Trachytherus, AMEGHINO, 1889

Billet, Guillaume, Muizon, Christian De & Quispe, Bernardino Mamani, 2008, Late Oligocene mesotheriids (Mammalia, Notoungulata) from Salla and Lacayani (Bolivia): implications for basal mesotheriid phylogeny and distribution, Zoological Journal of the Linnean Society 152 (1), pp. 153-200 : 156-172

publication ID	https://doi.org/ 10.1111/j.1096-3642.2007.00388.x
persistent identifier	https://treatment.plazi.org/id/03AA3A66-FFD9-FE57-FE83-BFD6A06635E3
treatment provided by	Felipe
scientific name	Trachytherus
status

Treatment

TRACHYTHERUS AMEGHINO, 1889

Trachytherus Ameghino, 1889a: 1–8 , non Trachytherium Gervais, 1849, Ameghino, 1889b: 919, plate 79 figs 1, 2, plate 97, fig. 3.

Proedium Ameghino, 1895: 623–624 , non Proedrus Förster, 1888.

Eutrachytherus Ameghino, 1897: 427–428 , nomen illegit. pro Trachytherus .

Proedrium Ameghino, 1897: 429 , lapsus pro Proedium .

Ameghinotherium Podestá, 1899: 1–8 .

Coresodon Ameghino, 1901: 374 (partim).

Isoproedrium Ameghino, 1904: 171 , nomen illegit. pro Proedium .

Anatrachytherus Reguero & Castro, 2004: 45–64 .

Type species: Trachytherus spegazzinianus Ameghino, 1889 .

Type locality: undetermined locality of Santa Cruz or Neuquén provinces, Argentina; see part on Trachytherus spegazzinianus .

Comments: As the subfamily ‘Trachytheriinae’, the genus Trachytherus probably represents a paraphyletic assemblage (see Discussion). The species of Trachytherus recognized in this study are:

Trachytherus spegazzinianus ( Ameghino, 1889a, b) a senior synonym of Anatrachytherus soriai ( Reguero & Castro, 2004) and Trachytherus curuzucuatiense ( Podestá, 1899) (expecting further remains, see Discussion).

Trachytherus ? mendocensis Simpson & Minoprio, 1949 ( Simpson & Minoprio, 1949; Simpson et al., 1962), but see Discussion.

Trachytherus subandinus Villarroel, Sempéré and Marshall, 1994 ( Villarroel et al., 1994).

Trachytherus alloxus sp. nov. (this study).

TRACHYTHERUS ALLOXUS SP. NOV.

FIGURES 1–15 View Figure 1 View Figure 2 View Figure 3 View Figure 4 View Figure 5 View Figure 6 View Figure 7 View Figure 8 View Figure 9 View Figure 10 View Figure 11 View Figure 12 View Figure 13 View Figure 14 View Figure 15 (EXCEPT 12B), APPENDICES 1–4 ( TABLES A1–A 3)

Trachytherus spegazzinianus Hoffstetter 1968: 1096 (partim).

Trachytherus sp. A MacFadden et al. 1985: 242, fig. 10.

Trachytherus sp. B MacFadden et al. 1985: 242, fig. 10.

Trachytherus spegazzinianus Sydow 1988 : PhD (partim).

Trachytherus sp. Marshall & Sempéré 1991: 635, Villarroel et al. 1994: 31.

Trachytherus spegazzinianus Shockey 1997a : PhD (partim).

Trachytherus spegazzinianus Reguero 1999 : PhD (partim).

Trachytherus spegazzinianus Reguero & Castro 2004: 45–64 (partim).

Trachytherus sp. nov. Reguero & Castro 2004: 45–64 (but no mentioned specimens).

Trachytherus spegazzinianus Shockey et al. 2007: 227–247 (partim).

Holotype: MNHN-BOL-V 006355: subcomplete skull with right and left I1, P2-M3 ( Fig. 2A–C View Figure 2 ); nasals missing.

Paratypes: SAL 7: broken rostrum with left I1–2 and P1-M3 and right I1 and P2-M1, right part of auditory region with epitympanic sinus, external auditory meatus and postglenoid complex, left humerus.

SAL 11: broken mandible with left i1 and broken i2 and p3, and right broken i1–2, p3–m3.

SAL 191: upper and lower jaw in occlusion with upper right dP2–3, upper left dI1, dC, P1, dP2–4, lower right di1–2, p1, and lower left di1–2, p1, dp2–4.

SAL 235: broken mandible with right i1–3, p3-m1, and left i1–2, p2-m1.

SAL 279: subcomplete skull with right P4-M3 and left M1–3.

MNHN-BOL-V 006356: subcomplete skull with left M2–3 and right M1–3, and proximal epiphysis of femur.

MNHN-BOL-V 003825: associated fragments of upper and lower jaws, with sectioned I1, erupting I2, alveolus I3-C, P1, dP2–3 roots, dP4, M1–2 left and right, dp3 roots, dp4, m1–2 left and right.

Referred specimens: Appendix 1.

Locality: Salla, Department of La Paz, Bolivia.

Distribution: Deseadan SALMA, late Oligocene. Specimens were collected in different parts of the Salla section ( MacFadden et al., 1985) but a great majority of them are from the unit 5, Branisella level.

Etymology: For alloxa, meaning abundant in Aymara language, to refer to the great number of specimens found in Salla.

Diagnosis: Differs from Trachytherus spegazzinianus in: smaller size; loss of M1 central fossette occurs when M2 presents contacting protoloph and metaloph or isolated central fossette (M1 central fossette always disappears before protoloph and metaloph contacts lingually in M2 of T. spegazzinianus ); molars proportionally labio-lingually narrower; mesial lobe of the upper molars (protoloph) little rounded lingually after moderate wear, with a pointing disto-lingual extremity (as for deciduous premolars) and much larger than the distal lobe (unlike in T. spegazzinianus ); lingual sulcus of the upper molars shallower and lack of lingual groove when central fossette is isolated; M3 poorly trilobed before enclosing the central fossette (M3 distinctly trilobed and with oblique mesial and distal faces in T. spegazzinianus ); protoloph of M1 with smaller and much less persistent lingual enamel folds.

Differs from Trachytherus ? mendocensis in: larger size; P3 with a central fossette contemporaneous with M1–2 opened labially or even with M1 with isolated central fossette and M2 opened labially; upper premolars parallelogram-shaped, not triangular; poorly marked median lobe on upper molars, especially on M3.

Differs from Trachytherus subandinus in: larger size; upper premolars parallelogram-shaped, not triangular to pentagonal.

Differs from the undetermined (and undescribed) taxon reported by Shockey et al. (2006) from Moquegua Peru in its larger size.

Skull

Patterson (1934c) provided a usefull description of the skull of Trachytherus spegazzinianus ; however, it is not exhaustive and many features that were likely to be insignificant to Patterson are treated here. The following description is also a partial review of cranial anatomical observations on mesotheres made by Patterson (1934a, c, 1936, 1952). When appropriate, the cranial anatomy of Trachytherus alloxus sp. nov. is compared with that of Pseudotypotherium pseudopachygnathum (Ameghino, 1904) (in Patterson, 1934a, 1936). Direct comparison is also made with the MNHN specimens of Mesotherium cristatum Serres, 1867 and Plesiotypotherium achirense Villarroel, 1974 .

Nasal–facial region ( Figs 2B, C View Figure 2 , 3 View Figure 3 )

This region presents a moderate degree of transversal narrowing across the P1, a condition which is much more developed in later mesotheriids. The anterior part of the premaxillaries, below the narial aperture, shows an interesting feature for this species. Patterson (1977) argued that ‘in nearly all notoungulates, the anterodorsal portion of the premaxillaries anterior to the narial opening extends upward into a median process’. This process is usually formed by medial union of two small bulges formed by each premaxillary. However, this structure, which could be called the narial process of the premaxillaries, is either poorly marked or absent in T. alloxus , in young individuals (SAL 803) as well as in older ones (SAL 283, SAL 732, MNHN-BOL-V 006355). This condition may constitute an exclusive character of T. alloxus within mesotheriids although it strongly varies in size among notoungulates.

In anterior view, the narial aperture has a piriform shape (reversed pear). Dorsally, the anterior extremity of the nasals tapers anteriorly and protrudes well anterior of the premaxillary process. On the rostrum, the premaxillary–maxillary suture is subvertical in its ventral part. In the dorsal part, it is more slanting and posteriorly orientated. It meets the nasals approximately above the level of the mesial region of dP 2 in young individuals (SAL 803) and most often above the mesial part of P 3 in adults (SAL 280, SAL 289). Except for the anterior tapering, the nasals have roughly subparallel edges. The fronto-nasal suture is slightly posterior to the level of the anterior orbital rim, except on juvenile individuals (SAL 297), where it lies at this level. In all ontogenetic stages for which this feature is known, the frontals present a median anterior process wedged between the nasals ( Fig. 3 View Figure 3 ). This frontal anterior process is present in all individuals but its size varies.

The infraorbital foramen is situated above M 1 in adult specimens (SAL 237, SAL 280, SAL 289, SAL 296) although this condition may vary slightly because of the anterior displacement of cheekteeth (see Ontogeny). In the juvenile individual SAL 297, it is above dP3.

As in ‘archaeohyracids’, hegetotheriids and other mesotheriids, the lacrymal has an important facial extension. Both this bone and the frontal exclude the maxillary from the superior orbital rim. The posterodorsal part of the maxillary (the ascending process) is reduced to a narrow strip of bone between the nasal and lacrymal, and, even if not far from it, it does not reach the posteriormost point of the nasals. More ventrally, the maxillary is also excluded from the orbital rim by the jugal, which contacts the lacrymal on the rim. The lacrymal foramen opens just in front of the orbital rim, in the inferior part of the lacrymal bone.

The anterior root of the zygomatic arch is opposite to M1–2. At this point, the arch presents ventrally a wide, flat area. It is lateraly bordered by a small crest running from the level of M3 to the anteriormost point of the arch. A this point, this crest forms a protuberance probably homologous with the welldeveloped descending process of the interatherids ( Patterson, 1934c). Patterson (1934c), although stating that the zygomatic arch of Trachytherus was very different from that of mesotheriines, interpreted this flat area as an incipient development of the zygomatic plate of mesotheriines. This feature has been subsequently used in later phylogenetic studies ( Cifelli, 1993). However, the structure of the zygomatic arch of T. alloxus (and T. spegazzinianus ) does not differ radically from that observed in nonmesotheriid typotherians such as Ultrapithecus , Oldfieldthomasia or Archaeohyrax . In fact, T. alloxus only presents a plesiomorphic condition of the anterior root of the zygomatic arch, which shows a lateral expansion as in many typotherians. The development of a true zygomatic plate occurred later in mesothere evolution. For example, Eotypotherium chico Croft, Flynn & Wyss, 2004 , a basal mesotheriine ( Croft et al., 2004), presents no such structure.

Palate ( Figs 2A View Figure 2 , 4 View Figure 4 )

Just behind the overgrown incisors, the incisive foramina present a particular configuration, characteristic of mesotheriids. They are triangular and their posterior ends are pointed and converge posteromedially ( Fig. 4 View Figure 4 ).

The premaxillary–maxillary suture is largely separated from these foramina. It is sigmoid; laterally, it is gently convex anteriorly and medially, and is strongly concave anteriorly. Medially, this suture runs forward and comes nearer to the incisive foramina.

The major palatine foramina open at the level of the mesial edge of M1, well anterior to the maxillary– palatine suture ( SAL 10 , SAL 237, SAL 280, MNHN- BOL-V 003478 ). They are closer to the tooth row than to the medial suture of the maxillae. Only one specimen ( SAL 10 ) presents a minor palatine foramen just ahead of the maxillary–palatine suture, at the level of the mesial region of M2. The palate is broad and, because the tooth rows slightly diverge posteriorly, it is wider posteriorly than anteriorly. The palatine extension on the palate is reduced, and the palatal portion of the palatines is wider than long. Anteriorly, they do not extend beyond the distal edge of M1. Transversely, they almost reach the lingual side of M3. The maxillary–palatine suture on the palate is somewhat parabolic .

Medially to the distal edge of M3, the palate is always but inconsistently excavated. This fossa partially corresponds to the postpalatal notch that separates M3 from the pterygoid apophysis of the palatine. Between the postpalatal notches, the palatine expands posteriorly and forms a wide medial triangular platform. This platform expands the palate posteriorly up to the level of the anterior edge of the glenoid cavity. T. alloxus has a narrower median platform than later mesotheres such as Plesiotypotherium and Mesotherium but is anteroposteriorly longer. The posterior extremity of this platform is formed by two large and strongly diverging palatine processes. Between the palatine processes the ventral edge of the choanae is markedly W-shaped, i.e. parabolic (opening posteriorly) with a small caudal spine medially.

Orbitotemporal region and skull roof ( Figs 2–4 View Figure 2 View Figure 3 View Figure 4 )

The zygomatic arch is roughly horizontal and has a straight inferior edge (unlike Mesotherium ) in its anterior part, below the orbit. Posterior to the orbit, the arch is oblique, and raises toward the top of the skull. On the zygomatic arch, the jugal constitutes a thin lateral blade of bone contacting the lacrymal and maxilla anteriorly; it occupies the entire height of the arch in the middle region, and only the ventral part posteriorly. The medial face of the zygomatic arch is formed by the maxillary (anteriorly across its full height, and posteriorly, the ventral part only) and the squamosal (posteriorly, the dorsal part only). Posteriorly, the jugal almost reaches the anterior level of the posterior zygomatic root. The jugal–squamosal suture is approximately horizontal on the oblique portion of the arch. The squamosal does not extend forward beyond the level of M3. Medially, the maxillary also bears a long contact with the squamosal; it almost reaches the posterior extremity of the jugal ( Fig. 4 View Figure 4 ).

The dorsal rim of the arch is continuous with the lateral part of the lambdoid crest. The lambdoid crest is well developed and medially meets the strong sagittal crest butting against it.

Between the orbits, the frontals widely expand dorsally. This region is well preserved on the holotype. Dorsal openings of the supraorbital foramina are well marked, close to each other. The postorbital processes are robust, large and triangular; they nearly contact the zygomatic arch in its oblique part. Posterior to the supraorbital processes, the frontals are strongly constricted. Posterior to this constriction is the braincase and the skull widens considerably. The frontal– parietal suture is poorly distinct, the bone being almost fused in the holotype. It is located between the postorbital apophyses and the postorbital constriction and is approximately transversal .

The parietal–squamosal suture is also poorly marked on SAL 279; it extends antero-posteriorly and is concave laterally. The sagittal crest is thick and well developed. It extends posteriorly up the lambdoid crest. On the skull roof, the squamosals form most of the lateral portion and the parietals are limited to a median position. Accessory vascular foramina are present on these bones.

SAL 237, MNHN-BOL-V 005710 and MNHN- BOL-V 006356 document the internal anatomy of the orbit, although its preservation is often very poor. The crown of M3 occupies a large place in the orbital floor. In front of this, a small suborbital fossa is limited anteriorly and externally by the orbital and zygomatic rims; it is much less developed than in Plesiotypotherium . At the bottom of the orbit, an internal groove occupies the place where various foramina usually open (infraorbital foramen, sphenopalatine foramen, caudal opening for the major palatal canal), which are not dictinct in T. alloxus .

No bone sutures can be observed in the orbit of the available specimens. MNHN-BOL-V 006356 presents a peculiar triangular area located just anteriorly to the posterior root of the zygomatic arch. The nature of this structure is difficult to determine, although it may represent some muscular attachment. It is underlined by a crest running from the basis of the postorbital apophysis and ending at the ventral extremity of this triangle.

Basicranium and auditory region ( Figs 4–6 View Figure 4 View Figure 5 View Figure 6 )

The description of this region is mostly based on SAL 279 ( Fig. 5A–C View Figure 5 ). Posterior to the postpalatal platform, and lateral to the choanae, a thick and large crest extends posterolaterally. This crest is made of the palatine on most of its length and its apex, posteriorly, is formed by the alisphenoid. This apex has been erroneously identified by Gabbert (2004: fig. 14.4) as pterygoid. In fact, the pterygoid is medial to this crest as clearly seen on the skull of Puelia sp. (MLP 67 II 27 27) figured by Gabbert (2004: fig. 14.4). Although not exclusively made of the alisphenoid and including an important portion of the palatine, this crest is in a similar position to the ectopterygoid crest of the alisphenoid as defined by Novacek in Leptictis ( Novacek, 1986: fig. 14). Because this crest is mainly formed by the palatine ventrally we refer to it herein as the palatine crest. It bears a well-marked muscle scar, which probably received part of the origin of the pterygoideus medialis as is observed in many recent ungulates. Apparently, the entopterygoid crest of the pterygoid and its hamular process (as defined by Novacek, 1986: fig. 14) are missing on SAL 279. The entopterygoid crest and the hamular process are clearly preserved on the skull of Puelia sp. (MLP 67 II 27 27) figured by Gabbert (2004: fig. 14.4) on the medial side of the palatine crest.

On the posterior edge of the postpalatal platform and dorsal to the palatine crests are two deep fossae excavated in the alisphenoid. This structure has been described by Gabbert (2004) in the Toxodontia as the scaphoid fossa. According to Gabbert (2004), the scaphoid fossa of notoungulates receives the origin of the tensor veli palatini muscle. However, the pterygoid fossa of ungulates and primates also receives the pterygoideus medialis and therefore it is probable that the scaphoid fossa of notoungulates also received the origin of this muscle. Therefore, the scaphoid fossa of Gabbert is probably homologous to the pterygoid fossa of other ungulates (see Barone, 1997, for a study of modern domestic ungulate anatomy).

The pterygoid fossa of Trachytherus alloxus is excavated in the alisphenoid. It is located dorsally to the palatine crest. It is reniform and is limited laterally and medially by a thin crest of the alisphenoid. The lateral edge of the pterygoid fossa is strongly concave laterally and reaches ventrally the apex of the palatine crest (i.e. the alisphenoid portion of the palatine crest). The ventral part of the pterygoid fossa is deep and more excavated than its posterodorsal portion. The latter receives the sphenotympanic fissure.

Because of the posterior extension of the postpalatal platform, the choanae open well behind the M3 and the ventral edge of the choanae is close to the level of the anterior edge of the glenoid cavities, while it is well anterior in Plesiotypotherium and Mesotherium . The basisphenoid forms the medial floor between the pterygoid fossae. This bone is narrow anteriorly, but widens posteriorly. Bordering the medial crest of the pterygoid fossa is a small but distinct groove excavated in the basisphenoid for the passage of the Eustachian tube.

The basioccipital is completely preserved on SAL 279 only, although on this specimen this bone is poorly preserved. Only the anterior part of the basioccipital is preserved on MNHN-BOL-V 006355, a specimen in which the basisphenoid–basioccipital suture is unfused. It is roughly straight and located at the level of the anterior angle of the tympanic bullae ( Fig. 4 View Figure 4 ). Just behind this suture are two symmetrical bulges delimitating a medial groove on the basioccipital for the attachment of the longus colli muscle. At the posterior end of the bone is an elevated triangular area, which bears a sharp crest extending anteriorly approximately 15 mm toward the longus colli tubercles. Because the skull SAL 279 is slightly crushed and distorted, the detailed observation of the anatomy of its basicranium is difficult. In particular, the foramina of this region are sometimes distorted or obliterated. Two foramina can be observed on the left side of the basioccipital: the hypoglossal foramen and the posterior lacerate foramen. The former is elongated anteroposteriorly, a condition that could be due to distortion. Three or four millimeters anterolateral to it, at the posteromedial angle of the tympanic and at the anteromedial angle of the base of the paroccipital process, is positioned the posterior lacerate foramen, which is approximately twice as large as the hypoglossal foramen. The posterior external carotid foramen is situated at the anterior edge of the posterior lacerate foramen and is confluent with it. This condition differs from that in Plesiotypotherium , where the carotid foramen is distinctly separated from the posterior lacerate foramen. However, it is noteworthy that the position of the posterior lacerate foramen of T. alloxus resembles that of Plesiotypotherium , Mesotherium and Pseudotypotherium in being not transversally elongated and not bounded anteriorly by the posterior vertical wall of the bulla (because it is medial to it). This condition differs from non-mesotheriid notoungulates.

At the anterolateral edge of the tympanic is an elongated opening, which corresponds to the sphenotympanic fissure as defined by Gabbert (2004) for the Toxodontia. It corresponds to the coalescence of the median lacerate foramen and the foramen ovale. This opening was designated by Patterson (1934a, 1936) as the ‘foramen ovale + foramen lacerum medium’ or simply foramen ovale for the mesotheriine Pseudotypotherium pseudopachygnathum . At the anteromedial angle of the sphenotympanic fissure and at the anteromedial angle of the tympanic is a small, circular foramen located at the posterior end of the eustachian groove on the basisphenoid and, which is therefore regarded as the eustachian foramen.

The bullae are roughly triangular. Their anterior angle is located at the posterior end of the eustachian groove of the basisphenoid; their posteromedial angle is at the anteromedial angle of the paroccipital process and their posterolateral angle forms the posterolateral angle of the skull. The crista meatus is posteroventral to the postglenoid process of the squamosal and corresponds to the ventral edge of the tympanic, which is pressed against the post-tympanic process. It extends from the anterior edge (level with the tympanohyal recess) of the bulla to the ventral surface of the external auditory meatus. At this point, the crista meatus thickens and develops a prominent ventral tubercle. This feature is a common trait for mesotheres, e.g. Pseudotypotherium , and is interpreted as a generalized feature for the family ( Patterson, 1936). This condition was also observed in Archaeohyrax (G. Billet, unpubl. data). Furthermore, Patterson (1936) also considered this feature to be a similarity between the Mesotheriidae and Toxodontidae . However, this resemblance is superficial and may represent a convergence as the two structures are rather different in the two families. In toxodontids it is a sharp spine of the crista meatus rather than a thickened tubercle as in mesotheriids.

Participation of a lip from the crista meatus to the ventral border of the external auditory meatus, mentioned by Patterson (1936) for mesotheriids, is also observed in T. alloxus . The edges of the external auditory meatus are somewhat wrinkled and swollen, a condition also present in mesotheriines. In front of the external auditory meatus opens a small suprameatal foramen. Such a foramen has previously never been mentioned for any mesotheres. It is located laterally on the posterior face of the postglenoid process and it has a similar position with the suprameatal foramen of some toxodontians ( Gabbert, 2004).

The tympanohyal recess, also called vagina processus hyoidei ( Patterson, 1932), is bounded anterolater- ally by the crista meatus, medially by the bulla and posterolaterally by the post-tympanic process of the squamosal. This recess is very deep as in all mesotheres (but also in toxodontids and some notohippids). The post-tympanic process and the crista meatus are oppressed against each another; their contact is oblique and extends between the tympanohyal recess and the external auditory meatus. It is as long as in Plesiotypotherium , but it is much longer than in Mesotherium . The stylomastoid foramen generally opens in the cleft between these two structures, approximately midway between the tympanohyal recess and external auditory meatus. The stylomastoid foramen of Trachytherus alloxus is embeded in that cleft. It is much more apparent in Mesotherium cristatum .

The paroccipital processes are broken at their apices. They are markedly separated from the tympanohyal recess. Their medial border is rather flat and anteroposteriorly orientated. Their section is roughly triangular and their lateral edge bears a low ridge, which extends dorsally on the tympanic and connects with the post-tympanic process, forming a semicircular crest on the posterior edge of the bulla. The external auditory meatus is located high on the skull, well above the occipital condyles. This condition is similar to the other mesotheriids and to hegetotheriids and Archaeohyrax but differs from the toxodontids, where the meatus is located much higher. In many other notoungulates, the external auditory meatus is usually located lower than in the abovementioned taxa.

Laterally, a groove is visible in a conical depression anterolateral to the tympanohyal recess. It is a slightly bowed laterally and obliquely orientated. Its lateral edge presents a small ridge probably made by the squamosal as in Plesiotypotherium and Mesotherium . At its anterior end is a foramen, which we interpret as the opening of the canal of Huguier, and which has been observed in many notoungulates.

As in most notoungulates, the dento-squamosal articulation is not a true glenoid cavity. It is composed of two articular facets separated by a small fossa called the mandibular fossa by Gabbert (2004). The anterior facet is slightly saddle-shaped and not distinctly concave. It is transversely elongated, being approximately three times wider than long. The posterior facet is half the width of the anterior and articulates with the posteromedial region of the dentary condyle. The two facets are not in contact (these facets are illustrated with dotted lines on the right side of the skull in Fig. 4 View Figure 4 ). The mandibular fossa is narrow and shallow and opens laterally. It is consistently present in notoungulates. The posterior facet is located on the anterior edge of a poorly individualized postglenoid process, which protrudes more laterally than ventrally. This structure, which is a broad bulge rather than a process, is located on the edge of the squamosal intermediately positioned between the anterior dentosquamosal facet and the external auditory meatus. Posteromedially to the postglenoid process, is a wellindividualized postglenoid foramen. It is of large size and transversely extended; it opens posterolaterally. The position of the postglenoid foramen varies within mesotheriids, and even within a single mesotheriid species or between the two sides of the skull of the same individual ( Villarroel, 1974). However, a general pattern has been found in mesotheriines, in which the postglenoid foramen is often ‘divided into two parts by a small bridge of bone lying transversely across it’ and situated medially or anteromedially to the postglenoid process as for Pseudotypotherium ( Patterson, 1934a) or Plesiotypotherium ( Villarroel, 1974) . In non-mesotheriine mesotheres (trachytheriines), this foramen has been described in a slightly more posterior position in T. spegazzinianus as compared with mesotheriines ( Patterson, 1934c). A similar condition is present in T. alloxus , as mentioned above. Furthermore, in mesotheriines such as Plesiotypotherium , the postglenoid foramen is situated just ventral to the opening for the canal of Huguier. In T. alloxus , it is approximately at the height of the canal of Huguier but lateral to it.

The petrosal is observable only on its cerebellar aspect on some specimens. It is in situ on the skull MNHN-BOL-V 006356. Other isolated petrosals (MNHN-BOL-V 006906, MNHN-BOL-V 009027, MNHN-BOL-V 005037) were found together with other remains of Trachytherus and are tentatively associated with them. They are all still attached to the tympanic. Although damaged, MNHN-BOL-V 006906 ( Fig. 6 View Figure 6 ) is reasonably informative. It is very similar to the petrosal of Plesiotypotherium . It differs from Mesotherium cristatum in its larger and deeper fossa subarcuata, and its less pronounced trigeminal impression. The internal auditory meatus opens immediately ventral to the fossa subarcuata and not antero-ventral as in Mesotherium . The posterior slit, intermediate in height between the fossa subarcuata and the internal auditory meatus, is tentatively identified as the vestibular aqueduct. This specimen is broken into parts that might concern the cochlear aqueduct.

Occiput

The foramen magnum and the occipital condyles do not significantly differ from other notoungulates and especially other mesotheriids. SAL 279 presents the internal concavity in the postero-dorsal part of the squamosals in which the epitympanic sinuses are usually positioned. In T. alloxus , they are distinctly above the level of the occipital condyles. The outline of the occipital in a posterior view has the shape of a broad hourglass. The bottleneck is bordered laterally by the epitympanic sinuses. On SAL 279, the posterior walls of the epitympanic sinuses have been sunk into the sinus itself during fossilization. Therefore, the lateral regions of the occipital face of SAL 279 are deeply concave. The dorsal part of the occipital bears a deep triangular fossa probably for the nucchal ligament (at least in part). The mastoid foramen opens just laterally to each curve of the bottleneck. A thin strip of bone is present medially to this foramen.

Lower jaw ( Fig. 7 View Figure 7 )

The symphysis is markedly procumbent and is excavated by a large gutter dorsally behind the incisors ( Fig. 7 View Figure 7 ). Posterior to the incisors and canine (when present, see Dentition) and in front of the premolars, the mandible presents a small transverse constriction, much less extended antero-posteriorly than in later mesotheres. There are two major mental foramina in T. alloxus as in Plesiotypotherium and Mesotherium . The anterior one is situated below the premolars, and the posterior one below the molars. Their position varies because of the inconsistency of cheekteeth positioning (see Ontogeny). The corpus of the mandible is stout and wide in adults, whereas it is much more slender in juveniles. Below the last molar, there is a small bulge, just anterior to a slightly depressed area that corresponds to a narrowing of the mandible leading to the ascending ramus. This region of the mandible is unknown in T. alloxus .

Dentition

General comments: Mones (1982) has proposed the terms protohypsodont for the high-crowned and limited growth teeth and euhypsodont for the highcrowned and ever growing teeth. However, Mones introduces, as he says, a ‘phylogenetic content’ in this terminology ( Mones, 1982: 110). Therefore, we prefer the terminology of Simpson (1970), which is more descriptive: ‘hypsodont (is) a tooth that eventually develops one or more roots but that has a crown definitely higher than those roots or than any of its horizontal dimensions and (...) hypselodont (is) a tooth that never forms a root but continues to grow and to extrude new parts from the alveolus throughout life’.

Upper incisors and canine ( Fig. 8 View Figure 8 )

The first incisor is hypertrophied and, probably, hypselodont ( Fig. 8 View Figure 8 ). It lacks enamel on all its lingual face. Its occlusal plane is ovale-shaped to subtriangular, pointed distally. The tooth is obliquely implanted and well separated from its counterpart. I2 is small and oval-shaped, like I1, but less obliquely implanted than the anterior tooth. It is separated from I1 by a small diastema. As with I1, it lacks enamel on the lingual face. It is hypsodont but undoubtedly rooted. When present (SAL 281), I3 is a vestigial tooth that resembles a small I2. It is separated from the latter by a small diastema on SAL 281. I3 is rarely present and this may depend on ontogeny (see below). No specimens seem to have a permanent canine except possibly SAL 7. Nevertheless, an alveolus for that tooth is frequently visible. It is possible that it was present in early wear stages, and subsequently lost because of a weak embedding in the maxillary. This may be a variable feature for this species.

Upper premolars ( Figs 9–11A, B View Figure 9 View Figure 10 View Figure 11 )

P1 is a small tooth. It is unworn on SAL 1022; at this stage, P1 presents a well-developed ectoloph, separated from a lingual relief by an important mesiodistal sulcus. The absence of wear on this erupting tooth precludes other observations. When slightly worn (SAL 803), P1 presents a weak premolar pattern with a tiny mesially individualized parastyle extension. With increasing wear, this tooth becomes featureless and is rapidly worn off and expelled from the tooth row. The posterior premolars show a gradual morphological trend. They resemble each other strongly but increase in size greatly posteriorly. These teeth exhibit a parastyle/paracone groove in the mesial part of their labial face. They are quadrangular, with a slightly rounded lingual face. Their labio-lingual width markedly increases as wear proceeds. MNHN- BOL-V 009027 exhibits erupting premolars ( Fig. 9A View Figure 9 ). On this specimen, all premolars present a distal edge of cingulum that increases from P2 to P4. The occlusal surface of P3–4 resembles the typical ‘face pattern’ of fossettes of typotherians ( Cifelli, 1993) but with a labio-lingually orientated lingual aperture when unworn. With wear (SAL 280) labial accessory fossettes disappear and the central fossette (‘the mouth’) is isolated ( Fig. 10 View Figure 10 ). Slightly worn P2 shows a more basic pattern with an ectoloph widely separated from the lingually continuous protoloph and metaloph ( Fig. 9A View Figure 9 ). Between these two lophs is a valley corresponding to the isolated central fossette. Thus, an isolated central fossette is present on all moderately worn premolars. Its orientation is oblique. It increases in size from P2 to P4.

With increasing wear, the fossette disappears on both premolars ( Fig. 9B View Figure 9 ). In the same way, the paracone/parastyle groove disappears in older individuals. The premolars of T. alloxus are strongly hypsodont ( Fig. 11A, B View Figure 11 ). They have a continuous growth but tend to form roots in old individuals (SAL 718, SAL 739, SAL 740).

Upper molars ( Figs 2A View Figure 2 , 11C, D View Figure 11 , 12 View Figure 12 )

The upper molars are built on the same general pattern. They are highly hypsodont with a strongly curved crown ( Fig. 11C, D View Figure 11 ). Nevertheless, they tend to develop roots in very old individuals (SAL 880). They are markedly larger than the premolars.

In early wear stages, the lingual border of M1–2 curves inward and presents three different lingual lobes: the largest anterior one is formed by the protoloph; the median one is formed by the crista2- crochet union (see below), and the posterior lobe comprises the metaloph and the distal cingulum (see below). The protoloph is well separated from the median lobe by an important oblique lingual sulcus. A much smaller sulcus separates the median lobe from the metaloph lingually. As wear proceeds ( Fig. 12A View Figure 12 ), the median lobe reduces and the two others increase lingually until contacting each other. In fact, the size of the protoloph increases more than that of the metaloph. At this stage, the protoloph most often presents an approximately straight lingual face with a pointed linguo-distal extremity, whereas it was more rounded in early stages. In this respect, Trachytherus alloxus differs from Trachytherus spegazzinianus in which the mesial and distal lobes (protoloph and metaloph) are markedly rounded and subequal ( Fig. 12B View Figure 12 ) even after moderate wear. At this stage (i.e. lobes just in contact), the lingual face of the M1 protoloph in T. alloxus may exhibit very small and hardly persistent enamel infoldings. With increasing wear, a large fossette (the central fossette) is formed by enclosure of the two sulci mentioned above. It has the shape of a boomerang pointing lingually with the distal arm being shorter than the mesial one ( Fig. 12A View Figure 12 ). The posterior arm rapidly disappears with wear, leading to a straight oblique fossette. After heavy wear, the central fossette disappears, producing a featureless occlusal surface. Likewise, as wear proceeds, a change of shape occurs: in M1, the mesiodistal length decreases while the labio-lingual one increases; in M2, no general tendency can be observed for the mesio-distal length, but the labio-lingual width increases (see sections on ontogeny and the effect of tooth wear and discussion of measurements).

Some specimens [SAL 902, SAL 1020, SAL 786 ( Fig. 11C, D View Figure 11 )] show erupting M1 or M2 on which the unworn occlusal pattern is visible. This provides important information (as well as the deciduous premolars, see below) to infer mesothere relationships. These erupting teeth show the presence of a distal edge of cingulum in the earliest stages, a widely spread feature in basal notoungulates. Likewise, unworn or little worn M1–2 display a small fossette labial to the crista 2–crochet junction. This fossette emphasizes the distinction between crista 2 and the metaloph, only joined originally by the crochet. This fossette is probably homologous to the posterior labial fossette of the typical ‘face’ pattern in typotherians ( Cifelli, 1993). There is no trace of an anterior labial fossette resulting from the lingual junction of the protoloph and the crista 1. If it still existed, it would be the feature least resistant to wear on upper molars in all mesotheriids. At this early wear stage, the median lobe is well developed, better than in many other stages. It is composed of attached crochet and crista 2 [SAL 786 ( Fig. 11C, D View Figure 11 )]. At this stage, this structure is attached to the metaloph by a very narrow isthmus running from the crochet. It enlarges with additional wear. With increasing wear, the distal cingulum disappears. The posterior labial fossette may persist after the loss of the distal cingulum, as is observed on the M2 of SAL 808, although it is somewhat inconsistent.

M3 presents the general pattern discussed above for M1–2 but with small differences. Even on an erupting M3, the posterior labial fossette and distal cingulum have not been observed. M3 wears first in its mesial part, while its distal extremity is still unerupted ( Fig. 2A View Figure 2 ). A strong anteroposterior wear gradient characterizes the eruption of M3. When the complete tooth is subjected to wear, M3 differs from M1 and M 2 in its less individualized median lobe because of a shallower posterior sulcus between it and the metaloph. It marks also a clear difference with Trachytherus spegazzinianus , the median lobe of which on M3 is clearly individualized although it is smaller than that of M1-M2 (as in T. alloxus ). The next wear stage is the isolation of the central fossette. This stage is marked by the lingual junction of the protoloph and metaloph, with the protoloph being much more developed than the metaloph. At this stage the M3 acquires a subtriangular shape. The newly isolated central fossette is straight and oblique with a minute posterior arm. Simultaneously, in the labial part of the distal face, a posterior expansion is individualized. Its increases throughout later wear stages as the central fossette reduces and finally disappears. Thus, M3 also reaches a featureless morphological stage in old individuals, except for an important distal extension formed by the metastyle.

The anterolabial angle of each molar slightly overlaps the preceding tooth labially.

Deciduous upper teeth ( Fig. 13 View Figure 13 )

Some specimens (SAL 191, SAL 803, MNHN-BOL-V 004271) exhibit upper deciduous incisors that greatly resemble their definitive counterpart, except in their smaller size. The first incisor is strongly developed and differs from its permanent counterpart by being clearly less curved. As mentioned above, the only canines observed are probably deciduous. They are minute, rooted and larger mesio-distally than labio-lingually.

The upper deciduous premolars constitute a gradual series. The unworn deciduous premolars found clearly present the face pattern of fossettes of typotherians ( Cifelli, 1993). All seem to have the same original occlusal pattern and merely differ in size, time of eruption and other minor differences ( Fig. 13A, B View Figure 13 ). Thus, on dP2, in early stages of wear (SAL 187), an anterior labial fossette is present, isolated by the junction of the protoloph and crista 1 ( Fig. 13B View Figure 13 ). This fossette is hardly distinguishable on an erupting dP4 (SAL 898) and rapidly disappears with wear.

The unworn dP3–4 presents an unusual central fossa. This fossa has wrinkled borders because of the presence of multiple minute crests on its edges (SAL 898). This fossa rapidly disappears with wear. It originally isolates the lingual part of the median lobe (crochet and crista 2) from the ectoloph. These structures are tightly linked to each other by crista 2, thus dividing the fossa ( Fig. 13A View Figure 13 ). Therefore, the posterior part of the fossa is homologous with the posterior labial fossette described above on the molars. Moreover, the sulcus separating lingually the protoloph from the crista 2–crochet extends anterolabially deeply within the tooth crown. It then turns posterolabially and reveals the crista 1–crista 2 disjunction. This corresponds to the anterior part of the fossa. Apparently, the crista 2–crochet or median lobe is not directly attached to the metaloph in early wear stages ( Fig. 13A View Figure 13 ); the junction of these two structures occurs lower on the tooth.

The distal cingulum is visible on dP3 (SAL 898) and dP4 (SAL 207). It results in the isolation of a small fossette with additional wear ( Fig. 13A View Figure 13 ). The deciduous premolars always present a mesially well-expanded parastyle (particularly on dP2), clearly distinguishing them from their permanent counterpart.

In addition to these features, the deciduous premolars resemble the molars and present the same morphological succession pattern versus wear, i.e. isolation of an oblique central fossette follows the original stage of a trilobate tooth with lingual aperture (this original stage is not present on dP2). None of the specimens shows a featureless occlusal surface, and the central fossette does not disappear even in advanced stages of wear.

These teeth are rectangular in early stages of wear, while they are more quadratic in advanced stages in dP3–4; dP2 always has a rounded shape with an important parastylar extension.

Both molars and premolars (deciduous and permanent) lack enamel mesially and distally on a small labial strip.

Lower incisors and canine ( Fig. 7 View Figure 7 )

The first and second pair of incisors are large ( Fig. 7A View Figure 7 ), occupying together approximately the same space as the first pair of upper incisors. The i1 is slightly longer mesio-distally than i2, but it may be the opposite on old individuals (SAL 11). Both are highly hypsodont, rooted in oldest individuals (SAL 11), and present a pronounced vertical groove dividing their lingual face, when slightly worn. These teeth are strongly procumbent. They lack enamel on their mesial and distal faces.

A vestigial and rooted tooth is adjacent to i2 until moderate wear stages. A similar condition is observed in T. spegazzinianus . Patterson (1934c) identified it as an i3 or a canine, but seemingly preferred the second hypothesis. However, it is more likely to be the third incisor given its adjoining position to i2 and its separation from p1 by a diastema. Moreover, the third incisor and the canine are both present (but broken) on SAL 858. This specimen is less worn than others without canines. Thus, the canine would only be present in early stages of wear of the adult dentition and was probably expelled rapidly.

Lower premolars ( Figs 7A View Figure 7 , 14A View Figure 14 )

A complete first premolar is not visible in any specimen. In SAL 200, p1 is erupting. It is pointing mesially, with a crest descending distally to what might be considered a talonid. This tooth is rooted and small. It is absent in old individuals such as SAL 8 or SAL 11 (see part ontogeny).

All p2–4 are bilobate, with a labial sulcus marking the separation of the two lobes (equivalent to trigonid and talonid). On p2 and p3, the anterior lobe is always longer than the posterior one. On p4, the posterior is the longest during early stages of wear ( Fig. 7A View Figure 7 ), but this condition is reversed in later stages ( Fig. 14A View Figure 14 ). Unworn permanent premolars (except p1) have not been observed in the MNHN and MNHN-BOL-V collection and the first stage known for p2–4 is the one with an isolated trigonid–talonid fossetid. This fossettid has a transverse (labio-lingual) orientation. It disappears with strong wear and the bilobate pattern is hardly visible or totally absent. The p2–4 are hypsodont but rooted in old individuals.

Lower molars ( Figs 7A View Figure 7 , 14B View Figure 14 , 15 View Figure 15 )

The molars are markedly longer than the premolars and the talonid is much longer than the trigonid, contrary to the condition of the premolars.

Very little worn m1s present isolated trigonid and talonid ( Figs 14B View Figure 14 , 15A, B View Figure 15 ). The edges of the trigonid are convex labially and concave lingually. The disto-lingual extremity of the metalophid is slightly pointed. There is a small fossa in the middle of the occlusal surface of the trigonid. This is produced by the lingual closing of the trigonid, probably due to an expansion of a premetastylid structure ( Reguero & Castro, 2004). The talonid is primarily composed of a labial convex hypolophid and a symmetrical entolophid. These two structures seem to be linked both distally and mesially ( Fig. 15A, B View Figure 15 ), which produces a central fossa on the talonid in these early stages. At this stage of wear, there is a broad lingual groove on the distal part of the talonid that emphasizes the distal prolongation of the hypolophid beyond the entolophid. The same pattern is observed also on slightly worn m2 (SAL 817).

On more worn molars, the trigonid and talonid fossae disappear as well as the above-mentioned lingual groove. The trigonid and talonid first contact lingually, which results in a deep labial sulcus separating the trigonid and the talonid across the tooth except for a thin lingual isthmus ( Fig. 15C View Figure 15 ). This condition corresponds to the permanent mesotheriine configuration. As wear proceeds, the trigonid and talonid also come into contact labially, isolating the trigonid–talonid fossetid. This first appearance of a trigonid–talonid lingual connection is a derived feature in mesotheriids, but is also present in hegetotheriids. The molar trigonid–talonid fossettid differs from that of the premolars in being oblique whereas it is more transverse on the premolars ( Fig. 7A View Figure 7 ). Also, at this stage of wear, the molars become bilobate both labially and lingually.

With increasing wear, the fossetid disappears and the occlusal surface becomes featureless ( Fig. 14A View Figure 14 ). The m1 widens transversely (labio-lingually) with increasing wear, but does not show any mesio-distal change. The second molar widens as well, and also slightly lengthens. The third molar shows the same trend with a more important mesio-distal lengthening (see Discussion on measurements).

The lower molars are highly hypsodont. No roots have been observed on any specimen, but this part of the crown is less available for observation than in upper molars.

Deciduous lower teeth ( Figs 14B View Figure 14 , 15A, B View Figure 15 )

Only SAL 200 preserves di1–3 and dc. The deciduous i1–2 are similar but less strong than their permanent counterpart. The third deciduous incisor and the deciduous canine are minute incisiform teeth. They occupy a very small space and are rapidly taken off the teeth row. Moreover, their presence in early wear stages is inconsistent.

SAL 200 also exhibits an unworn dp2. It presents a clear distinction between trigonid and talonid. These two lobes are originally linked labially on dp2, unlike on molars and also on dp3–4 (see below). The paralophid is very poorly developed and is just represented as a small mesio-lingual fold. The protoloph is mesiodistally orientated and joins the transverse metalophid distally. There is an important mesio-lingual aperture of the trigonid due to the weak development of the paralophid. The talonid shows a similar structure to that described for the erupting m1 above, except for the labial attachment to the trigonid. Lingually, there is a prominent space separating trigonid and talonid.

The available third and fourth deciduous premolars present, when slightly worn, no trigonid–talonid attachment ( Fig. 15A View Figure 15 ). The first junction between these two lobes is lingual, as on the molars. A lingual sulcus persists on the trigonid until moderate stages of wear. Likewise, a wide and shallow groove is present on the distal part of the talonid lingual face in early stages of wear. With increasing wear, a trigonid– talonid transverse fossetid becomes isolated and all other features disappear ( Figs 14B View Figure 14 , 15B View Figure 15 ). The mesiodistal length of these teeth decreases as wear proceeds.

All cheek teeth present a thick layer of cement, even the deciduous one, but its presence is somewhat inconsistent.

Postcranial skeleton

Numerous postcranial remains are present in the MNHN collection. Nevertheless, the postcranial skeleton of this species has been previously described, especially in terms of functional adaptations ( Shockey et al., 2007), under the name Trachytherus spegazzinianus , and thus are not repeated here.

Ontogeny and the effect of tooth wear (Appendix 2) Upper jaw (Appendix 2A): Sixteen different ontogenetic stages were recognized by us on the upper dentition. As wear proceeds, I3, C and P1 are generally taken off the tooth row by anterior displacement of distal teeth. This is produced by the increasing area occupied by distal cheek teeth (especially because of the eruption of M3?) thus reducing available space for I3–C or P1. Changes in upper molar shape are significant, as shown by the measurements of mesio-distal and labio-lingual width at different stages of wear (see the discussion section on measurements and Appendix 2).

Lower jaw (Appendix 2B): We have recognized 11 different ontogenetic stages for the lower jaw. As for the upper teeth, some lower teeth (i3–p2) tend to be taken off the tooth row. It is very difficult to determine if c or p1 is taken off first (i3 and p2 remain longer in the tooth row). One interesting observation is the forward shifting of the cheek teeth through ontogenetic stages as shown by a relative comparison of the cheek teeth position versus that of the posterior mental foramen. In early stages (SAL 899, SAL 900), this foramen is situated below the mesial or middle part of dp4. In more advanced stages (SAL 870), it is below the anterior part of m1, and then (SAL 235, SAL 733) below the distal part of m1. Finally, in old individuals (SAL 11), this foramen is situated just below the distal part of m2. The changes in dimensions versus wear also exist for lower molars. However, the trend for m1 and m2 are not as clear as that for their upper counterpart (see Appendix 2).

MNHN-BOL-V 003825 and MNHN-BOL-V 004711 are the only specimens presenting associated upper and lower jaws in this study. They show the correspondence of upper jaw stages 7 and 12 with lower jaw stages 5 and 8, respectively (Appendix 2).