Peru•1 specimen (a subcomplete cranium lacking part of the right orbit, part of the basicranium, the ear bones, and teeth); East Pisco Basin, Santa Rosa locality; 14°47’22.9”S, 75°30’22.5”W; Pisco Formation, P0 allomember; Langhian to Serravallian (Middle Miocene); VII.1978; C. de Muizon leg.; MNHN[MNHN.F.PPI282].

Peru•1 specimen (a cranium lacking parts of a 55 mm-long section of the rostrum, part of the palate, of the basicranium, and of the occipital shield, as well as all the ear bones and teeth); same data as for the holotype; MNHN[MNHN.F.PPI283].

None of the three studied specimens preserves postcranial parts, hence the lack of information on the degree of vertebral epiphyseal fusion. The three crania display robust and well-ossified bones, as well as well-defined dental alveoli, and we could not find any indication of fully open cranial sutures, though the suture lines of the interparietal with the parietals and occipital remain visible in the holotype (see below). Cranial dimensions are generally slightly smaller in the paratype, compared to the two other specimens (Table 1), but no clues for an earlier ontogenetic stage could be found. The detached tooth associated to MUSM4693 does not display any significant apical wear and its pulp cavity is nearly closed by dentine. All things considered the three specimens are interpreted as representing subadult to adult individuals (Perrin 1975; Perrin & Myrick 1980).

Bizygomatic width ranges between more than 164 mm and 200 mm, corresponding to an odontocete in the cranial size range of the modern common dolphin, Delphinus ‌delphis. The rostrum is moderately elongated, making 66-67.5 % of the condylobasal length (Table 1), a ratio falling within the ‘longirostrine’ category of McCurry & Pyenson (2019). While the cross section of the rostrum at mid-length is as high as wide in both the holotype and the paratype, the anterior tip of the rostrum is slightly wider in the paratype. The mesorostral groove is broadly dorsally open (though slightly less so in the holotype and MUSM4693), with the maximum opening being located a short distance anterior to the level of the antorbital notches, at about the anterior tip of the anteromedial sulcus (Figs 2, 3, 10A). In lateral view the ventral margin of the rostrum is convex for most of its length, while the dorsal margin is roughly rectilinear. Only complete in the holotype, the upper alveolar series include 40 alveoli each (Fig. 6). In lateral view, the facial region is moderately concave, with a gradual rise (that is somewhat steeper in the holotype) towards the vertex (Figs 4, 5, 10B). The vertex is located at about the same anteroposterior level as the postorbital process. Medial sutures between the bones exposed at the vertex (i.e., nasals and frontals) are situated along the midline (a character best seen in the holotype), and the right and left premaxillary sac fossae are either equal or subequal in width, thus implying little to no facial asymmetry. The minimum transverse width of the vertex is observed across the frontals; it is greater than the width of the bony nares. The temporal fossa is only slightly longer than high, with the roof reaching a dorsal level intermediate between those of the orbit and vertex in the three specimens. Making the posterior edge of the temporal fossa, the nuchal crest extends behind the posterior surface of the occipital shield. The nuchal and temporal crests are modestly developed, i.e., low and thin in cross section. Both the occipital shield and the medial wall of the temporal fossa are convex, thus testifying to an inflated endocranial cavity.

The anterior apex of the rostrum is comprised of the premaxillae alone. The maxilla-premaxilla suture can be followed on the left side of the holotype until 22 mm posterior to the tip of the rostrum, where it turns abruptly ventrally. Only three upper alveoli are found in the premaxilla. The first premaxillary alveolus is only slightly larger than the two others, a feature best seen in the paratype, where the outline of the tip of the rostrum in ventral view is more rectangular than in the holotype. This first alveolus is directed more anteriorly than the others, being possibly subhorizontal in both the holotype and paratype (though alveolar walls are too poorly preserved for a precise assessment of the direction). The first upper incisor may thus have been procumbent, but without reaching the size of the apical upper tusks seen in the holotype of Kentriodon ‌pernix ‌(Kellogg 1927).

Along the lateral surface of the rostrum, the maxilla-premaxilla suture runs into a narrow sulcus, which broadens locally at levels where a small dorsal infraorbital foramen is probably present (though covered by sediment) and followed anteriorly by a larger sulcus (Figs 4; 5). In dorsal view, the premaxilla is wider than the maxilla before the level of the antorbital notch (Figs 2; 3, 10A). At this level, both bones are subequal in width in the holotype and MUSM4693, while the right premaxilla is narrower than the right maxilla in the paratype, making about 75 per cent of the width of the latter. At the level of the antorbital notch, the lateral margin of each premaxilla is only slightly constricted in correspondence to a main dorsal infraorbital foramen that pierces the adjacent maxilla just lateral to the premaxilla-maxilla suture (see below). The transversely narrow premaxillary foramen (13 mm-long in the holotype) is also located at this anteroposterior level in the three specimens. The 40 mm-long anteromedial sulcus laterally defines a transversely concave prenarial triangle with an irregular surface ornamented by thin longitudinal crests and grooves. The broad and well-defined posteromedial sulcus raises obliquely towards the markedly elevated anteromedial edge of the premaxillary sac fossa. Right and left anteromedial edges are at a short distance to each other (more so in the holotype, see Table 1), nearly completely closing the U-shaped bony nares anteromedially. Each premaxillary sac fossa is only slightly concave transversely, a feature that is locally accentuated by the elevation of the medial edge thereof. Each fossa is also only slightly concave anteroposteriorly. The dorsal surface of the fossa is elevated compared to the lateral suture with the maxilla, being separated from the latter by a 7 to 11 mm-high lateral wall. Best developed in the holotype, this wall is excavated by a shallow posterolateral sulcus that vanishes at about mid-length of the bony nares. The lateral margin of the ascending process of the premaxilla is constricted before thickening laterally and anterodorsally to form a transversely directed (in dorsal view) and lateroventrally sloping (in anterior view) small crest (Figs 2-3). This crest shortly overhangs the more ventral part of the bone in all the specimens, being reminiscent of the transversely directed premaxillary crest in some ziphiids (e.g., the modern Berardius spp.). This crest broadly contacts the anterolateral margin of the nasal. It is separated from the high medial edge of the maxilla along the vertex by a curved, deep, and narrow sulcus. A transverse line joining the blunt posterior ends of the right and left premaxillae is at the same anteroposterior level as the posteromedial edge of the bony nares.

The lateral margin of the maxilla is made of spongy bone in the region just anterior to the antorbital notch, making the distinction between bone and hardened sediment – and as such, mechanical preparation of the fossil – locally challenging. The outline of the maxillary flange is therefore difficult to describe; it is interpreted as somewhat laterally convex, dorsoventrally thick, and only little elevated dorsally compared to the slightly transversely concave more median region of the maxilla. The antorbital notch is deep (12-15 mm) and V-shaped in the three specimens, drawing an angle of 45°-50° in dorsal view (Figs 2, 3, 10A). Several small dorsal infraorbital foramina (at least two on each side in the holotype and at least three on the right side and two on the left in the paratype) open anteriorly just lateral to the maxilla-premaxilla suture in the posterior third of the rostrum. A major dorsal infraorbital foramen (diameter ranging between 5 and 7 mm) opens anteriorly to anterolaterally on each side of the holotype and paratype, at about the same level as the premaxillary foramen, at a very short distance (1.5 to 2 mm) from the maxilla-premaxilla suture. A second large foramen (diameter ranging between 4 and 6 mm) is located posterolaterally on both sides of MNHN.F.PPI282 and MNHN.F.PPI283, and it opens laterally. On the left side of the paratype, a smaller (diameter of 3 mm) foramen is observed between the two main foramina, while the holotype also includes a smaller (2 to 2.5 mm) foramen, but posterior to the second main foramen, seen on both maxillae. The posteriormost dorsal infraorbital foramen is found at the center of the ascending process, posterior to the level of the postorbital process. Its diameter ranges from 5 to 7.5 mm, and the broad sulcus that leaves posterolaterally from this foramen is partly bisected by a shallow crest on the right side of both the holotype and paratype. The dorsal surface of the maxilla in the antorbital region is only slightly more elevated than the medial portion of this bone, a feature that is more developed in the paratype. Better defined in the latter specimen, two low and thin maxillary crests, run parallel 9-10 mm from each other, in a posterolateral direction, from this poorly elevated region. Their course is more difficult to follow in the area dorsal to the temporal fossa, but they can be distinguished along the anterior flank of the nuchal crest, likely corresponding to separate origin areas for the facial muscles (possibly, the pars anteroexternus and pars anterointernus, see Heyning 1989). No dorsal exposure of the maxilla medial to the premaxilla could be detected along the bony nares and/or the posterior part of the mesorostral groove.

Along the vertex, the medial edge of the maxilla is abruptly elevated dorsally and dorsolaterally, reaching roughly the same dorsal level as the nasal, and markedly overhanging the more ventral part of the maxilla, providing thus this region with a pinched aspect in posterodorsal view. This edge of the maxilla is anteroposteriorly longer in the paratype (29.5 mm on the left side vs 26 mm on the holotype), corresponding to a longer vertex compared to the holotype. The posterior edge of the maxilla is damaged in both the holotype and paratype, but the preserved parts indicate a narrow exposure of the frontal between the maxilla and the interparietal or parietal on the anterior flank of the nuchal crest. Each maxilla reaches a level that is distinctly posterior to the anteromedial part of the nuchal crest, giving a deeply sigmoidal outline to the anterodorsal margin of the occipital shield in dorsal view.

All the 37 alveoli for maxillary teeth counted on each side of the holotype correspond to single-rooted teeth (Fig. 6). The transverse diameter of the anterior alveoli ranges from 4.5 to 5 mm in the paratype and from 6 to 6.5 mm in the holotype. The corresponding interalveolar septa are high, well-defined, and 1 to 2 mm long. In the better-preserved holotype, the alveoli from the mid-part of the rostrum have a transverse diameter of 6.5 to 7 mm (slightly smaller in the paratype), with generally longer (1.5 to 3 mm) septa in the holotype (contra 1 to 1.5 mm-long septa in the paratype; Fig. 7). The diameter of the alveoli decreases posteriorly, with the last alveoli having a diameter of 4.5 to 5 mm in the holotype. The last five alveoli gradually detach from the lateral margin of the rostrum. Only partly visible in the holotype, the vomerine trough is deep and narrow along the median third of the rostrum length. The maxilla-premaxilla sutures are visible along the flanks of the trough, diverging anterolaterally and reaching the medial margin of the alveolar row. Only a few palatine foramina are visible: a small one, 1.5 mm in diameter, is found 23 mm anterior to the tip of the right palatine of the holotype, and at least another is found along the palatine-maxilla suture on both sides of the same specimen. The ventral surface of the maxilla is transversely concave in the area between the posterior tract of the alveolar series and the palatine. This depressed area may indicate the presence of an anterior sinus, though the outline of such a depression at the rostrum base does not always correspond to the outline of the anterior sinus in extant delphinidans (see Fraser & Purves 1960).

In both the holotype and paratype, the anterior tip of the palatine reaches a level anterior to the last posterior maxillary alveolus (Figs 6-7). The tips of the right and left palatines are distant from each other, the maxilla-palatine sutures drawing a W-shaped outline. From this tip, the suture runs posterolaterally for a short distance, before turning anterolaterally, making a secondary anterior projection before running posteriorly and posteromedially. The constriction of the palatine resulting from this posteromedial direction of the suture is maximal at the level of the antorbital notch. From there, the suture leaves posterolaterally towards the medial edge of the ventral infraorbital foramen. This region features a large and deep depression that marks the dorsalmost portion of the outer surface of the lateral lamina of the palatine. We interpret this depression as part of the site of origin for the pterygoideus internus muscle, based on topological similarities with extant delphinids (Seagars 1982). The narrow anterior part of the pterygoid sinus fossa extends in the palatine until a level just posterior to the last maxillary alveolus, thus well anterior to the level of the antorbital notch, in the three specimens. The lateral lamina of the palatine is high (35 mm from the medial edge of the ventral infraorbital foramen to the ventromedial edge of the lamina) and extends posterolaterally until the level of the anterior edge of the frontal groove, posterior to mid-orbit length.

The whole hamular region of the pterygoid is lost in both the holotype and the paratype, but a section in the pterygoid just anterior to the choana in the paratype indicates that at least the posteromedial base of the hamular process was relatively robust (Fig. 7). Similarly, no trace of the lateral lamina of the pterygoid could be found. This bone is preserved along the medial wall of the pterygoid sinus fossa, only covering the palatine for less than half the distance between the choana and the anterior apex of the pterygoid sinus fossa. Better preserved on the paratype, the medial lamina of the pterygoid extends as a thin plate for at least half the length of the basioccipital basin. In this region, the base of the lateral surface of the medial lamina is excavated by a shallow fossa, starting 10 mm anterior to the ventral carotid foramen, which likely corresponds to an extension of the peribullary/pterygoid sinus fossae.

The base of the slender styliform process of the jugal is posteromedial to the antorbital notch (Figs 6-7). The sutures with both the maxilla (anteromedially) and the lacrimal (posteriorly) can be distinguished at least locally. The jugal is thus not fully fused with the lacrimal, though the lateral portion of their contact is hidden by sediment.

In lateral view the lacrimal underlies the preorbital process of the frontal as a thick (up to 7.5 mm) oblique plate that contributes to the robust and ventrally rounded outline of this region, a feature that is visible in the three specimens (Figs 4, 5, 10B). In ventral view, the lacrimal-maxilla suture is directed posteromedially towards the posterior edge of the antorbital fossa, where its outline is more difficult to follow. The lateral margin of the ventral infraorbital foramen is interpreted as being made by the lacrimal. The ventral surface of the lacrimal makes more than one third of the total ventral surface of the orbit. Posterolateral to the base of the styliform process, this surface is depressed, forming a distinct lacrimo-maxillary fossa.

The nasals are much transversely wider than the bony nares and slightly wider than the exposure of frontals on the vertex in the holotype, paratype, and MUSM4693; they occupy a dorsal surface greater than that of the frontals (Figs 2-3, 10A). Due to a shorter exposure of the frontals in the holotype compared to the paratype, the nasals reach posteriorly a level closer (5.5 mm) to the occipital shield in the former specimen. The highest point of the nasals is only slightly higher than the frontals. A well-defined internasal fossa that slopes anteroventrally excavates the medial third of both nasals (Figs 2-3, 8A, 9A). This internasal fossa is deeper in the holotype. The high anterior margin of each nasal is cut by a deep vertical notch, which defines an anterolateral spine of the nasal extending anteriorly along the medial edge of the ascending process of the premaxilla. The spine is longer on the holotype, with a tip at least 9 mm anterior to the anteriorly convex anteromedial edge of the nasal. The dorsal surface of the spine is roughly flat, and slopes anteroventrally and slightly medially. Best seen in the holotype, the joined nasal-frontal sutures display a complex, interdigitated outline, drawing a widely open V that points forwards in their medial portion, while the lateralmost part of the suture is roughly laterally directed. In the paratype, the joined sutures display a more regularly convex outline in their medial part.

On the vertex, the exposed dorsal surface of the frontals is transversely convex and irregular, marked by numerous thin grooves and crests (Figs 2-3). Anteroposteriorly shorter in the holotype, the frontals are separated in the latter specimen by a deep and narrow (4 mm wide) fossa for most of their medial margin. Such a fossa, which is absent in the paratype, may testify to a subsequently lost separate center of ossification, possibly corresponding to a vestigial anterior median interparietal (Roston et al. 2023), an interpretation that should be tested with a CT scan across the vertex region. The internasal fossa does not invade the frontals. The posterior margin of the frontals on the vertex, along the nuchal crest, is slightly anteriorly convex (more distinctly so in the holotype).

Best preserved on the paratype, the robust preorbital process is triangular in lateral view, with a swollen lateral surface that is also seen in MUSM4693 (Figs 4-5, 10B). Its ventral extent (17 mm on the left side of the paratype) reaches not far from the dorsoventral level of the tip of the postorbital process, giving to the dorsal outline of the highly arched orbit a semi-elliptical shape. Only complete on the right side of the holotype, the postorbital process is long (31 mm), slender (with an anteroposterior thickness of 4.5 mm at 6 mm from the tip), and subvertical, only displaying a slight anteroventral curve of the apical portion. The tip of this process is at a close distance (no more than 2 mm) from the anterodorsal margin of the zygomatic process of the squamosal. None of these two parts display any indication (e.g., a fossa or a crest) for a contact, and the condition observed in the holotype is thought to represent the original relationship between the frontal and the squamosal, given the similar position and direction of the nearly complete postorbital process on the left side of the paratype and the lack of any clue for deformation in this area (consider the regular outline of the temporal fossa and of the occipital shield in the holotype). In ventral view, the infratemporal crest is well defined, and somewhat more acute in its medial portion in the holotype (Figs 6-7). Observed in both specimens, the fossa for the postorbital lobe of the pterygoid sinus is anteroposteriorly broader in the paratype, in which it probably partly crossed through the infratemporal crest. A shallower fossa with poorly defined edges is detected between the frontal groove and the antorbital fossa, only shortly invading the lacrimal; it is tentatively interpreted as the fossa for the preorbital lobe of the pterygoid sinus. Anterior to the infratemporal crest, the ventral surface of the frontal groove is pierced by several tiny foramina, some of which may correspond to ethmoidal foramina (or, for the more lateral, to foramina for the frontal diploic vein); on the right side of the holotype, four foramina are detected between the infratemporal crest and the thin oblique plate of the orbitosphenoid that makes the anteromedial margin of the frontal groove. The posteriorly pointed, transversely concave ventral surface of the roof of the temporal fossa is made by the frontal alone, the interdigitated frontal-parietal suture following the dorsomedial corner of the temporal fossa in a posterolateral direction. Best seen on the right side of the paratype, this suture reaches the highest point of the temporal fossa before turning dorsally along the nuchal crest (Fig. 5A).

Due to the presence of indurated sediment in the mesorostral groove and part of the vomerine trough in both the holotype and paratype, as well as to the damaged state of the basicranium in the former specimen, the vomer is only visible in the basioccipital basin of the latter, extending posteriorly for the same distance as the medial lamina of the pterygoid (Fig. 7).

Combining the right side of the paratype for the frontal-parietal suture and the right side of the holotype for the squamosal-parietal suture, the parietal makes most of the medial wall of the temporal fossa (Figs 4; 5A). This surface is anteroposteriorly and, to a lesser degree, dorsoventrally convex. In ventral view, a narrow stripe of parietal is visible posterior to the alisphenoid and medial to the squamosal on the basicranium of the holotype, being partly covered by sediment (Fig. 6). On the better-preserved occipital shield of the holotype, the parietals may contribute to the dorsolateral portions of the shield as small triangular exposures (longest side of each triangle approximately 15 mm long) (Figs 2; 8B). Medial to this triangle, a dorsal stripe of the shield displays an irregular surface that contrasts with the smooth surface more posteroventrally. The limit between the irregular and smooth surfaces is marked by a small step, with an outline that is reminiscent of the interparietal-occipital suture identified for example in Tursiops ‌truncatus (Mead & Fordyce, 2009, fig. 5). This dorsal region with thicker bone is thus interpreted here as being formed by the interparietal. A roughly similar separation is locally found on the right side of the paratype (Fig. 3). In the holotype, medial to the proposed suture between the parietal and interparietal, the corner of the interparietal bears a small prominence, most likely for neck muscle attachment. Medially, the interparietal of the holotype and MUSM4693 is marked by a thin but prominent external occipital crest, starting on the nuchal crest and extending posteroventrally for a short distance on the occipital (Figs 2, 8B, 10C). Though lower, a similar crest is present on the paratype. Lateral to the crest, the occipital shield is transversely and anteroposteriorly concave.

Just posteroventral to the external occipital crest, the occipital shield becomes markedly convex transversely and anteroposteriorly in all specimens. Further posteroventrally, a longitudinal groove broadens and deepens down to a short distance to the foramen magnum, separating two posteriorly bulging areas (Figs 2; 3; 8B; 10C). The right and left temporal crests converge slightly ventromedially until a level 20 mm above the foramen magnum, thus providing the occipital shield with a pinched aspect in posterior view.

In both the holotype and paratype, a single large oval opening occurs posteromedial to the frontal groove (Figs 6-7); it does not allow for distinguishing between the optic foramen, the inferior and superior orbital fissures, and the foramen rotundum. As the edges of this confluent opening are all very thin, the differences between the holotype and paratype with respect to its outline (e.g., the opening is more extended anterolaterally in the paratype) are most likely not meaningful. Though the frontal-orbitosphenoid suture is difficult to follow in this region, the latter bone is at least partly preserved in both specimens as a thin, oblique plate making the anteromedial wall of the frontal groove and extending posteromedially along the anterior edge of the large confluent opening. The frontal-alisphenoid suture is easier to follow, zig-zagging posterodorsally for a short distance on the medial wall of the temporal fossa up to the frontal-parietal suture. The alisphenoid extends farther anterodorsally in the holotype, reaching the posteromedial margin of the fossa for the postorbital lobe of the pterygoid sinus. The alisphenoid is exposed as a narrow stripe in the temporal fossa, and this part of the bone is separated from its basicranium part by a thin anterior extension of the subtemporal crest (Fig. 7B). Medioventral to this crest, the ventrolaterally facing surface of the alisphenoid is irregularly concave. Along the posterior margin of the alisphenoid, the large foramen ovale (6 mm of diameter in the paratype) appears to be confluent with the cranial hiatus (see below). The path for the mandibular nerve is broad and well-defined for about 10 mm, passing laterally dorsal to the tip of the falciform process.

The best-preserved squamosal is the right one of the holotype (Fig. 4). The relatively robust (for an early delphinidan) and moderately elongated zygomatic process is directed anterodorsally. In lateral view, the dorsoventrally high (21 mm) anterior margin of this process is subvertical. It draws an angle of about 90 degrees with the ventral margin, making an angular anteroventral corner, while the anterodorsal corner is more rounded. The maximum transverse width of this part of the zygomatic process is 26 mm, taken dorsally. The transversely broad dorsal margin of the process is approximately rectilinear in lateral view, being followed posteriorly by a slight elevation of the supramastoid crest. The thinner ventral margin of the zygomatic process is also roughly rectilinear. It bears an 8 mm-long anteroventral prominence that corresponds to the contact surface with the styliform process of the jugal. The lateral surface of the zygomatic process is covered by small tubercles, as observed in several inioids (e.g., the extinct Brujadelphis ‌ankylorostris and the extant Inia spp.; Lambert et al. 2017). The postglenoid process is robust, with an anteroposterior length of 11 mm and a mediolateral width of 10.5 mm. Its apex reaches a level more ventral than the posttympanic process, but dorsal to the apex of the paroccipital process. In posterolateral view, the posttympanic process is broad, the external auditory meatus being 18 mm distant from the exoccipital. The subcircular sternocephalicus fossa is deep but does not extend anteriorly past the level of the external auditory meatus, thus facing posterolaterally. Though its anterior part remains filled by sediment, the squamosal fossa — making the floor of the temporal fossa (not to be confused with the squamosal fossa on the periotic of some mysticetes; Mead & Fordyce 2009) — is transversely broad (more than 26 mm) and deep, with a marked anteroventral slope of its surface. Like the rest of the medial wall of the temporal fossa, the squamosal plate is anteroposteriorly convex. In this region, the parietal-squamosal suture rises for 28 mm dorsal to the level of the supramastoid crest.

In ventral view, the mandibular fossa is transversely wide (26 mm), and it faces anteriorly and slightly ventrally (Fig. 6). The tympanosquamosal recess is vast, extending dorsal to the medial edge of the mandibular fossa, deeply excavating the medial surface of the postglenoid process, where it extends nearly as far as the ventral apex of the latter (also seen in MUSM4693). The recess reaches the ventromedial surface of the zygomatic process anterior to the mandibular fossa, at least 9 mm anterior to the level of the anterior edge of the squamosal fossa. In its main medial region, the floor of the recess is irregular, being covered with small, deeper fossae. The plate-like falciform process bears a pointed, narrow anteroventromedial projection with a length of 12.5 mm (taken from the anteromedial margin of the tympanosquamosal recess) in the holotype. Together with the shape of the lateral lamina of the palatine (see above), this condition of the falciform process suggests that, like in most other delphinidans (e.g. Muizon, 1988b), there was no contact between the lateral lamina of the pterygoid (that would be lost in these two specimens) and the squamosal at this level. On the narrow external auditory meatus (2 mm wide at mid-length in the holotype), the anterior meatal crest appears to be higher than the posterior meatal crest, though this region is not optimally prepared. The spiny process is interpreted as ending 27 mm medial to the outer opening of the meatus.

Only the upper part of the right occipital condyle is preserved in the holotype (Figs 4, 8B), showing a moderately well-defined (i.e. short but distinct) neck. Based on the natural endocast, the foramen magnum should have been wider (more than 33 mm) than high (26 mm), oval in outline, and with an only slight dorsomedial elevation (i.e. no distinct opisthion). The dorsal condyloid fossa is very deep, extending up to 10 mm anterior to the anterolateral edge of the condyle. Only partly preserved on the right side of the holotype, the more lateral part of the exoccipital is directed anterolaterally, being overhung by the temporal/supramastoid crest. Also incomplete, the ventral margin of the exoccipital was originally anterior to the level of the occipital condyle and reached more ventrally than the postglenoid process.

Only the anterodorsal portions of the basioccipital crests are preserved in the paratype (Fig. 7). Their section is transversely thin. The left and right crests are 51 mm distant from each other anteriorly. The jugular notch is partly preserved on the right side of the holotype; its anterior section is broad and U-shaped. Dorsolateral to the basioccipital crest, the basioccipital contributes to the partial closure of the space in between the foramen ovale and the cranial hiatus (best seen on the left side of the paratype). With a transverse diameter of 2 mm, the ventral carotid foramen is 10 mm anteromedial to the foramen ovale along the lateral surface of the basioccipital crest, just posteromedial to the oblique suture between the medial lamina of the pterygoid and the basisphenoid.

The left periotic of MUSM4693 is well preserved, only missing most of the posterior bullar facet and the distalmost part of the posterior process and displaying a couple of fractures across the whole bone with minimum to no displacement and some slight damage to the anterior portion of the lateral tuberosity (Fig. 11). The accessory ossicle of the tympanic is preserved attached in the fovea epitubaria and the partial stapes is held in situ in the fenestra ovalis (see below). The periotic has a length of more than 24.8 mm (see all dimensions in Table 1). Based on the dorsal outline of the posterior process in lateral view, the size of the break cross section, and the general shape similarities with the periotic of Kentriodon ‌pernix ‌USNM 8060 (Kellogg 1927, plate 3), the total length probably did not exceed 25.3 mm.

The anterior process is shorter than the pars cochlearis. At mid-length, it is slightly dorsoventrally thicker than mediolaterally wide (not taking into account the accessory ossicle). In ventral view (Fig. 11A), it turns regularly anteromedially, with slightly convex medial wall of the parabullary ridge. The anteroventral tip of the anterior process bears a few small tubercles that make its outline locally somewhat more robust. In medial view (Fig. 11B), the anterodorsal angle is prominent, with an angle of 120° between the rectilinear dorsal and anterior margins. The anterodorsal angle is posterior to the level of the anteroventral tip of the process. Best seen in dorsal view (Fig. 11D), as it is partly hidden by the accessory ossicle in ventral view, the anterior incisure is broad and U-shaped, the medial surface of the anterior process being well removed from the anterior surface of the pars cochlearis. In lateral view (Fig. 11C), the surface of the anterior process is slightly depressed just dorsal to the recurved parabullary ridge.

The pars cochlearis is proportionally high in the medial direction and moderately tilted anteriorly in ventral view. Its anteromedial margin is more angular than its posteromedial margin. The ventral and medial surfaces of the pars cochlearis are regularly convex, with no trace of a median promontorial groove. In medial view, the pars cochlearis has a nearly flat dorsal outline, corresponding to the medial edge of the internal auditory meatus (IAM), and its long axis in this view is oblique, directed anteriorly and slightly anterodorsally. The maximum ventral bulging is located anterior to the fenestra rotunda. The latter is large, with a diameter of 2.8 mm and a semi-circular outline. Its mediodorsal corner is followed mediodorsally by a shallow, short sulcus. The aperture for the cochlear aqueduct is small and roughly circular (diameter 1.5 mm), opening dorsomedioposteriorly at a short distance (1 mm) from the IAM. More distant from the IAM (2.8 mm), the aperture for the endolymphatic duct has similar dimensions (maximum diameter 1.4 mm), is slightly dorsoventrally flattened and opens posterolaterodorsally, being followed by a broad groove for 3 mm towards the posterodorsal corner of the periotic (= dorsal tuberosity in Mead & Fordyce 2009). The IAM is vast, occupying a large part of the dorsal surface of the pars cochlearis, and tear-shaped, extending as a deep, U-shaped sulcus up to the anterolateral limit of the pars cochlearis (Fig. 11D, G). The anteromedial edge of the IAM is more rounded in section than the acute posterolateral edge. The tractus spiralis foraminosus and the area cribrosa media (see Orliac et al. 2020 for terminology, as well as Ichishima et al. 2021 for an alternative set of anatomical terms) are separated from the laterally located area cribrosa superior (singular foramen in Mead & Fordyce 2009) by a thin crista transversa (or transverse crest). The latter is much narrower, but similar in height to the perpendicular crest, which separates the area cribrosa superior from the anterolaterally located internal opening of the facial canal. The top surface of the broad (about 1 mm wide) perpendicular crest is excavated by a shallow fossa. The internal opening of the facial canal is more anterior than the tractus spiralis foraminosus, but does not reach the anterolateral corner of the pars cochlearis, which is excavated by the U-shaped sulcus that leaves from this opening. Posterolateral to the IAM, the superior process is barely developed, only projecting 1.5 mm dorsal to the level of the medial margin of the IAM and displaying a nearly flat (only slightly convex) dorsal outline in medial view, without any distinct crest.

The fenestra ovalis (with the in situ partial stapes) has an oval outline, with an anteromedially directed long axis (Fig. 11A). Posterior to the fenestra ovalis, the lateral caudal tympanic process (sensu Lambert et al. 2018) is barely defined; the medial edge of the fossa for the stapedial muscle undulates slightly, without any distinct lateral projection. The fossa for the stapedial muscle is broad (2.6 mm) and extends posteriorly to the incomplete posteromedial margin of the posterior process, being separated from the facial sulcus by a low, thin oblique ridge. From this ridge, the short facial sulcus turns anterolaterally and then anteriorly to the distal opening of the facial canal. The latter is anterior to the level of the fenestra ovalis. Dorsal to the posterior end of the fossa for the stapedial muscle, the posteromedial surface of the posterior process is excavated by a shallow but well-defined stylomastoid fossa, with a dorsoventral diameter of about 3 mm. In medial view, the dorsal margin of the posterior process turns abruptly posteroventrally, indicating a short main part of the posterior process. The posterior break surface of this process (Fig. 11E) suggests that the dorsoventrally thin posterior tip of the process and the posterior bullar facet should have turned posteroventrally to some extent, as e.g. in Kentriodon ‌pernix ‌USNM 8060. The lateral tuberosity is anteroposteriorly long (8 mm) but projects only little laterally. Consequently, the epitympanic hiatus is shallow (at least with respect to its anterior wall, as the posterior wall is incomplete, due to the missing posterior bullar facet). The broad (mediolateral diameter 3.1 mm) and shallow mallear fossa faces ventrally and slightly medially. The fossa incudis could not be detected, but is thought to have been lost together with the posterior bullar facet.

Preserved in situ within the fovea epitubaria of the periotic, the detached left accessory ossicle is 7.3 mm long anteroposteriorly and 4.3 mm wide mediolaterally (Fig. 11A). Anteriorly, it reaches a level 2.6 mm posterior to the anteroventral tip of the anterior process of the periotic.

The main preserved part of the left tympanic (preserved length = 27 mm) is the involucrum, which only lacks its anterior tip (Fig. 12). At about mid-length of the bone, the involucrum is 9.5 mm thick dorsoventrally. Its dorsal and ventral margins are subparallel for most of its length, except for a moderate anteroventral slope of the dorsal margin in the anteriormost 5 mm. The inner posterior prominence has a regularly rounded outline in medial view, with no distinct corner. In dorsal view, the medial margin of the tympanic is convex along its posterior half, transitioning anteriorly to a slightly concave region that is further followed by the convex anteriormost part. The maximum mediolateral width of the involucrum (10.6 mm) is found in its posterior region, just anterior to the partly preserved inner pedicle. From that level, the lateral margin of the involucrum gradually turns anteromedially, with a thickness decreasing down to 5 mm at the preserved anterior tip. In ventral view, the inner posterior prominence is narrow (5-6 mm) and laterally margined by a deep median furrow (Fig. 12C). Only partly preserved, the latter extends for at least two thirds of the preserved length of the tympanic.

The basal part of the stapes, which is kept in situ in the fenestra ovalis, has an oval outline (the long axis is 2.3 mm long) (Fig. 11A). It is proposed that the stapes broke across the anterior crus, possibly at the level of a vestigial stapedial foramen.

One subcomplete detached tooth of MUSM4693, interpreted as a posterior lower right tooth based on the crown-to-root proportion, crown curvature, and root curvature (see below), has been prepared and is described here (Fig. 13). Only missing the tip of the root, it is 19.6 mm long, including a 7.8 mm-long crown. Labiolingual and mesiodistal diameters at the crown base are 4.3 and >4.0 mm, respectively. The crown is curved apicolingually, with a slight bulge on the lingual surface, just above the root-crown boundary. The thin (0.2 mm, measured along a minor break surface) enamel layer is consistently smooth. The root is roughly rectilinear. While no apical wear facet could be detected, shallow wear surfaces along the distal and mesiolabial surfaces of the crown likely witness the occlusion with the preceding and succeeding upper teeth. The pulp cavity is nearly completely filled, with only a tiny (0.1 mm) opening observed at the center of the cross section of the tip of the root.

Several broken upper teeth are preserved in situ in the region of the left upper alveolar series about 160 mm anterior to the level of the antorbital notch in MUSM4693 (Fig. 10B). Their diameter at the crown base is 4.5 mm.