor has not been fully prepared). The aperture into the fossa is constricted, especially laterally, compared to the space invading the pars canalicularis (based on the CT scans). The dorsal rim of the aperture is formed by the anterior semicircular canal (“asc” in fig. 18); the medial rim by the crus commune (not visible in the figures), which is the conjoined anterior and posterior semicircular canals; and the lateral rim is formed by the prominent crista petrosa (fig. 18). Lateral to the crista petrosa are visible the continuation of the anterior semicircular canal and the anterior ampulla (“aa” in fig. 18), a dilation where the anterior canal meets the vestibule. The dorsal surface of the crus commune has a narrow, short groove that runs anteroventrally into a small opening, the vestibular aqueduct for the endolymphatic duct (not visible in the figures).

The bony surfaces adjacent to the rim of the subarcuate fossa are either covered with matrix (medially) or damaged (dorsally). The bony surface immediately posterior and parallel to the lateralmost anterior semicircular canal and anterior ampulla bears a well-developed vascular groove for a large vein, the prootic sinus (“ps” in figs. 11, 16, 18). The ventral end of the prootic sinus groove is on the posterolateral aspect of a small, quadrangular horizontal shelf, which connects the main body of the pars canalicularis with the part of the petrosal in the roof of the external acoustic meatus. Here the prootic sinus meets three other vascular structures: the posttemporal canal, running posteriorly to the occiput between the pars canalicularis and overlying caudal process of the squamosal; the postglenoid canal, running anteroventrally to the suprameatal and postglenoid foramina; and the diminutive prootic canal (“pc” in figs. 11, 16, 18), running medially into the middle ear. There is a small crest running from the anteroventral aspect of the prootic sinus groove to the dorsal aspect of the lateral prootic canal opening that directs some of the prootic sinus into the prootic canal. For more detail about these vessels, see the Basicranial Vascular Reconstruction.

As noted above, the posterior braincase roof is entirely missing. However, some matrix within the posterior braincase is preserved. At the posterior limit of the preserved matrix is a roughly trapezoidal area, the posterior and longest border of which lies dorsal to the top of the foramen magnum. This surface is smoother and darker than the surrounding matrix and represents an endocast of part of the brain (“ver” in fig. 10). Based on comparisons with endocasts of other Mongolian Late Cretaceous eutherians (Kielan-Jaworowska, 1984b; Kielan-Jaworowska and Trofimov, 1986), this is an endocast of the vermis of the cerebellum.

Basicranial Vascular Reconstruction

Various foramina, grooves, and canals associated with the basicranial vasculature have been described above. We provide a comprehensive account here to aid the reader in understanding this system, which has provided numerous characters in previous phylogenetic analyses (e.g., Rougier et al., 1998; Luo and Wible, 2005) as well as ours (see appendix 2). The bases for our restoration and the terminology employed are our own published and unpublished studies of the basicranial vessels in extant mammals (e.g., Wible, 1984, 1986, 1987, 1990, 2003, 2008; Rougier et al., 1992; Wible and Hopson, 1995; Rougier and Wible, 2006).

The basicranial venous system varies considerably in extant mammals (Gelderen, 1924; Wible, 1990; Rougier et al., 1992; Wible and Hopson, 1995; Rougier and Wible, 2006). The principal conduits in extant placentals are the postglenoid foramen, the jugular foramen, and foramen magnum, with secondary conduits including the mastoid foramen, condyloid canal, posttemporal canal, suprameatal foramen, foramen for ramus temporalis, and foramen for the inferior petrosal sinus (Sisson, 1910; Evans, 1993; Wible and Gaudin, 2004; Giannini et al., 2006; Wible, 2008). Of these, Maelestes lacks only a condyloid canal and foramen for the inferior petrosal sinus. Although the latter opening was absent, the sinus itself must have been present in light of the universal incidence of this vessel among extant therians (see more below). The incidence of a mastoid foramen cannot be excluded, because the petrosal, exoccipital, supraoccipital juncture is not fully preserved. Additionally, Maelestes has a venous channel not present in extant placentals, a prootic canal (“pc” in figs. 11, 16, 18). Figure 19 illustrates our reconstruction of the major basicranial vessels in Maelestes.

Maelestes has evidence for only one dural sinus in the form of the groove running lateral to the subarcuate fossa on the endocranial surface of the petrosal (“ps” in figs. 11, 16, 18). Naming this venous channel is not straightforward, because different vessels with different ontogenies serve as the anterior distributary of the transverse sinus in extant mammals (Gelderen, 1924; Wible, 1990; Wible and Hopson, 1995; Rougier and Wible, 2006). In monotremes and marsupials, it is the prootic sinus (the middle cerebral vein), which exits the skull differently in the two clades. In monotremes, the prootic sinus leaves via a well-developed prootic canal and joins the lateral head vein in the middle ear; in marsupials, the prootic canal (and lateral head vein) is reduced or absent (Wible, 1990, 2003; Sánchez-Villagra and Wible, 2002), with the primary exit for the prootic sinus through the postglenoid foramen via a neomorphic addition, the sphenoparietal emissary vein of Gelderen (1924). In placentals (with the exception of the extant solenodon, see below), the entire prootic sinus is replaced by a neomorphic vessel, the capsuloparietal emissary vein of Gelderen (1924), which exits the postglenoid foramen. For Maelestes, we identify the vein in question as a prootic sinus, because the presence of a reduced prootic canal along with a postglenoid foramen indicates a marsupial-like venous pattern, as has been described by some of us (Wible et al., 2001) in the Early Cretaceous eutherian Prokennalestes and in isolated petrosals allocated to Late Cretaceous zhelestids (Ekdale et al., 2004). Wible (2008) recently has documented the existence of a reduced prootic canal and lateral head vein in the Hispanolan solenodon, the first report of this structure in a placental, extinct or extant.

Some justification for our identification of a prootic canal in Maelestes is warranted, because another vessel, the ramus superior of the stapedial artery, runs through the same general area in some extant placentals (MacPhee, 1981; Wible, 1987, 2008). Moreover, in Prokennalestes, the prootic canal and the canal for ramus superior are contiguous, the posterior and anterior ends of a single dumbbell-shaped canal (Wible et al., 2001). Regarding the canal in question in Maelestes, given its extremely small size, we believe that it transmitted only one structure. Nevertheless, the ramus superior cannot be excluded as that sole occupant. However, justifying our identification of a prootic canal in Maelestes is the low crest connecting the endocranial aperture with the groove for the prootic sinus. Such a connection is to be expected between the prootic sinus and prootic canal, but not between the prootic sinus and ramus superior.

From the tympanic aperture of the prootic canal, we reconstruct a small lateral head vein (“lhv” in fig. 19) passing through the middle ear to join the internal jugular vein (“ijv” in fig. 19) below the jugular foramen as in extant monotremes and marsupials (Wible and Hopson, 1995), and the solenodon (Wible, 2008). However, the prootic canal was not the principal exit for the prootic sinus in Maelestes. The bulk of the venous blood in the prootic sinus entered the postglenoid canal between the squamosal and petrosal to exit the postglenoid foramen (“pgv” in fig. 19). As noted above, the vessel in the postglenoid foramen is termed the capsuloparietal emissary vein in extant placentals and the sphenoparietal emissary vein in extant marsupials, reflecting different ontogenetic origins. We use the noncommittal term postglenoid vein to identify this vessel in Maelestes, because we do not know its ontogeny. Other well-developed conduits of the prootic sinus are via the suprameatal foramen (“vt” in fig. 19), foramen for ramus temporalis, and posttemporal canal (“vdm” in fig. 19). In light of the pattern in extant mammals (Wible and Hopson, 1995; Wible, 2003; Wible and Gaudin, 2004), it is most likely that the first two passed blood from the temporalis muscle into the prootic sinus, but the direction of blood flow in the last, the vena diploëtica magna, is more uncertain.

The area on the endocranial surface of the petrosals of Maelestes where a groove would be expected for the sigmoid sinus, the posterior distributary of the transverse sinus in extant mammals, is covered with matrix (fig. 18) and the CT scans do not provide sufficient detail. Therefore, it is uncertain whether the sigmoid sinus left the skull via the foramen magnum as in monotremes and marsupials or via the jugular foramen as in most placentals (Wible, 1990). However, given the remarkably small size of the jugular foramen in Maelestes (fig. 15), which is subequal to the fenestra cochleae as in extant marsupials (Rougier et al., 1998), it seems likely that the principal exit of the sigmoid sinus was into the vertebral veins via the foramen magnum.

Also covered with matrix and distorted by the displacement of the petrosals and occipital in Maelestes is the area where evidence would be expected of the inferior petrosal sinus (figs. 15, 18), which drains posteriorly from the cavernous sinus around the hypophysis in extant mammals. Based on the CT scans, the most likely course for the inferior petrosal sinus (“ips” in fig. 19) is on the dorsal (endocranial) surface of the medial flange of the petrosal with an exit via the jugular foramen into the internal jugular vein. However, given the small size of the jugular foramen, which also transmitted the glossopharyngeal, vagus and cranial accessory nerves, this vein was probably not a major drainage route.

The principal basicranial arteries in extant placentals include the internal carotid, which supplies the brain, and the stapedial artery, the primary extracranial branch of the internal carotid, which sends superior and inferior rami that are distributed with the trigeminal nerve system (Tandler, 1899, 1901; Bugge, 1974, 1978, 1979; Wible, 1984, 1987). In extant mammals, as noted by Wible (1986), the internal carotid artery (with accompanying vein and sympathetic nerve) follows one of three pathways en route to the cranial cavity beneath the basicranium: (1) on the promontorium or transpromontorial; (2) through the substance of the tympanic wall or perbullar; or (3) medial to the tympanic wall or extrabullar. The presence in Maelestes of a groove traversing the promontorium aimed at the carotid foramen within the basisphenoid (fig. 15) unequivocally supports reconstruction of a transpromontorial internal carotid artery at some point in the animal's life (“ica” in fig. 19). An option is that the internal carotid artery involutes during ontogeny and the internal carotid nerve is the sole occupant of the transpromontorial groove in the adult as reported by Conroy and Wible (1978) in the lemur Varecia variegata (Kerr, 1792).

As noted by Wible (1987), the only structure in extant placentals to occupy a groove on the promontorium aimed at the fenestra vestibuli is the stapedial artery. Moreover, this artery invariably penetrates the intracrural foramen in the stapes in extant placentals (Novacek and Wyss, 1986; Wible, 1987). Maelestes has a groove originating from the transpromontorial carotid groove and directed at the fenestra vestibuli (fig. 15), which unequivocally supports reconstruction of a stapedial artery (“sa” in fig. 19) arising from the internal carotid within the middle ear. The stapedial groove notches the fenestra vestibuli near its midpoint, further suggesting that the stapedial artery traversed an intracrural foramen in the stapes.

The only physical evidence for the further course of the stapedial artery is the morphology of the horizontal shelf lateral to the primary facial foramen. This shelf is mildly concave at its anterior end, suggesting the presence of a well-developed vessel or nerve. Because the facial nerve lies medial to this concavity and the vein in the prootic canal is small, the most likely occupant is an artery, which would have to be the stapedial artery or one of its primary rami.

The stapedial artery in extant placentals may divide into a ramus superior and ramus inferior or it may end as one or the other of these vessels; the ramus superior enters the cranial cavity, whereas the inferior ramus may enter the cranial cavity or run beneath the skull base, either in a canal or open (Tandler, 1899, 1901; Bugge, 1974, 1978, 1979; Wible, 1984, 1987, 2008). If a ramus superior (“rs” in fig. 19) is present in Maelestes, the only possible point of entry into the cranial cavity is the piriform fenestra; if a ramus inferior (“ri” in fig. 19) is present, it might use the piriform fenestra or another route beneath the skull base not preserved in the fossil.

A final artery that is quite large in a few extant placentals (e.g., armadillos) is the arteria diploëtica magna (Hyrtl, 1853, 1854; Wible, 1984, 1987; Wible and Gaudin, 2004). When well developed, it arises from the occipital artery and enters the posttemporal canal between the squamosal and petrosal and runs forward through the braincase to the orbit, which it achieves via the anterior opening of the orbitotemporal canal; en route it provides meningeal rami that supply the meninges and temporal rami that leave the skull to feed the temporalis muscle. Maelestes has evidence supporting a similar reconstruction (“adm” in fig. 19): a well-developed posttemporal canal with a large posterior opening entirely within the mastoid exposure of the petrosal (figs. 16, 17); a groove on the frontal associated with the anterior opening of the orbitotemporal canal (fig. 11); a foramen for a ramus temporalis in the caudal process of the squamosal off the posttemporal canal (fig. 16); and a broad groove on the mastoid exposure leading to the posttemporal canal (figs. 16, 17), which suggests connection with the occipital artery. It is not known, however, whether the arteria diploëtica magna has a connection to the stapedial artery via a ramus superior as interpreted, for example, in Zalambdalestes (Wible et al., 2004).

Atlas (fig. 20) and Axis (fig. 21)

A partial atlas and axis have been prepared from their articulation with the skull. The atlas is preserved in left and right pieces, with most of the neural arch; the axis is a fragment of the body.

As occurs sporadically throughout mammals, e.g., Late Triassic–Early Jurassic morganucodontids (Jenkins and Parrington, 1976), Early Cretaceous Vincelestes (Rougier, 1993), and early Miocene Necrolestes Ameghino, 1891 (Asher et al., 2007), the neural (dorsal) arch of the atlas of Maelestes is preserved as left and right halves. The left half is the more complete and illustrated here (fig. 20); on the right half, the articular facets are damaged and the transverse process is entirely missing. The right neural arch is more complete at the midline, where it is flat with no indication of a dorsal tubercle. The left neural arch is more than twice as long at the midline as at its lateral extent (2.17 and 1.03 mm); its cranial margin is mildly convex and its caudal margin is mildly concave (“da” in fig. 20C). The base of the left transverse process is oval in outline, longer than wide (“tp” in fig. 20C); it is slightly longer than the narrowest part of the neural arch. The transverse process is short, but its lateral end is damaged and it is uncertain how much longer this element was in life. Ventral to the base of the transverse process is a sulcus indicating the course of the vertebral artery (“sva” in fig. 20A); there is no transverse foramen. The left cranial articular fovea for the left occipital condyle is roughly ovate, with the wider end the prominent cranial edge (“craf” in fig. 20B, D). The articular surface is strongly concave and directed cranially, ventrally, and medially. The left caudal articular fovea for the axis is also roughly egg shaped, with the dorsal end wider (“caf” in fig. 20A). The articular surface is flat and directed caudally and slightly medially. The intercentrum is not preserved.

The axis (fig. 21) is represented by an asymmetrically broken piece of the cranial portion of the body, with more preserved on the right side than the left. The dorsal surface is well preserved (fig. 21B), but bone has flaked off the ventral surface (except for the dens) exposing an endocast (fig. 21A). In light of where the suture between the atlantal and axial parts of the body (Jenkins, 1969) occurs in other taxa, such as Asioryctes (Kielan-Jaworowska, 1977), Zalambdalestes (Kielan-Jaworowska, 1978), and Pucadelphys Marshall and Muizon, 1988 (Marshall and Sigogneau-Russell, 1995), it is likely that the bulk of the element in Maelestes is developmentally the homolog of the body of the atlas. In fact, there may be a remnant of the suture on the dorsal surface about one-third the way up from the caudal margin (fig. 21B). The preserved element is dominated cranially by the peglike, dorsocranially directed dens, the tip of which is imperfectly preserved. Both the dorsal and ventral surfaces of the base of the dens are smooth. Caudolateral to the dens and separated from it is the cranial articular fovea for the atlas, which is fully preserved on the right side. It is oval, slightly convex, and obliquely oriented, such that it is more cranial medially. The dorsal surface of the body (fig. 21B) is flat in the middle with wings curving dorsally at the lateral margins, which bear the facies articularis dorsalis. The endocast on the ventral surface indicates the presence of a low median longitudinal ridge, flanked by narrow depressions (fig. 21A).

Other Vertebrae (figs. 22, 23)

A series of 12 vertebrae are preserved in articulation (figs. 22, 23). All but the cranialmost vertebra have free rib facets and, therefore, are identified as thoracic vertebrae. The cranialmost vertebra is imperfectly preserved and the presence of rib facets cannot be confirmed or denied; it is either the last cervical or a thoracic (abbreviated “C” or “T” in the following). Because the left transverse process (“tp” on “C last” in fig. 22) as preserved appears to be long and horizontal (see Lessertisseur and Saban, 1967a), we identify it as the last cervical (presumably C7). It is exposed in cranial and ventral views (fig. 22), and represented by the body and the laterally directed dorsal roots of the transverse processes; the right process has been broken and shifted posterodorsally. There is no indication of a ventral root or a transverse foramen. The left side may include the base of the pedicle as well, but this is not certain due to damage. The cranial and right ventral surfaces of the body are polished and rounded, the result of postmortem weathering. The left ventral surface is also rounded, but this is natural.

The first three thoracic vertebrae are exposed in ventral view (fig. 22) and represented by partial bodies and on T1 and T3 by the base of the right pedicle (“ped” in fig. 22). The body of T1 has a narrow, flat median surface bounded by shallow lateral concavities. T2 has a broader, flat median surface with low ridges separating it from shallow lateral concavities. T3 has a broad, flat median surface, but lacks lateral ridges and concavities, the lateral surface being flat. The head and neck of the left first rib are only slightly displaced from their natural position. The broken head of the right second rib is preserved in the caudal costal fovea of T1; the left cranial costal fovea of T2 is partly exposed. Visible in dorsal view are parts of T1, T2, and the body of the left second rib (fig. 23). T2 is presented by the base of the broken left transverse process and T1 by the broken left transverse process, the complete left lamina, and the nearly complete left side of the spinous process (“spT1” in fig. 23). The spinous process is long (more than 2 mm), thin, and posteriorly directed.

The bodies of T4, T5, and the cranial half of T6 are exposed in ventral view (fig. 22). These bodies lack a flat median surface and are more rounded. The head, neck, damaged tubercle, and proximal body of the right fourth rib lies between T4 and T5; the head is near its natural position, but the tubercle has moved ventrally and caudally. The cranial surface of the costal angle is gently concave. A distal piece of the body of the right fourth rib lies between the body of T6 and much of the body of the third rib. The third rib is 9.4 mm in length, faintly curved, and anteroposteriorly compressed. Broken parts of the neural arch of T4 are visible in dorsal view (fig. 23). Lateral to this on the left side are the proximal body and tubercle of the left fourth rib. Cranial to this is a similar sized piece of the left third rib, which also includes the head and neck.

Little remains of T7. The bodies of T8–T11 are exposed ventrally (fig. 22). These are gently rounded, but the median surfaces are flatter than those of T4–T6. The damaged head, neck, and tubercle of the right eighth and ninth ribs are near their natural positions. The ninth rib also preserves the proximal body, and the cranial surface of its costal angle is gently concave, as noted for the third rib above. The damaged heads of the right tenth and eleventh ribs lie near their natural positions. The preserved portion of the tenth rib lacks a tubercle, which appears to have been the condition in life. Parts of the neural arches are preserved for T8–T10, with T9 essentially complete (fig. 23). All three vertebrae lack anapophyses and have postzygopophyses (“poz” in fig. 23) projecting caudally and short, narrow transverse processes projecting ventrocaudolaterally. Only T9 preserves the spinous process; it is low and slightly posteriorly directed. The prezygopophyseal-postzygopophyseal articulations between T8–T11 are of the radial type (see Lessertisseur and Saban, 1967a).

Clavicle (figs. 22, 27)

Preserved between the proximal left humerus and the thoracic vertebrae is an approximately 5 mm long isolated bone that likely is part of the left clavicle (“cl?” in figs. 22, 27). If correctly identified, the clavicle is missing both ends and the exposed surface is the caudal one. The bone is craniocaudally compressed with the cranial surface concave and the slightly thicker lateral end more bowed. The alternative is that it is a displaced segment of a rib body.

Scapula (figs. 24, 25)

More than half of the left scapula is preserved. Missing is the dorsal end of the supraspinous fossa, and the acromion and coracoid process are damaged. Matrix has not been removed from the ventral area of the infraspinous fossa to buttress the thin bone, and the glenoid fossa is obscured by the articulated head of the humerus.

The preserved scapula is unlike that of the vast majority of modern therians, in which the lateral surface of the scapular lamina is essentially flat with the supraspinous and infraspinous fossae in the same plane (Lessertisseur and Saban, 1967b; Rougier, 1993; Horovitz, 2003). In contrast, in Maelestes, the supraspinous fossa overlies the infraspinous fossa, nearly completely obscuring it in lateral view (“ssf” and “isf” in fig. 24). A view down the scapula from the vertebral margin shows the lamina to be roughly S-shaped, with the infraspinous fossa under the supraspinous (and the subscapular surface under the infraspinous fossa). Enough of the scapular lamina is preserved (along the medial surface and cranial margin) to show that this S-shaped configuration is natural and not the result of postmortem distortion. Horovitz (2003: 860) reported the same configuration for Ukhaatherium, but noted that it “could be due to postmorten damage, although it is present to a slighter degree in Vincelestes neuquenianus and it has been interpreted to be the natural condition in the damaged scapula of Henkelotherium guimarotae (Rougier, 1993).” In light of the remarkably similar morphology in Maelestes, the condition in Ukhaatherium is probably also natural. Whereas the vast majority of modern therians have relatively flat scapular laminae, there are a few forms in which the supraspinous fossa has some degree of overlap of the infraspinous fossa. The most extreme examples are the chrysochlorids or golden moles (e.g., Chrysochloris asiatica (Linnaeus, 1758) CM 94946, fig. 26; Amblysomus hottentotus (Smith, 1829) CM 40782; Eremitalpa granti (Broom, 1907) CM 10897), but these do not show the total overlapping as occurs in Maelestes and Ukhaatherium. Moreover, in the chrysochlorids, the infraspinous fossa is significantly smaller than the supraspinous; in Maelestes and Ukhaatherium, the reverse is true (see below).

In Maelestes, in lateral view, the ventral end of the narrow, relatively flat supraspinous fossa lies between the scapular neck cranially and the base of the acromion process caudally (“nsc” and “ap” in fig. 24). The caudal and lateral margins are intact, but most of the cranial and vertebral margins are broken. The dorsal extent of the supraspinous fossa is not known; the preserved part is less than half the length of the infraspinous fossa. In Ukhaatherium, the infraspinous fossa extends slightly more dorsally than the supraspinous. The missing dorsal part of the supraspinous fossa in Maelestes exposes the approximate dorsal half of the narrow infraspinous fossa, of which the exposed surface is subtly concave. The caudal and vertebral margins of the infraspinous fossa (“cm” and “vm” in fig. 24) although damaged are natural, but the cranial margin is broken.

The ventral end of the spine or acromion is considerably damaged (fig. 24). As the caudal margin of the supraspinous fossa approaches the acromion, it broadens and then narrows to the acromion base. The incompletely preserved broadened part may represent the base of a missing metacromion (processus suprahamatus of NAV). There is an isolated fusiform bone fragment between the acromion base and humerus that likely represents the tip of the acromion (processus hamatus of NAV). If so, then the acromion projected ventrally about 2 mm beyond the glenoid fossa.

A short, narrow neck, which is damaged along its cranial margin, separates the scapular glenoid fossa (“gf” in figs. 24, 25) from the supraspinous fossa. The humeral head is in articulation, obscuring the rim of the glenoid. From the outer margins, the glenoid is elliptical; its long axis is in the plane of the subscapular surface (“sscf” in fig. 25) and oblique to the supraspinous fossa. In medial view, the glenoid surface is not flat but curved with the ventrolateral end projecting farthest and forming a prominent supraglenoid tubercle.

Medial to the neck at the ventral end of the subscapular surface is the broken triangular base of the coracoid process (“ccp” in fig. 25). The base is broad, but the size and orientation of the coracoid process cannot be determined.

The dorsal half of the medial or subscapular surface is obscured by matrix left to buttress the infraspinous fossa (fig. 25). Its caudal and vertebral margin although damaged are intact, but its cranial margin, which would be continuous with the supraspinous fossa, is damaged. The ventral half of the subscapular surface is exposed. Its caudal margin is complete and ends in the broken base of the coracoid process. Its cranial margin is complete near the scapular neck only. The surface preserved next to the coracoid process and neck is slightly concave.

Humerus (figs. 27, 28)

A relatively complete left humerus (fig. 27) articulates with the scapula. Proximally, periosteal bone is missing from the ventral surfaces of the humeral head, the greater and lesser tubercles, and the proximal 40% of the diaphysis. Distally, the lateral epicondyle and the capitulum are broken and preserved as an isolated piece attached to the head and proximal body of the radius (fig. 28).

The humeral head is not fully exposed (“hh” in fig. 27), but appears to be circular in proximal view. The head is slightly higher (cranial) than the greater tubercle and distinctly higher (cranial) than the lesser tubercle (“gt” and “lt” in fig. 27). Because of damage, the presence of an infraspinous fossa on the greater tubercle cannot be evaluated. Bone in the area of the deltopectoral crest (“dc” in fig. 27) and intertubercular sulcus is missing. In medial and lateral view, the diaphysis in the area of the missing bone is anteriorly convex and broad, whereas distally it is essentially straight and subcylindrical. In light of this contour, the deltopectoral crest likely occupied the area of the missing bone, i.e., the proximal 40% of the diaphysis, and was not much elevated. Also, it is likely that the intertubercular sulcus was shallow.

The preserved medial epicondyle is prominent and thickened dorsoventrally (“mec” in fig. 27). Its medial edge is damaged, which means it was even more prominent. The lateral epicondyle is less prominent and not as thickened dorsoventrally (“lec” in fig. 28). Proximomedial to the trochlea (“tr” in fig. 27), the anterior surface of the medial epicondyle bears a distinct depression and proximal to that is an elongate, oblique entepicondylar foramen (“eef” in fig. 27). Horovitz (2003: 863) described “a well-defined depression between the medial epicondyle and the trochlea on the posterior aspect of the humerus” in Ukhaatherium. This area is damaged in Maelestes, but a small, shallow depression was probably present (fig. 23). The bulk of the trochlea is preserved with the main piece of humerus (figs. 23, 27), but the extreme lateral end is on the isolated piece with the capitulum (“cap” in fig. 28A) and lateral epicondyle. Although there is damage to the lateral trochlea, it appears that there is a slight discontinuity ventrally between the trochlear and capitular surfaces. The trochlea is spindle shaped, tapering from its large medial end. Proximolateral to the trochlea is an oval supratrochlear foramen (“stf” in fig. 27), the distolateral border of which is completed by the broken piece with the capitulum and lateral epicondyle (fig. 28). The olecranon fossa (“of” in fig. 23) is deeper than the radial fossa (“raf” in fig. 27). The capitulum in ventral view is barrel shaped (fig. 28A); it is hidden in dorsal view by the head of the radius (fig. 28B). The lateral end of the capitulum has only a narrow separation from the lateral epicondyle. On the ventral surface between the lateral end of the capitulum and the supratrochlear foramen is a small triangular depression. Proximal to the lateral epicondyle is a distinct lateral supracondylar (supinator) crest, which extends slightly more proximally than the entepicondylar foramen (“suc” in figs. 23, 27).

Ulna (fig. 22)

A fragment of the proximal left ulna is preserved in medial view. Visible are the prominent coronoid process, the distal part of the trochlear notch (“trn” in fig. 22), the broken proximal body, and the broken olecranon process (“op” in fig. 22). The body is approximately half the anteroposterior diameter of the olecranon process. The medial surface of the olecranon is concave; in the dog (Evans, 1993), this surface is the origin for the ulnar heads of the flexor carpi ulnaris and flexor digitorum profundus.

Radius (fig. 28)

A fragment of the head and proximal body of the left radius is attached to the capitulum of the distal left humerus. The hidden articular surface of the head, the capitular depression, appears to be shallow. Viewed from the distal end, the outer circumference of the articular surface is subcircular, slightly wider dorsoventrally then mediolaterally. The rim of the articular surface is not flat but has a low capitular eminence ventrally (fig. 28A) and a low elevation dorsally (fig. 28B). Medial to the capitular eminence is a triangular surface (“ul fac” in fig. 28A) that articulated with the ulna lateral to its trochlear notch (the corresponding surface on the ulna, the radial notch, is not exposed). The preserved radial fragment has no indication of a radial tuberosity. The body is mediolaterally compressed (fig. 28).