Institutions and fossil collections: CRBMC: Centre de Restauració de Béns Mobles de Catalunya (Valldoreix, Barcelona, Spain); ICP: Institut Català de Paleontologia Miquel Crusafont, Universitat Autònoma de Barcelona (Spain); IPS: acronym of the collections of the ICP; MGSB: Museu Geològic del Seminari de Barcelona (Spain); MVT: field acronym for Mas Vilageliu (Tona, Barcelona, Spain).

Anatomical abbreviations: BO: basioccipital; BS: basisphenoid; CL: crista lambdoidea; CT: crista temporalis; EO: exoccipital; FM: foramen magnum; FR: frontal; J: jugal; LAC: lacrimal; MX: maxilla; N: nasal; NPP: nasal process of premaxilla; OC: occipital condyles; PA: parietal; PAL: palatine; PM: premaxilla; PT: pterygoid process; SO: supraoccipital; SQ: squamosal.

This paper is based on the description of the partial skull IPS81881a (field catalog number: MVT10), which was found associated with 45 fragmentary postcranial remains presumably from the same individual. These postcranial remains are currently unprepared because of funding shortage, and therefore cannot be described here. They include: a relatively complete vertebra (MVT27); three partial vertebrae (MVT7, MVT18 and MVT28); partial vertebrae and vertebral fragments (MVT15, MVT29, MVT32, MVT39); partial ribs and rib fragments (MVT1, MVT3–MVT6, MVT12–MVT14, MVT16, MVT17, MVT20–MVT26, MVT33–MVT36, MVT40–MVT42); and unidentified bone fragments (MVT8, MVT19, MVT30, MVT31, MVT37, MVT38). All these fossil remains are currently housed at the ICP.

Other sirenian specimens examined at the MGSB in the course of this study are the following: type specimens of P. solei from Tona (here assigned to P.? intermedium), including MGSB48657-2 (holotype, left mandibular ramus with three molars), MGSB48657-1 (paratype, skullcap with supraoccipitals, parietals and frontals) and MGSB48657-4 (paratype, partial maxilla and two associated fragments with three molars and several tooth alveoli); and the holotype of P.? montserratense (nomen dubium) from Castellbell i el Vilar (MGSB44892, partial skeleton with cranial and axial elements, still embedded in matrix).

Anatomical nomenclature for the cranium follows Domning (1994), Rose (2006: fig. 2.1) and Zalmout and Gingerich (2012: fig. 8), whereas dental terminology follows Smith and Dodson (2003) and Sorbi et al. (2012: fig. 10A).

The cladistic analysis performed in this paper is based on the data matrix of 69 craniodental characters employed by Vélez-Juarbe and Domning (2014: Supplementary data, Table S1). As in Vélez-Juarbe and Domning (2014), the proboscidean Phosphatherium escuilliei Gheerbrant et al., 1996 and the desmostylian Cornwallius sookensis (Cornwall, 1922) were employed as outgroups. Besides prorastomid, protosirenid, trichechid and dugongid species already coded by Vélez-Juarbe and Domning (2014), we further coded additional halitheriine dugongids of genera Eotheroides, Eosiren and Prototherium, based on published literature and the observation of fossil material. The newly coded taxa are the following: Eotheroides lambondrano Samonds et al., 2009 (based on the original description); Eosiren libyca Andrews, 1902 (further based on Andrews, 1906 and Domning, 1994); Eosiren imenti Domning et al., 1994 (based on the original description); P. veronense (based on Domning, 1994, Pilleri et al., 1989 and de Zigno, 1875); P.? intermedium (including P. “solei”; based on Bizzotto, 2005, Domning, 1994 and Pilleri et al., 1989; and the material examined by us at the MGSB); and P. ausetanum sp. nov. (based on this study). The resulting data matrix is presented in Supplementary data, Table S1. Of the 69 characters included in the data matrix, 7 characters were constant and 12 parsimony non-informative, although they are provided here for the sake of completeness. A cladistic analysis based on maximum parsimony was performed by means of PAUP* 4.0b10 for Unix (Swofford, 2003), with the branch-and-bound option. Following Vélez-Juarbe and Domning (2014), all characters were treated as unordered. Clade stability was assessed by means of Bremer's support indices (which compute the extra steps required in order to collapse a clade in the consensus of near-most-parsimonious trees) as well as with boostrap analysis (1000 replicates). The Consistency Index excluding uninformative characters (CI), the Retention Index (RI) and the Rescaled Consistency Index (RCI) are reported.

The cranium (Fig. 1 and Fig. 2) is overall quite complete, although the right side is crushed and only partially preserved (so that several portions, such as the zygomatic arch, are lacking). It is also slightly deformed, with the basal region displaced slightly to the left. The well-preserved left side of the skull (Fig. 1 and Fig. 2) includes both the splanchnocranium and the neurocranium. It probably belongs to a subadult or young adult individual (based on the lightly worn M3 and the unfused basioccipital-basisphenoid suture). Cranial and dental measurements are reported in Table 1 and Table 2, respectively. Preserved cranial length is 309 mm; only its rostral-most portion is missing, so that we estimate that its maximum length would not have surpassed 330 mm. Based on a mirror image from the left side, we estimate maximum skull breadth at 158 mm (see also Table 1). Rostral deflection (following Velez-Juarbe et al., 2012: fig. 1C), measured on the left (undeformed) side, is 47°.

Premaxilla: This bone is broken on its anterior-most portion (Fig. 1 and Fig. 2), particularly on the right side, so that the nasal process is completely preserved only on the left side. The rostrum seems to have been long, but small relative to the cranium (c. 3[1] according to our estimate of condylobasal length), with its dorsal surface forming a keel, and the posterodorsal surface forming a boss when viewed laterally (c. 10[1]). The nasal process contacts the frontal bone (c. 9[1]; as is typical of sirenians, see Domning, 1994) and laterally embraces the nasal (c. 6[0]), displaying a sinuous shape when viewed dorsally (Fig. 1 and Fig. 2). The premaxilla is eroded on its ventral side (Fig. 1 and Fig. 2), so that the anterior dentition is missing. The premaxillary-maxillary suture is behind (instead of below) the premaxillary symphysis.

Maxilla: This bone is robust and displays a curved basal contour in lateral view (Fig. 1 and Fig. 2), especially above the molars. The zygomatic-orbital bridge is situated slightly above the alveolar plane (6 mm, c. 11[0]; Fig. 1 and Fig. 2). Due to incomplete preservation, it is not possible to conclusively ascertain whether the zygomatic-orbital bridge is long or shortened (c. 14[?]). The rostral palatal surface of the maxilla (Fig. 1 and Fig. 2) is broad on its anterior portion, progressively narrowing towards the premolars (c. 23[0]), where it has a groove-like depression (9 mm in width and 42 mm in length) that is deepest at the level of P1–P2. The maxilla slightly widens backwards towards its contact with the palatines, at the mesial level of the M1.

Upper cheek teeth: On the right side, the alveolus of the canine (c. 144[1]; 11.8 mm in mesiodistal length), which preserves tiny dentine fragments, can be discerned. Distally from it, there is a short diastema of 7 mm, followed by the cheek teeth series (c. 151[0]). It includes the empty alveoli of four single-rooted premolars (c. 157[1]), which are interpreted as P1–P4 (c. 145[0], 146[1]) and have a mesiodistal length of 8.4 mm (P1), 7.8 mm (P2), 6.5 mm (P3) and 8.5 mm (P4). They are followed by the molar series (M1 to M3; c. 150[0]), which is preserved on both sides, although the M1 and M2 are very worn and somewhat broken (Fig. 4). The M1 is shorter than the M2, whereas the M3 is longer than the preceding molars (Table 2). The M1 displays a quadrangular occlusal outline, with the protoloph and metaloph apparently of the same width—as far as their worn state allows us to ascertain it. The M2 is more rectangular (broader than long), with the protoloph and the metaloph also similarly wide. Due to its lesser degree of wear (least accentuated on the buccal side), occlusal morphology can be best ascertained on the M3, which clearly displays an oval (distally-tapering) outline and a bunolophodont occlusal pattern. The mesial precingulum displays a large secondary cuspule, which is mostly worn away and connected to the protoloph. The latter is constituted by the worn protocone and paraconule (obliterated by wear) as well as the more distinct and protruding, conical paracone. The metaloph, narrower and less protruding than the protoloph, is also tricuspidated, being constituted by hypocone and metaconule (partly obliterated and merged with one another due to wear), as well as the more distinct and conical metacone (smaller than the paracone). The central valley between the protoloph and metaloph is narrow but continuous on the right side (where it is only partly interrupted by a short spur originating from the metacone), and more distontinuous on the left side (where the ridge constituted by the hypocone and metaconule is linked to the protoloph). A postcingulum originates distally from the hypocone; it curves distolabially and is open labially (well separated from the metacone), being constituted by two secondary cuspules situated on the distolabial corner of the crown. There is a very narrow labial cingulum around the bases of the paracone and metacone, as well as a wider, but discontinous and wrinkled lingual cingulum between the protocone and hypocone.

Palatine: The anterior end of the palatine (Fig. 1 and Fig. 2) is situated at the level of the mesial end of the M1 (c. 99[0]). This bone is eroded but complete, and the posterior border of the palatine, slightly deformed, is rounded (c. 97[0]) and situated 13 mm behind the M3. The posterior moiety of the palatine displays a central shallow depression that is 11 mm in width and 20 mm in length. Posteriorly, the palatine constitutes, at both sides, a long process of posterolateral direction that progressively widens to contact anteriorly the pterygoid process.

Nasal: Only the left nasal is preserved (Fig. 1 and Fig. 2), although it is eroded. It is quadrangular and displays a small anterolateral process.

Frontal: The left frontal is complete, whereas the right one is only partially preserved (Fig. 1 and Fig. 2). It is narrow in its posterior portion, gradually widening anteriorly until constituting the supraorbital process (c. 44[0]). The latter process displays a prominent and dorsoventrally-flattened posterolateral corner (c. 36[0]), and a thickening on the anterolateral corner. A mirror reconstruction based on the left side indicates that the two frontals together would have displayed an overall Y-shape, with an estimated maximum width of 130 mm.

Parietal: Both parietals are preserved, although more fragmentarily on the right side. The braincase is long and narrow, with subparallel lateral sides that broaden only slightly towards the occipital region (Fig. 1B). Dorsally, the two temporal lines shape an elongated and narrow X (Fig. 2B). The sagittal and frontoparietal sutures are practically fused, although their course can be still discerned at several points (so as to provide a reliable reconstruction). The temporal crests are not prominent and moderately curved (type B of Domning, 1988: fig. 3) and closely approach each other (without contacting) along 2.5 cm on their dorsal-most portion. At the intersection of the sagittal suture with the lambdoid crest, there is a shallow, arrowhead-shaped concavity (30 mm long) on the interparietal area (which looks asymetrical due to distortion). Overall, in lateral view, the cranial vault displays a slightly dorsally convex profile (Fig. 1 and Fig. 2), with the highest point being situated just posterior to the frontal.

Lacrimal: The lacrimal (Fig. 1 and Fig. 2) is fragmentary and little informative. It contacts the jugal bone, forming the anterior border of the orbit. Due to incomplete preservation, it is not possible to ascertain whether the lacrimal contact the premaxilla or not (c. 93[?]). The lacrimal foramen (c. 91[0]), situated anteriorly from the orbit, is oval (7.1 mm long and 4.5 mm wide).

Jugal: The anterior portion of this bone displays a triangular section, with a thin edge on its external portion (c. 88[0]; Fig. 1 and Fig. 2). Due to preservational reasons it is unclear whether the jugal contacts the premaxilla (c. 87[?]). At about the posteroventral edge of the orbit, the jugal becomes dorsoventrally highest (c. 85[0]), with a maximum height of 28 mm behind the orbit (c. 89[0]). The jugal further narrows posteriorly, giving rise to a long and tapering zygomatic process (Fig. 1 and Fig. 2).

Squamosal: The articulation area with the parietal and supraoccipital is very fragmentary and not informative. In dorsal view, the zygomatic process of the squamosal (91 mm long) is straight, and in lateral view its anterior end forms a triangular process above the jugal (c. 81[0]). Medially, the zygomatic process of the squamosal is not swollen (c. 84[0]), while posteriorly it widens dorsoventrally and curves in posterodorsal direction, reaching to a point higher than the dorsal border of the orbital rim and extending to the temporal crest (c. 76[1]). A processus retroversus is present, forming a small bulb moderately inflected in the dorsal moiety of posterior end of the zygomatic process (c. 77[1]; Fig. 1 and Fig. 2).

Occipital region: The supraoccipital (Fig. 3) displays a marked trapezoidal shape, being narrower on the dorsal than on the basal margin (c. 64[1]; maximum width 53 mm), and displaying very rounded contours. The low, nearly inconspicuous external occipital protuberance divides the supraoccipital into two shallow oval concavities. The parietal-supraoccipital angle is 127°. The right exoccipital is quite completely preserved and only slightly deformed, unlike the left one, which is much more fragmentary. A shallow supracondylar fossa is observed dorsal to the condyle (c. 67[1]), with an ellipsoidal shape and a mesioventral to dorsolateral orientation. The dorsolateral border of exoccipital is rounded and seems continouous until the condyle (c. 70[0]). The preserved occipital condyle is 34 mm in height and 20 mm in maximum width. The two exocipitals contact each other above the foramen magnum along 11 mm (c. 66[0]), and overall have an estimated maximum width of 63 mm. The foramen magnum is rounded with a dorsal peak or indentation, with an estimated maximum width of 19 mm and a height of 26 mm.

Alisphenoid: This bone can only be observed on the right side, where it is however fragmentary. Its lateral surfaces are smooth, and the alisphenoid canal is absent (c. 101[1]).

Pterygoid: It is preserved on both sides, although more clearly observable on the right (Fig. 1 and Fig. 2). In lateral view, it displays an inverted triangular outline and a deep pterygoid fossa (c. 102[1]). Its distal tip is rounded with a short posteriorly-oriented hamular process (c. 105[1]).

Basioccipital and basisphenoid: The basicranium is very incompletely preserved, because only some fragments of the basioccipital and basisphenoid, which are not informative, can be observed on the right side (Fig. 1 and Fig. 2). These two bones appear not to be completely fused with one another, although their fragmentary preservation precludes an unambiguous assessment.

The cranium IPS81881a displays the diagnostic features of the Dugongidae (see Domning, 1994), such as the lack of the alisphenoid canal, an extension of the squamosal up to the temporal crest, a well-developed processus retroversus, and absence or loss of P5. Its attribution to genus Prototherium is further justified by its elongate and narrow cranial shape (Samonds et al., 2009 and Zalmout and Gingerich, 2012). Other features proposed by Bizzotto, 1983 and Bizzotto, 2005, such as the presence of seven postcanine teeth (i.e., presence of four permanent premolars and three molars in adults) and a longer M3 compared to the preceding upper molars, are also present in the new species. However, these features are also displayed by other Eocene sirenians and thus not diagnostic of Prototherium. In turn, the presence of an angular process in the posterior edge of the mandible, previously considered diagnostic of Prototherium (de Zigno, 1887) but in fact absent from P.? intermedium s.l. (Bizzotto, 1983, Bizzotto, 2005 and Pilleri et al., 1989), cannot be ascertained in IPS81881a (which lacks the mandible). Similarly, other cranial features previously considered diagnostic of Prototherium, such as a zygomatic-orbital bridge higher than the palate or the length of the nasals roughly one-third of total skull length (Sickenberg, 1934), are not reliable because they are too variable among species currently assigned to this genus (Bizzotto, 1983, Bizzotto, 2005 and Pilleri et al., 1989) and doubtful in its use.

An attribution of IPS81881a to genus Prototherium is further confirmed by the fact that the new species differs from other Eocene dugongid genera in several respects. Thus, compared to the monotypic genus Sirenavus Kretzoi, 1941, IPS81881a differs by diplaying shorter nasals, and a rectangular (instead of ellipsoidal) cranial vault in dorsal view (see Kordos, 2002 and Kretzoi, 1941: fig. 1b). In turn, IPS81881a differs from the scarce material of Anisosiren Kordos, 1979 in the smaller upper molars and the less lophodont occlusal pattern of the check teeth (Kordos, 1979: 323). It differs from species of Eosiren Andrews, 1902 by the shallow instead of deep supracondylar fossa and the shorter (not elongated) M1 and M2 (Zalmout and Gingerich, 2012). An attribution of IPS81881a to Halitherium taulannense Sagne, 2001 can be discounted based on the lesser rostral deflection, lack of marked temporal crests and the anterior margin of the palatine not extending to the mesial level of the M1. Finally, the described specimen differs from species currently assigned to genus Eotheroides (see Samonds et al., 2009) in the shorter nasals of the former as well as in the position of the premaxillary-maxillary suture (located behind the premaxillary symphysis in the new species, instead of below as in Eotheroides). However, the new species shares with E. aegyptiacum (type species of the genus) several features that are lacking in other species of Prototherium, such as the slightly concave posterior border of the palatine (more incised in P. veronense and P.? intermedium) or the shallow supracondylar fossa of exoccipital (absent in P. veronense). In spite of these similarities with Eotheroides, an ascription of the new species to this genus instead of Prototherium does not seem advisable in the light of the differences with Eotheroides noted above, at least according to how these genera are currently conceived.

IPS81881a further resembles other species of Prototherium by not retaining the DP5 after the eruption of the upper molars (resulting in an adult dental formula with seven instead of eight postcanine teeth), although the condition of several species of Prototherium in this regard has proven controversial. The presence of a P5 has been clearly documented in Prorastomidae and Protosirenidae (Domning, 2001b, Domning and Gingerich, 1994, Domning et al., 1982 and Savage et al., 1994). Among Eocene dugongids, the retention of the DP5 (without substitution by the P5) is quite frequent, being documented in Eotheroides (Abel, 1913, Samonds et al., 2009 and Zalmout and Gingerich, 2012), Eosiren (Andrews, 1902 and Domning et al., 1994), Halitherium taulannense (Sagne, 2001) and probably Sirenavus (Kordos, 2002). In contrast, the lack of DP5 in adults is unusual among Eocene sirenians. The latter condition is displayed by some individuals of Eotheroides aegyptiacum described by Abel (1913: 352) and, according to Bizzotto (2005), all the species of Prototherium would similarly lack a DP5 in adults. Regarding the type species of the genus, however, Sickenberg (1934) favored the retention of a DP5, based on the presence of an empty space between the inferred P3–P4 and M1–M3 preserved series, but Bizzotto (2005) questioned this interpretation and concluded that P. veronense would also lack a DP5. Pilleri et al. (1989), in turn, proposed a dental formula with eight postcanine teeth (retaining the P5 in adults) for P. solei, based on their reconstruction of the maxilla. However, our inspection of the paratype MGSB48657-4 indicates that the preserved alveoli cannot be reliably assigned to incisors, canines or single-rooted premolars and that a diastema is displayed mesially from the molars. As such, we consider more likely that this nominal taxon does not differ from other specimens of P.? intermedium in this regard, thereby supporting the contention that they are best synonymized. At present, however, it is uncertain whether the apparent lack of DP5 in adult specimens of Prototherium spp. (including the holotype of the new species described here) has any taxonomic value, being alternatively interpretable as an ontogenetic change with no phylogenetic significance.

The description of a new species, Prototherium ausetanum sp. nov., for IPS81881a is justified by the fact that the described holotype cranium differs in various respects from all other species currently included (even if tentatively) in the genus Prototherium. The holotype of the new species, based on condylobasal length, is slightly smaller (9%) than P. veronense and larger (22%) than P.? i. intermedium (on the basis of adult specimens with erupted but not very worn M3; see Fig. 5 for further details). It also displays a relatively wider skull than these taxa (both at the level of the supraorbital process and of the zygomatic arches; Table 1) and a slightly concave posterior edge of the palatine (more incised in P. veronense, although not observable in P.? intermedium). Prototherium ausetanum sp. nov. also retains a lacrimal foramen, although this feature cannot be ascertained in other species of the genus (Domning, 1994). The presence of a supracondylar fossa further distinguishes the new species from P. veronense (Domning, 1994). Finally, P. ausetanum sp. nov. further differs from all other species of Prototherium in the occlusal morphology of the M3, which distally displays a postcingulum with two cuspules that encloses a labially open valley. This morphology is not present in other species of Prototherium, which merely display a small cuspule distally from the metaloph forming the postcingulum (Bizzotto, 1983, Pilleri et al., 1989 and de Zigno, 1875).

Compared to other species of Prototherium, the cranial vault of IPS81881a (Fig. 5B) displays a rectangular shape with the two temporal lines shaping an elongated and narrow X, thus more closely resembling the condition of P. veronense (Fig. 5A; de Zigno, 1875). In this regard, the new species differs from P.? intermedium (Fig. 5D), which displays protruding temporal crests that result in a marked depression in between, as well as a marked widening on the posterior arms of the X (Bizzotto, 2005: pl. 2B). However, it should be taken into account that cranial vault shape is quite variable among sirenians, so that the above-mentioned differences can be even observed among specimens of a single species (e.g., Domning, 1988) and, a such, they have little taxonomic value. Prototherium ausetanum sp. nov., in any case, also more closely resembles P. veronense in displaying a moderate degree of rostral deflection (47° for the new species and 50° for P. veronense, whereas P.? intermedium shows a more marked deflection of 67°), an elongated but small rostrum relative to cranium (being enlarged in P.? intermedium), as well as a trapezoid supraoccipital (instead of quadrangular as in P.? intermedium; Pilleri et al., 1989: pl. XVI; Bizzotto, 2005: fig. 4A). The rounded foramen magnum of P. ausetanum sp. nov. further differs from that of P.? intermedium (which is ellipsoidal; Bizzotto, 2005: fig. 4B), but it cannot be confidently compared with that of P. veronense due to distortion in the latter. In contrast, P. ausetanum sp. nov. more closely resembles P.? intermedium in the short nasals, contrasting with the longer condition displayed by those of P. veronense (presumably representing the plesiomorphic condition; Domning, 1994), as well as in the position of the zygomatic-orbital bridge (at about the level of the palate, instead of higher above the latter as in P. veronense; Domning, 1994).

With regard to the two nominal species previously described by Pilleri et al. (1989) from Eocene remains of two Catalan localities, due to preservational reasons, detailed comparisons are only feasible with P. solei. This taxon, originally described as a distinct species of Prototherium from the Bartonian of Tona, was subsequently synonymized with P.? intermedium by Domning, 1996 and Domning, 2014, whereas Bizzotto (2005) merely distinguished these taxa at the subspecies rank. Given that P. solei only differs from the nominotypical subspecies in size and a few other morphologic features of little taxonomic value (Bizzotto, 2005), with no differences in the character codings for the matrix employed in this paper (at least, regarding those features preserved in both nominal taxa), here we concur with Domning (2014) that P. solei is best considered a junior subjective synonym of P.? intermedium, in spite of the age differences between their respective type localities. Given that both P. ausetanum and P. “solei” come from the same area and are older (from the Bartonian) than the type locality of P.? intermedium (from the Priabonian), the differences between the new species and specimens of P.? intermedium previously attributed to P. solei deserve particular attention. Thus, in spite of a similar overall cranial size, the holotype of P. ausetanum sp. nov. differs from P. “solei” (paratypes MGSB48657-1 and MGSB48657-4) in several cranial measurements (Table 1), with the former displaying much longer parietals as well as a much wider skull (both at the level of the supraorbital processes and at that of the zygomatic arches). Moreover, the skullcap of IPS81881a further differs from that of the paratype MGSB48657-1 (see Pilleri et al., 1989: pl. 5), which displays (like other specimens of P.? intermedium) thickened temporal crests that flank two deep depressions along the sagittal line (one in the frontoparietal region, and the other on the interparietal region). These depressions, almost continuous in P. “solei”, are almost lacking in P. ausetanum sp. nov., which merely displays a shallow interparietal depression. Furthermore, P. ausetanum sp. nov. also differs in some dental features from the paratype of P. “solei” MGSB48657-4 (see Pilleri et al., 1989: pl. 2), with the latter displaying a more trapezoidal (instead of rectangular) occlusal profile of the M2 and further lacking the distinctive morphology of the M3 of P. ausetanum sp. nov. It is not possible to ascertain differences in the degree of rostral deflection, since it cannot be measured in any of the type specimens of P. “solei”.

In turn, the holotype (and only available specimen) of P.? montserratense (MGSB44892), from the Bartonian of Castellbell i el Vilar, consists of a partial skeleton with some craniodental elements, which are however still embeded within the matrix, so that no detailed comparisons are possible. Domning, 1996 and Domning, 2014 synonymized this taxon with P.? intermedium, but in fact our inspection of the holotype of P.? montserratense indicates that it displays no diagnostic features to discount an alternative attribution to P. veronense. Accordingly, we prefer to consider this taxon as a nomen dubium, until further preparation or inspection of the holotype with non-invasive techniques enables the clarification of its taxonomic identity as a distinct, taxonomically valid species. However, while the possibility cannot be currently discounted that P.? montserratense is a junior subjective synonym of P.? intermedium or P. veronense, the inspection of the holotype of the former indicates that this nominal taxon is unlikely to be a senior synonym of P. ausetanum sp. nov., from which it differs by the trapezoidal (instead of rectangular) contour of the M2 and the much wider foramen magnum (see Pilleri et al., 1989: table 9).

The cladistic analysis yields 15 most parsimonious trees of 125 steps: CI = 0.66 (excluding uninformative characters); RI = 0.80; RCI = 0.56. The topology of the strict consensus cladogram (Fig. 6) recovers the monophyly of dugongids and shows the same branching order for non-dugongid sirenians as in Vélez-Juarbe and Domning (2014), i.e., with prorastomids, protosirenids and trichechids being the successive sister taxa of dugongids. Among the latter, our analysis fails to recover the monophyly of most of the included genera (Prototherium, Eotheroides, Eosiren and Halitherium). The new species of Prototherium is recovered as the sister taxon of Eotheroides aegyptiacum (the type species of genus Eotheroides). In turn, the type species of Prototherium (P. veronense) constitutes a basal polytomy with Eotheroides lambondrano, the clade of P. ausetanum sp. nov. + E. aegyptiacum, and that including the remaining dugongids (other halitheriines plus hydrodamalines and dugongines, including P.? intermedium), which differ from the more plesimorphic dugonginds Prototherium and Eotheroides in the larger rostrum relative to the cranium. In other words, our analysis does not resolve the relationships between the species of the basal dugongid genera Prototherium and Eotheroides, except for the fact that P.? intermedium occupies a more derived position (as the sister taxon of Halitherium taulannense). This suggests that P.? intermedium should be removed from genus Prototherium, although we refrain from providing an alternate genus ascription until the genus Halitherium, which apparently is not monophyletic (Vélez-Juarbe and Domning, 2014 and Voss, 2013), is revised. In fact, the results of our analysis further highlight that the more plesimorphic genera Prototherium and Eotheroides are also in need of revision. A more comprehensive analysis, including additional species of Eotheroides, would be required to more confidently resolve the phylogenetic relationships between these taxa. In the meantime, we prefer to include the newly described species into Prototherium, based on the morphological features discussed in the preceding section.