Craniodental anatomy of late Oligocene archaeohyracids (Notoungulata, Mammalia) from Bolivia and Argentina and new phylogenetic hypotheses

Transcription

1 Zoological Journal of the Linnean Society, 2009, 155, With 27 figures Craniodental anatomy of late Oligocene archaeohyracids (Notoungulata, Mammalia) from Bolivia and Argentina and new phylogenetic hypotheses GUILLAUME BILLET 1 *, BRYAN PATTERSON and CHRISTIAN DE MUIZON 1 1 UMR5143 (MNHN, CNRS, UPMC), Département histoire de la Terre (CP 38), Muséum National d Histoire Naturelle, 57, rue Cuvier, Paris, F France Received 5 September 2007; accepted for publication 29 November 2007 The rich Deseadan fauna from the locality of Salla Luribay (Bolivia) documents the last occurrence of archaeohyracids, a poorly known group of small typotherian notoungulates. Exceptionally, archaeohyracid remains are extremely abundant in the Salla deposits and are assigned to a single new species Archaeohyrax suniensis sp. nov. The anatomy of the new taxon is presented and the ontogeny of the cheek teeth is described. Archaeohyrax patagonicus Ameghino, 1897 from the Deseadan of Patagonia is also redescribed and compared with the new Bolivian species. Additionally, juvenile teeth of Sallatherium altiplanense (Hegetotheriidae) are described because they provide crucial phylogenetic information for understanding the phylogeny of archaeohyracids. A cladistic analysis performed on archaeohyracids and hegetotheriids supports for the first time the existence of a clade of late archaeohyracids (post-mustersan), which is the sister taxon of all hegetotheriids. It also argues for the origin of mesotheriids within archaeohyracids and for the existence of a hegetotheriine clade. These conclusions fit well with temporal data known for each taxa. Nevertheless, the present analysis also underlines the fact that the lack of data concerning the cranial anatomy of many archaeohyracids (Eohyrax, Pseudhyrax, Archaeotypotherium, Protarchaeohyrax) weakens the phylogenetic signal The Linnean Society of London, Zoological Journal of the Linnean Society, 2009, 155, ADDITIONAL KEYWORDS: Archaeohyrax cranium Deseadan Hegetotheria ontogeny Sallatherium systematics teeth. INTRODUCTION The Deseadan (late Oligocene) Salla Beds of Bolivia have yielded an extremely abundant mammal fauna (Hoffstetter, 1968, 1969). They have been assigned an age of between 27 and 25.8 Myr for the richest fossiliferous levels (Kay et al., 1998). Notoungulates remains are especially abundant in these deposits. The Notoungulata are, among the four groups of *Corresponding author. billet@mnhn.fr Bryan Patterson died in He has been the author of numerous valuable papers on South American mammals between 1931 and His contribution to the present study comes from his unpublished manuscript on Deseadan archaeohyracids in which he notably described Archaeohyrax patagonicus. Biography link: South American endemic mammals, the taxonomically richest and often the most abundant in Cenozoic palaeofaunas. Therefore, knowledge of this group is of crucial importance for our understanding of the singular history of South America in Early Cenozoic times. Most of the notoungulate fauna from Salla has been described: notohippids (Shockey, 1997), hegetotheriids (Reguero & Cerdeño, 2005), leontiniids (Shockey, 2005), interatheriids (Hitz, Billet & Derriberry, 2008) and mesotheriids (Billet, Muizon de & Mamani Quispe, 2008). However, the most abundant of all notoungulates, and indeed of all mammals, in Salla (MacFadden et al., 1985) have not thus far been described. These are notoungulates commonly referred to an enigmatic family, namely the Archaeohyracidae Ameghino, Little is known of this 458

2 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 459 peculiar family, remains of which are usually very scarce in South American palaeofaunas. Their abundance in the Salla Beds is an exception that may underline some palaeoenvironmental or taphonomic peculiarities. Archaeohyracidae are part of the Typotheria, which include small to medium-sized notoungulates exhibiting the following apomorphic traits: a rodentlike emphasis on the anterior dentition with I1 somewhat enlarged and upper molars presenting a characteristic face pattern of fossettes (Cifelli, 1993). Oldfieldthomasiidae, Archaeopithecidae, Interatheriidae, Mesotheriidae and Hegetotheriidae are also part of Typotheria (Cifelli, 1993). Among these, Hegetotheriidae and Archaeohyracidae are considered by Cifelli (1993) and Croft et al. (2003) as a clade the Hegetotheria an entity that was previously regarded as a separate suborder by Simpson (1967). The evolution of archaeohyracids is poorly understood and the definition of Archaeohyracidae is unclear (see Croft et al., 2003: 32, for a list of the currently recognized archaeohyracid species). They are known from the Casamayoran [probably middle or late Eocene (Kay et al., 1999)] to the Deseadan, but many recent studies have suggested that another group of notoungulates, the hegetotheriids, directly arises from the latest archaeohyracids (Cifelli, 1993; Hitz, 1995; Croft, 1998, 2000; Reguero, 1999; Croft et al., 2003). Therefore, Archaeohyracidae are considered to be paraphyletic. Identification of archaeohyracids remains in collections is difficult. In the collections of the Paris and La Paz Museums, for example, many specimens referable to archaeohyracids were labelled as interatheriids, mesotheriids, hegetotheriids and even notohippids. Even the holotype of the new species described herein was referred to an interatheriid. Archaeohyracid remains from Salla currently represent the last record of this group. They were preliminarily studied by Reguero & Cifelli (1997) who referred them to two species: Archaeohyrax sp. nov. and Protarchaeohyrax sp. nov. (see also Croft et al., 2003). However, MacFadden et al. (1985) previously referred these remains to a unique new species of Archaeohyrax. More than 250 archaeohyracid specimens from Salla were available to us for study. Among this sample, we recognized a single new species of Archaeohyrax. Below, we describe in particular a partial skull and several rostrums that provide critical new information and characters. The large number of specimens also provided crucial ontogenetic data. These are particularly relevant to our tentative delimitation of intra-specific variation of cheek teeth within this new species. The type species of Archaeohyrax, A. patagonicus Ameghino, 1897, is based on a holotype composed of an associated cranium and mandible. These were figured by Ameghino (1897) in the original description. However, neither the drawings nor the text was accurate enough to describe the specimen correctly and precisely. Here, we thoroughly describe this specimen and incorporate data from an unpublished manuscript of one of us (B.P.). A direct comparison is made with the new species from Salla. In light of the knowledge brought by these descriptions, we performed a cladistic analysis on archaeohyracids and hegetotheriids. The description of juvenile teeth of the hegetotheriid Sallatherium altiplanense Reguero & Cerdeño, 2005 provided important new data and phylogenetic information. The new hypotheses of relationships obtained among archaeohyracids and hegetotheriids are discussed, and compared with temporal data. MATERIAL AND METHODS The studied material belongs to the American Museum of Natural History, New York, USA (AMNH), the Field Museum of Natural History, Chicago, Illinois, USA (FMNH), the Museo Argentino de Ciencias Naturales Bernadino Rivadavia, Buenos Aires, Argentina (MACN), the Museo de La Plata, Argentina (MLP), the Salla Montehermosan and Santa-Cruz collections of the Muséum national d Histoire naturelle (respectively MNHN-SAL referred to as SAL in the text, MNHN-MHR as MHR and MNHN-SCZ referred to as SCZ), Paris, France, the Museo Nacional de Historia Natural, La Paz, Bolivia (MNHN-BOL), the Princeton University collection housed in the Yale Peabody Museum, Yale University, New Haven, CT, USA (YPM- PU), and the Florida Museum of Natural History, Gainesville, FL, USA (UF). Teeth measurements were taken to the nearest 0.1 mm at the occlusal surface. The mesio-distal length was the maximal length on the labial face, and the labio-lingual width is perpendicular to this (Fig. 1). When information provided by Bryan Patterson in his unpublished manuscript were not verified by us, we refer to these personal observations as Patterson, ms. List of anatomical abbreviations: al., alisphenoid; ant-lab. fos., anterior labial fossette; basiocc., basioccipital; bsph., basisphenoid; bul., tympanic bulla; bul. cons., bullar constriction; can. Hug., canal of Huguier; car. f., carotid foramen; centr. fos., central fossette; centr. val., central valley; cro., crochet; cr. 1, crista 1; cr. 2, crista 2; cr. 2-cro., united crista 2 and crochet; c. m., crista meatus; dist. cing., distal cingulum; d. pr. mx., descending process of the maxillary; entlphid.,

3 460 G. BILLET ET AL. Figure 1. Teeth measurements, illustration of the method. A, upper cheekteeth, SAL 189, left maxilla with P3-M3. B, lower cheekteeth, SAL 309, right mandible with p3-m2. Mesial to right. Scale bar = 1 cm. entolophid; ent-hyp. fos., entolophid hypolophid fossettid; epy. sin., epitympanic sinus; e. a. m., external auditory meatus; fr., frontal; gr., groove on the ventral border of the e. a. m.; gr. trig-tal., lingual groove demarcating the trigonid and talonid; hyp. f., hypoglossal foramen; hyplhid., hypolophid; inc. f., incisive foramina; infr. f., infra-orbital foramen; int. fos., intermediate labial fossette; ju., jugal; lac., lacrymal; lac. f., lacrymal foramen; ll M1, first upper molar labiolingual width; ll m1, first lower molar labio-lingual width; maj. pal. f., major palatine foramina; md M1, first upper molar mesio-distal length; md m1, first lower molar mesio-distal length; mtlph., metaloph; mtlphid., metalophid; mx., maxillary; nas., nasal; nas. bu., nasal bulge; occ. cd., occipital condyle; opt. f., optic foramen; os., orbitosphenoid; pa., parietal; parlphid., paralophid; pl., palatine; pmx., premaxillary; poc. ap., paroccipital apophysis; pos-cing. fos. postcingulum fossette; post-lab. fos., posterior labial fossette; prtlph., protoloph; prtlphid., protolophid; pstgl. f., postglenoid foramen; pstgl. pr., postglenoid process; pstpal. pl., post-palatal plate; pst-tymp. pr., posttympanic process; ps. gr., parastyle groove; pt., pterygoid; p. l. f., posterior lacerate foramen; sc. fos., scaphoid fossa; sm. f., supra-meatal foramen; sq., squamosal; styl. f., stylomastoid foramen; suprorb. f., supra-orbital foramina; tri-tal. fos., trigonid-talonid fossettid; tri. fos., trigonid fossettid; tymp. ext., tympanic extension (bounding the tympanohyal recess); tymp. rec., tympanohyal recess; vent. e. a. m., ventral border of the e. a. m. SYSTEMATIC PALEONTOLOGY NOTOUNGULATA ROTH, 1903 TYPOTHERIA ZITTEL, 1893 HEGETOTHERIA SIMPSON, 1945 According to the results obtained from the analysis perfomed therein, the Hegetotheria (sensu Simpson, 1945) are considered to be polyphyletic (see Discussion). ARCHAEOHYRACIDAE AMEGHINO, 1897 According to the results obtained from the analysis perfomed therein, the Archaeohyracidae (sensu Simpson, 1945) are considered to be polyphyletic (see Discussion) ARCHAEOHYRAX AMEGHINO, 1897 Diagnosis: Members of Archaeohyrax are characterized by the presence of a well-marked labial sulcus on the talonid of m3, an unambiguous synapomorphy of our cladistic analysis. This character is also found as a convergence in hegetotheriids. However, the sulcus

4 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 461 is shallower and wider in Archaeohyrax and therefore the two conditions may not be homologous. Other characters may diagnose this clade phylogenetically, but they depend on optimization (see discussion of our phylogenetic analysis). Type species: Archaeohyrax patagonicus Ameghino, 1897 Type locality: Cabeza Blanca (see below), Chubut, Argentina. Included species: Archaeohyrax patagonicus Ameghino, 1897, Archaeohyrax suniensis sp. nov. ARCHAEOHYRAX SUNIENSIS SP. NOV. FIGURES 1 17, APPENDIX S1 4 (TABLES 1, 2, 4) Archaeohyrax sp. nov. Reguero & Cifelli, Protarchaeohyrax sp. nov. Reguero & Cifelli, Archaeohyrax sp. nov. MacFadden et al., Holotype: SAL 4: skull and mandible of a young adult, with right I1, I3-M2, left di1-2, I3, P1-M3, right p1-m2 (m3 is covered with matrix), left p2-m3 (Figs 2, 3). Paratypes: MNHN-BOL-V : skull and mandible associated; SAL 183: rostrum with right and left c-m3; MNHN-BOL-V : mandible with right i1, Figure 2. Archaeohyrax suniensis, cranium of the holotype, SAL 4. A, ventral view. B, dorsal view. C, right lateral view. Scale bar = 1 cm.

5 462 G. BILLET ET AL. p1-m3 and left i1-m3; YPM-PU 23701: rostrum with right and left p2-m3; SAL 310: mandible with right i2, p2-m3 and left p3-m3. Referred specimens: See Appendix S1. Locality: Salla-Luribay, Bolivia. Figure 3. Archaeohyrax suniensis sp. nov., schematic representation of the dentition of SAL 4. A, upper jaw with right I3-M2, left di1-2, I3, P1-M3. B, lower jaw with right p1-m2 (m3 is covered with matrix), left p2-m3. Mesial at the top. Scale bar = 1 cm. Diagnosis: Differs from A. patagonicus in: smaller size; smaller facial extent of the premaxillaries, not extending further posteriorly than the level of C; smaller number of teeth exhibiting simultaneously a fossette or fossetid (i.e. P3-M2 & p3-m2 in A. suniensis; P2-M3 & p2-m3 in A. patagonicus); more hypsodont cheek teeth, almost hypselodont; second mental foramen situated more anteriorly than in A. patagonicus; stylomastoid foramen situated closer to the tympanohyal recess. Etymology: from suni, an Aymara word that refers to the highland Altiplano in Bolivia. Distribution: Deseadan SALMA, late Oligocene. Specimens all come from the Salla section (MacFadden et al., 1985) but most are from unit 5, Branisella level. DESCRIPTION SKULL From a general perspective, the rostrum is only slightly shorter than the distance from the anterior edge of the orbits to the occiput (Fig. 4). It is massive and elevated in lateral view with a convex dorsal edge. In dorsal view the rostrum is triangular and conspicuously narrows anteriorly toward the rather small apical narial opening. At the base of the rostrum, the skull widens distinctly just anterior to the orbit. The orbits are large especially in dorsal view. In lateral view, the massive facial region in relation to the cranial region is readily apparent. The dorsal profile of the skull is concavo-convex with a distinct concavity at the level of the postorbital processes; the dorsal edge of the braincase is as convex as the rostrum. In ventral view the palate is large, especially in relation to the comparatively small teeth. In occipital view the skull is approximately twice as wide as high. Figure 4. Archaeohyrax suniensis sp. nov., paratype, cranium and mandible, MNHN-BOL-V A, right lateral view. B, dorsal view. Scale bar = 1 cm. NASAL FACIAL REGION (FIGS 2, 4 6) The rostrum conspicuously but regularly widens posterior to the canines up to the level of the anterior orbital rim. In lateral view, the elevation of the rostrum decreases anteriorly from the level of P3. On the lateral side of the rostrum, the premaxillary maxillary suture is located above I3. It is straight and

6 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 463 Figure 5. Archaeohyrax suniensis sp. nov., paratype, rostrum, SAL 183. A, ventral view. B, dorsal view. C, right lateral view. Scale bar = 1 cm. Figure 6. Interpretative and simplified partial reconstruction of the skull of Archaeohyrax suniensis sp. nov., right lateral view, from SAL 4, MNHN-BOL-V and SAL 183. Scale bar = 1 cm. vertical in its ventral third; further dorsally it runs posteriorly and is markedly convex posteriorly in its dorsal portion. It meets the nasal dorsally above the level of I3. The outline of the narial aperture is cordiform (MNHN-BOL-V ). Below this aperture, the narial processes of the premaxillaries (Billet et al., 2008) are convergent on the specimens on which they are partially preserved (MNHN-BOL-V , SAL 1010). Dorsally, the nasals are long and narrow with a slight posterior widening. They are posteriorly separated from each other by a medial, triangular to semicircular projection of the frontals (Figs 4, 5). This condition may vary more, as Sinclair (1909) has

7 464 G. BILLET ET AL. Figure 7. Interpretative and simplified reconstruction of the skull of Archaeohyrax suniensis sp. nov., ventral view, from SAL 4, SAL 1010, SAL 183, MNHN-BOL-V and YPM-PU (the pterygoid region is highly simplified because it is unknown in many details). Scale bar = 1 cm. already shown that there is an important variation in the fronto-nasal suture of Hegetotherium. The posteriormost extremity of the nasal does not quite reach the level of the anterior orbital rim. The nasals are markedly arched transversely, this condition being especially pronounced posteriorly, in the vicinity of their suture with the frontals. This gives the skull an original shape with an elevated median bulge (Fig. 6). The anterior extremity of the nasals projects anteriorly beyond the level of the premaxillaries and forms the anterior tip of the skull. The infra-orbital foramen opens above the level of M1 in adults. Its position relative to cheek teeth varies throughout ontogeny with the forward shifting of cheek teeth (see dentition). For instance on SAL 4, a young adult, it opens above the distal part of P4. This foramen is well anterior to the orbital rim. As in hegetotheriids and mesotheriids, the lacrymal bone has an important facial extension, even more so than in these groups. In A. suniensis, it presents a semi-oval shape. The lacrymal foramen opens just in front of the orbital rim, in the inferior part of the lacrymal bone and is surrounded by a small tubercle. Both the lacrymal and the frontal exclude the maxillary from the superior orbital rim and the postero-dorsal part of the maxillary is therefore reduced to a narrow strip between the nasal and the lacrymal. It extends posteriorly approximately as far as the nasals. The frontomaxillary contact is very short. Ventrally, the maxillary is also excluded from the orbital rim by the contact between the lacrymal and the jugal. The anterior root of the zygomatic arch is essentially formed by the maxillary ventrally and medially, and by the jugal, which covers the maxillary laterally and dorsally. The ventro-lateral margin of the arch bears a small crest that ends anteriorly in a small maxillar protuberance. This protuberance may be homologous to the well-developed descending process of the maxillaries in some interatheriines. The anterior root of the zygomatic arch is opposite to M(1)-3. Ventrally, the anterior root of the zygomatic arch

8 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 465 Figure 8. Archaeohyrax suniensis sp. nov., details of the auditory region (right side), SAL 4. A, ventral view. B, postero-lateral (right) view. C, ventro-lateral (right) view. Scale bar = 1 cm. Figure 9. Archaeohyrax suniensis sp. nov., incomplete posterior part of the skull, SAL 551. Ventral view. Scale bar = 1 cm. presents a broad flattened area, a synapomorphy of typotherian notoungulates (Cifelli, 1993). PALATE (FIGS 2, 5, 7) The incisive foramina are situated immediately behind I1. They are incompletely preserved on SAL 4 but appear to have been small. This small size of the incisive foramina is confirmed on SAL 1010, a juvenile specimen tentatively attributed to A. suniensis. Likewise, the extension of the premaxillaries on the palate is reduced: the premaxillary maxillary suture runs medially and slightly backward immediately medially to I3, and rapidly turns forward and directs toward the incisive foramina. As a result, the premaxillaries are medially separated by a large triangular area of the maxillaries (Fig. 7). Behind the canines, the palate, which is delimited by internally concave tooth rows, widens considerably. However, the tooth rows can be rather linear in late wear stage (SAL 313). The major palatine foramina open at the level of P3-4, well ahead of the maxillary palatine suture. The minor palatine foramina open close to the maxillary palatine suture, just ahead of it. The palatal exposure of the palatine bone is a large parabolic surface, which is almost as long as half the length of the palate. Anteriorly, it reaches approximately the edge of P3. As in Trachytherus sp. nov. from Salla (Billet et al., 2008), a post palatal platform extends the palatal floor postero-medially to M2. This platform is directly continuous with the palate and presents as an important posterior extension. The antero-lateral edges of

9 466 G. BILLET ET AL. Figure 10. Archaeohyrax suniensis sp. nov., isolated slightly worn premolars. A, YPM-PU 23788, left P4 (?). B, YPM-PU 23789, right P4 (?). Mesial to left and labial at the top. Scale bar = 1 cm. Figure 11. Archaeohyrax suniensis sp. nov., upper dentition. A, left P3-M3, SAL 189. B, right P3-M3 (reversed), YPM-PU C, rostrum with right and left P2-M3, YPM-PU Mesial to the left. Scale bar = 1 cm. this platform are salient above the postero-lateral portions of the palate. This medial platform widens posteriorly as the palatine crests are markedly divergent. No specimen preserves the complete posterodorsal part of this platform. ORBITOTEMPORAL REGION AND SKULL ROOF (FIGS 4 6) The orbits are greatly enlarged but to a lesser extent than in hegetotheriids. Because of the great develop-

10 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 467 Figure 12. Archaeohyrax suniensis sp. nov., deciduous upper dentition and slightly worn M1. A, left dp3-4, SAL 480. B, right P3, dp4, M1 (reversed), SAL 471. C, right dp2-4, M1 (reversed), YPM-PU Mesial to left and labial at the top. Scale bar = 1 cm. ment of the orbits, the orbital floor is large. It consists of the maxillaries and is a plate-like extension, which covers the coronal part of M2-3. The orbital floor is bordered laterally by the root of the zygomatic arch. In the orbit, a large elongated fossa separates the horizontal orbital floor from the vertical anteromedial orbital wall. The orbital aperture of the infraorbital canal is situated at the antero-lateral extremity of this groove. The sphenopalatine foramen and the caudal opening for the major palatal canal are not distinct within this groove, in which they should open. Anteriorly, the vertical orbital wall is exclusively formed by the lacrymal. The suture of this latter bone with the frontal is straight and orientated vertically from the dorsal orbital rim down to the orbital groove where it contacts the maxillary. At this point, the lacrimal maxillary suture is horizontal, passes above the orbital aperture of the infraorbital foramen and joins the jugal laterally. Posteromedial to the lacrymal, the frontal greatly contributes to the vertical orbital wall. In its posterior part, just above the orbital fossa, the frontal contacts the perpendicular blade of the palatine ventrally (lamina perpendicularis). It is not possible to determine whether the frontal does contact the maxillary anteriorly in this groove or if a thin extension of the palatine prevents this contact. Observation of most orbitotemporal fossa bones is difficult on SAL 4. The morphology seems to be similar to the structures described below for A. patagonicus, in which these bones are more distinct. The orbitosphenoid extends on the vertical orbital wall immediately posterior to the level of M3. Its anterior suture with the frontal is vertical (Fig. 6). Ventrally, it contacts the palatine perpendicular blade. The dorsal part of the fronto-orbitosphenoid suture is nearly horizontal just beneath the postorbital apophysis. The posterior extremity of this suture is not distinct on

11 468 G. BILLET ET AL. Figure 13. Archaeohyrax suniensis sp. nov., partial mandible, MNHN-BOL-V Scale bar = 1 cm. SAL 4 but it probably ends very close to the ventral extremity of the fronto-parietal suture and to the anterior extremity of the parieto-alisphenoid suture, which are not visible. This is due to the presence of a small breakage in this region, posteroventrally to the postorbital apophysis. As mentioned above, there is an oblique contact of moderate length between parietal and alisphenoid. The squamosal does not extend between the parietal and the alisphenoid unlike in hegetotheriids. Ventrally, the squamosal emits a small triangular projection into the alisphenoid, as in hegetotheriids. On the vertical part of the frontal orbitosphenoid suture (on SAL 299, YPM-PU 23701), a small ethmoid foramen is present. Postero-ventrally to it is the large optic foramen (clearly visible on SAL 4), which pierces the orbitosphenoid. The dorsal edge of the sphenorbital fissure [foramen orbitorotundum of Gabbert (2004)] is preserved but its ventral edge and most of the ventral region of the orbitosphenoid and alisphenoid have been broken off prior to fossilization of SAL 4. The sphenorbital fissure is distinctly posterior and slightly ventral to the optical foramen. This region corresponds to the narrowest part of the skull. On the dorsal face of the skull, there is an important flattened area between the orbits, usually called the frontal plate. Near its centre, it is pierced by two supraorbital foramina (Fig. 5B). These are preceded by wide and shallow grooves, which run anteriorly up to the nasofrontal suture. The ventral aperture of these foramina is situated beneath the orbital rim, just anteroventral to the post-orbital processes. The post-orbital processes are complete only on MNHN- BOL-V (Fig. 4B). They are more slender and smaller than in mesotheriids, although they approximate to the zygomatic arch posterolaterally. Posteriorly, the postorbital constriction is somewhat moderate as compared with other notoungulates, except hegetotheriids. The frontal parietal suture lies at this point. It runs posteromedially and follows the temporal lines with which it merges medially. The sagittal crest, formed by the medial fusion of the temporal lines, is well developed (MNHN-BOL-V ) and butts posteriorly against the transverse lambdoid crests. The course of the squamosal parietal suture on the cranial roof is unclear. The parietal extension probably decreases posteriorly. The zygomatic arch is preserved on MNHN-BOL-V Its anterior orientation is horizontal below the orbit (Fig. 4A). It turns oblique behind this up to its posterior root, which is high on the side of the skull. The dorsal crest of the zygomatic arch is continuous with the lambdoid crest on the lateral edge of the pars epitympanicus of the squamosals. Laterally, the jugal squamosal suture runs on the oblique part of the zygomatic arch. Its course is approximately horizontal. The jugal may not reach the glenoid cavity posteriorly. BASICRANIUM AND AUDITORY REGION (FIGS 7 9) As in Trachytherus alloxus (Billet et al., 2008), because of the posterior extension of the postpalatal platform, the choanae open well posterior to M3. No inference can be made on the presence of a hamular process of the pterygoid as its theoretical emplacement is not preserved on any of the specimens. The pterygoid fossa [= scaphoid fossa described by Gabbert (2004) on toxodontians; see Billet et al. (2008)] is not entirely preserved on SAL 4. The only preserved portion in this region of SAL 4 is the ventral part (except the ventralmost extremity) of a transverse blade that buttresses the posterolateral angle of the postpalatal platform (Fig. 7). The unpreserved dorsal part of this blade may have housed most of the pterygoid fossa. What remains of the pterygoid fossa is a conspicuous concavity in the preserved part of the posterior face of the transverse blade. The lateral edge of the blade is straight, unlike that in Trachytherus alloxus. The relative contributions of the alisphenoid and ptery-

12 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 469 Figure 14. Archaeohyrax suniensis sp. nov., lower permanent cheek teeth. A, left p3-m2, SAL 311. B, right p3-m2 (reversed), SAL 309. C, left p3-m3, YPM-PU (partial). Labial to left and mesial at the top. Scale bar = 1 cm. Figure 15. Archaeohyrax suniensis sp. nov., paratype, MNHN-BOL-V Scale bar = 1 cm. goid to this blade cannot be detailed. The abovementioned triangular process of the squamosal participates in the dorso-lateral part of this structure. The apex of the palatine crests (see Billet et al., 2008) is also not preserved. The region anterior to the bulla is not preserved and thus the sphenotympanic fissure (Gabbert, 2004) cannot be described. Most of the remaining auditory region is well preserved on SAL 4 (Fig. 8) and to a lesser degree on SAL 551 (Fig. 9). The medial floor of the basicranium is formed by the basisphenoid and the basioccipital, which define a median plate between the bullae (Fig. 7). The basisphenoid basioccipital suture is straight and transverse. It runs at the level of the anterior third of the bullae. Just behind this suture, the basioccipital presents a notch for the passage of the carotid artery medial to the bulla. A similar condition is observed in Archaeotypotherium and in hegetotheriids. No bulges or groove can be observed in the vicinity of the basisphenoid basioccipital suture. From the basisphenoid basioccipital suture to the foramen magnum, the basioccipital bears a sharp medial crest, the intensity of which is more elevated posteriorly. This crest separates two fossae for the origin of the recti capitis ventralis muscles. Lateral to the greatest width of the basioccipital is a posterior lacerate foramen that is markedly elongated transversally. It is bounded anteriorly by the posterior vertical wall of the tympanic bulla (ectotympanic) and posteriorly and laterally by the exoccipital (paroccipital process). The hypoglossal

13 470 G. BILLET ET AL. Figure 16. Archaeohyrax suniensis sp. nov., variation of the lower cheek teeth morphology. A, left p3-m2 (reversed), MNHN-BOL-V (partial). B, right p4-m3, MNHN- BOL-V C, right p3-m3, MNHN-BOL-V Lingual to left and mesial at the top. Scale bar = 1 cm. foramen is immediately postero-medial to the posterior lacerate foramen. It opens in a slightly depressed area between the occipital condyles and the posterior lacerate foramen. It is separated from the posterior lacerate foramen by a very thin crest of bone. The tympanic bullae are large, distinctly bilobate and approximately as wide as long. In ventral view, the outline of the bulla presents a strong constriction, near the middle of its anterior oblique face. This constriction is in fact an anteroposteriorly orientated wide groove, which starts anteriorly ventral to the sphenotympanic fissure and ends posteriorly just anterior to the paroccipital process. According to Patterson (ms), this constriction delimits the medial hypotympanic sinus from the dorso-lateral and smaller cavum tympanii. The bullae do not bear a styliform process at the antero-medial corner. This corner markedly narrows anteriorly to a small bony process (SAL 551) (Fig. 9). The groove for the auditory (Eustachian) tube is in its usual position running anteriorly from the antero-medial corner of the bulla. The canal of Huguier opens latero-dorsal to the cavum tympanii of the bulla and medio-ventral to the glenoid cavity. It is a minute foramen bounded anteriorly by a small bulge (crest) formed by the squamosal and postero-ventrally by a small triangular process, also formed by the squamosal, and that is applied against the bulla (Fig. 8C). Therefore, the canal of Huguier is bounded laterally by the squamosal and medially by the ectotympanic. It probably corresponds to the end of the Glaserian fissure, as is usually found in notoungulates (Patterson, 1936). The glenoid cavity is positioned high on the skull, well above the tooth row, approximately at the level of the lacrimal foramen. It is not entirely preserved on any of the specimens, but may be composed as in Trachytherus alloxus of a mandibular fossa and articular facets (Billet et al., 2008). SAL 4 retains only the postero-medial part of the cavity. Medially in the mandibular fossa, at the antero-dorsal base of the postglenoid process, is a small foramen of unknown nature (not visible on the figures). It is very small on the right side whereas it is much larger on the left side of SAL 4. Posteriorly, the postglenoid process abuts against the dorsal part of the crista meatus. These two structures define a thick vertical wall posterior to the mandibular fossa. The squamosal tympanic suture is distinct on this wall on SAL 4. It passes through the conspicuous and rounded postglenoid foramen, which opens in the middle of the vertical wall. Medially, the tympanic squamosal suture is less distinct. At the postero-lateral corner of the bulla, the crista meatus gradually arises medially from the lateral wall of the cavum tympanii. It is postero-laterally orientated and ends in a strong process antero-ventromedially to the external auditory meatus. At this point, the crista meatus splits dorsally into two processes that bound the external auditory meatus anteriorly and posteromedially (Fig. 8B). Because the crista meatus is widely separated from the weakly developed post-tympanic process of the squamosal, the two structures are separated by a wide channel rather than by a cleft. This channel houses the stylomastoid foramen, the tympanohyal recess and a postero-laterally orientated sulcus. The stylomastoid foramen opens very close to the tympanohyal recess. More precisely, it pierces the posterodorsal part of the lateral wall of the tympanohyal recess. Patterson (1932) described the stylomastoid foramen as situated midway from the tympanohyal recess to the external auditory meatus in many notoungulates, or slightly closer to the tympanohyal recess (Patterson, 1934a) but never as close as is observed in Archaeohyrax suniensis. Posterolateral to the stylomastoid foramen and lateral to the posttympanic process is a deep and wide groove probably for the passage of the facial nerve. As is usual in notoungulates (Patterson, 1932), the deep tympanohyal recess (vagina processus hyoidei)

14 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 471 Figure 17. Archaeohyrax suniensis sp. nov., deciduous lower teeth. A, left dp3-4, m1-2, YPM-PU B, right dp3-4, SAL 646. Labial to left for A, to right for B, mesial at the top for both. Scale bar = 1 cm. is situated at the postero-lateral corner of the bulla, anterolateral to the paroccipital process and anteromedial to the stylomastoid foramen. A small medially concave tympanic crest (clearly distinct from the crista meatus) forms the lateral edge of the tympanohyal recess (Figs 7, 8A, C). This crest is probably homologous (see cladistic analysis) to the structure that bounds the tympanohyal recess anteriorly in hegetotheriines, in which the crista meatus is vestigial. The lateral wall of the bulla bounds the tympanohyal recess antero-medially. Postero-medially, the tympanohyal recess is bounded by a thin blade of bone, running from the tympanic bulla, and which contacts the tip of the post-tympanic process posteriorly. The paroccipital process, even if very close from it, does not participate in the posterior wall of the tympanohyal recess, unlike the condition in Toxodontia (Patterson, 1932; Gabbert, 2004). This exclusion of the paroccipital process from the posterior edge of the tympanohyal recess is a widespread feature among typotherians (Patterson, 1934a). The paroccipital process is massive and transversally compressed. Its medial face is slightly concave. Its lateral face bears a tiny ridge, which appears to be a ventral extension of the post-tympanic process of the squamosal. The only entirely preserved paroccipital process is present on the right side of SAL 4. It extends much more ventrally than the bulla. The anterior edge of the paroccipital process almost contacts the posterior wall of the bulla, a condition which clearly separates the tympanohyal recess laterally from the posterior lacerate foramen medially. The external auditory meatus opens high on the skull and faces postero-laterally. As mentioned above, the ventral, anterior and posterior border of the meatus is formed by extensions of the crista meatus. The squamosal may participate dorsally. Anterior to the external auditory meatus in lateral view is a large rounded aperture (approximately one-third of the external auditory meatus in size). This aperture is bounded anteriorly by the posterior face of the postglenoid process (Fig. 8B) and posteriorly by the anterior border of the external auditory meatus. The suprameatal foramen opens at the bottom of this aperture. This differs from Trachytherus alloxus (Billet et al., 2008) and many toxodontians (Gabbert, 2004) in which it opens directly at the surface of the skull. OCCIPUT All occipital elements are fused. The foramen magnum, wider than high, presents an oval shape. It faces equally posteriorly and ventrally. The outline of

15 472 G. BILLET ET AL. the exoccipitals in posterior view resembles a sandglass but different from that in mesotheriids (Billet et al., 2008). The bottleneck corresponds to a slight dorsal narrowing between the epitympanic sinuses. The widest part of the occipitals that extends laterally to the foramen magnum forms the paroccipital process ventrally. The mastoid foramen is not visible on the available specimens. The epitympanic sinuses are clearly visible on SAL 551 because the part of the squamosal bone that normally covers them has been broken off. They have the usual position of notoungulates, i.e. in the posterior and dorsal part of the squamosals (Gabbert, 2004). Their major length axis is orientated subvertically; more precisely, the dorsal extremity of this axis is orientated slightly postero-medially. The medial occipital region is slightly bulged compared with its lateral counterparts. DENTARY (FIG. 4) As in many typotherians, the symphysis is long and hollow dorsally. It is also markedly procumbent. The corpus is massive. It increases evenly in height from incisors to molars (Fig. 4A). The angular region is rounded and descends slightly below the level of the horizontal ramus (MNHN-BOL-V ). There are two constant mental foramina. The anterior one is situated below the emplacement of i3-c and is deeply opened on the ventral surface of the symphysis. The position of the posterior one varies with wear: it is situated below the distal part of p2 when this tooth erupts (YPM-PU 23800, MNHN-BOL-V ) in early stages of wear; and it is situated below the p4-m1 transition in late wear stages (SAL 305, SAL 306, YPM-PU 23702, YPM-PU 23765, YPM-PU 22068, MNHN-BOL-V ). The opening of the mandibular canal is large and a well-defined groove leads into it. It is situated in the inferior part of the corpus, well posterior to the level of the distal extremity of m3. DENTITION The great number of known specimens allows us to study the different wear stages, which reveals valuable information on the development of the cheek teeth. In the following description the terms hypsodont and hypselodont are used in the same sense as that in the description of Trachytherus alloxus (Billet et al., 2008) UPPER INCISORS AND CANINE (FIGS 6, 7) The I1 is greatly enlarged compared with I2-3. These teeth are hypsodont although they are likely to be rooted in old individuals as their crowns distinctly taper on SAL 313. The first incisors are obliquely set in the premaxilla as in other archaeohyracids, hegetotheriids and mesotheriids (Cifelli, 1993). They bear enamel only on their labial face. The I2 and I3 are reduced, I2 being slightly larger than I3. Both are separated from preceding and following teeth by small diastema (Figs 6, 7). The I3 is erupting on SAL 4. It is originally incisiform with enamel cover only labially. This tooth is peg-like and transversely compressed. It presents a crest that runs mesio-distally from anterior to posterior base passing through the apex. This crest runs along the lingual edge of the enamel. The canine is a very small tooth, slightly larger than I3. It is also transversely compressed and has no enamel on its lingual side. UPPER PREMOLARS (FIGS 3, 10, 11) All specimens of P1 are heavily worn. This tooth is longer (mesio-distally) than wide (labio-lingually), being narrower mesially than distally with a hint of parastyle. The other premolars are markedly hypsodont and only form roots in old individuals (SAL 313). They present a well-individualized parastyle with a deep groove (but shallower than in interatherids) in early stages of wear. This sulcus disappears in late stages of wear, when the occlusal surface has become featureless. P2 is originally less elongate mesio-distally than the following premolars. This gives it a more squared shape in early stages of wear, whereas P3-4 are more rectangular (SAL 4). With increasing wear, all tend to become triangular with an oblique mesio-lingual face. Examples of unworn premolars of A. suniensis show that the first crista joins the protoloph, and the second crista joins the crochet of the metaloph, emphasizing the characteristic typotherian face pattern of fossettes (Cifelli, 1993) (Fig. 10). However, there is an important difference from the original pattern of the molars described below, i.e. the formation of a third labial fossette between the usual anterior labial fossette (formed by the union of crista 1 and the protoloph) and the posterior labial fossette (formed by the union of crista 2 and the crochetmetaloph); this third fossette is isolated lingually by the contact between the crochet-crista 2 and the protoloph. In fact, it separates crista 1 from crista 2. The anterior labial fossette may be the more resistant labial fossette on the premolars. Thus, unworn to slightly worn P2-4 present an anterior labial fossette, an intermediate labial fossette, a posterior labial fossette, a central valley with a lingual aperture and a postcingulum fossette. The posterior labial fossette is trifling and does not resist long into wear (YPM-PU 23188, YPM-PU 23789).

16 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 473 In later wear stages, the central valley is isolated in a central fossette and the anterior labial fossette reduces (YPM-PU 23743, SAL 4; Fig. 3). With increasing wear, the anterior labial fossette disappears (Fig. 11A). In late wear stages, the central fossette also disappears, which leads to a featureless occlusal surface (Fig. 11B). UPPER MOLARS (FIGS 4, 5, 11, 12) The molars of A. suniensis are highly hypsodont (Fig. 4A) and no roots appear on any of the specimens observed so far. The molars are more elongate mesiodistally than the premolars and they present a tiny parastyle/paracone groove that disappears after little wear. They are rectangular in early wear stages, and trapezoidal in later stages (with the labial face being the largest one). Throughout the wear process, M1 and M2 narrow mesio-distally and lengthen labiolingually (the trend is much less clear for M2 than for M1; see Discussion of measurements) whereas M3 broadens mesio-distally and labio-lingually. Unworn examples of M1 and M2 are present in the SAL and YPM-PU collection. On these teeth, crista 1 attaches to the protoloph, and crista 2 joins the crochet-metaloph (Fig. 12C). A deep central valley separates the protoloph-crista 1 from crista 2-crochetmetaloph. Additionally, a small sulcus marks the lingual attachment of the crochet to the metaloph. In later wear stages, this sulcus forms the small posterior arm of the V-shaped central fossette, the anterior arm being formed by the above-mentioned deep central valley. On unworn M1 and M2, a posterior ridge of cingulum is still present, thus isolating a postcingulum fossette. The anterior labial fossette wears off rapidly, followed by the postcingulum fossette and finally the posterior labial fossette (Fig. 12B). When the posterior labial fossette disappears, the central fossette is isolated. It produces a tooth with a single central fossette (Fig. 11A). Ultimately, this fossette disappears and leaves a featureless occlusal surface (Fig. 11B, C). The third molar is rather different and does not present a postcingulum fossette. The two labial accessory fossettes mentioned above are present (Fig. 5A). They are contemporaneous, the anterior one being more resistant than on M1 and M2. They disappear just before isolation of the central fossette. This isolation occurs in the distal part of the lingual face with the great development of the protoloph (Fig. 11B), unlike on M1 and M2 where it occurs in the middle of the lingual face. Finally, when the central fossette disappears, a small distal extension, corresponding to the metastyle, develops (Fig. 11C). The wear of the upper cheek teeth and especially that of M1 and M2 produces a particular relief on the labial part of the tooth. Two peaks are individualized labially which are grossly positioned in the place of the paracone and metacone, although these two cusps are not recognizable on these teeth. Therefore, these two peaks should not be termed paracone and metacone as previously done. The upper cheek teeth lack enamel on a vertical strip in the labial part of their mesial and distal face. UPPER DECIDUOUS TEETH (FIGS 3, 12) The di1-2 and dc are heavily worn on SAL 4 (and also on SAL 1010). They all approach root level at this stage. The di1 is much larger than the di2, as for their permanent counterpart. They do not differ from these latter except for their weaker hypsodonty level. The di2 (di3 broken on SAL 1010, and already replaced by I3 in SAL 4) and dc greatly resemble each other and are incisiform (Fig. 3). The deciduous premolars are hypsodont and rooted. The first deciduous premolar, dp2, bears a parastyle that greatly extends mesially (Fig. 12C). There is also a pronounced parastyle/paracone groove, emphasizing the individualization of the parastyle. The dp2 is trapezoidal to sub-triangular with increasing wear. The less worn specimen of dp2 (SAL 583) presents a disappearing posterior labial fossette and an isolated central fossette. Slightly worn specimens of dp3 and dp4 present the usual two accessory labial fossettes and a lingual aperture (Fig. 12A). A postcingulum fossette is also distinct. Even if they greatly resemble the molars in their general occlusal pattern, the deciduous premolars are different in at least one aspect: when the central fossette is enclosing, there are still labial accessory fossettes present (Fig. 12A, C). However, they do not resist long to additional wear. The dp3 and dp4 are trapezoidal. They possess a parastyle/ paracone groove but much less marked than on dp2. At late stages of wear, all the deciduous premolars show only a single central fossette that almost disappears when wear approaches root level (Fig. 12B). LOWER INCISORS AND CANINE (FIG. 13) The first and second pair of incisors are procumbent and have approximately the same size. They are cylindrical and lack enamel on small vertical mesial and distal stripes. The i1 and i2 are not much more developed than i3. The i3 is also cylindrical and procumbent. It is applied against i2 (Fig. 13). The canine is more reduced than the i3. It is incisiform when erupting (MNHN-BOL-V ) and is also procumbent, but less than i1 and i2. Very small diastemata separate the canine from i3 and p1.

17 474 G. BILLET ET AL. LOWER PREMOLARS (FIGS 13 15) Little can be said concerning p1. It is always poorly preserved and/or heavily worn. This tooth is compressed labio-lingually and elongated mesio-distally. The following premolars all exhibit a labially bilobed pattern. The trigonid of p2 is not fully developed as on the other premolars and molars because it lacks a transversally extended mesial crista (paralophid) (Fig. 13). Likewise, the transverse metalophid is reduced on p2. Thus, the trigonid of p2 originally comprises principally of a mesio-distally extended protolophid. It is likely that this tooth cannot isolate any fossette except on its talonid. The trigonid of p3 and p4 is composed of a transverse and low paralophid, a mesio-distally orientated and labial protolophid, and a transverse metalophid which extends slightly distally. Furthermore, from the lingual extremity of the paralophid runs distally a small crista which nearly closes the thus defined central fossa in approaching the metalophid. All these cristae and lophids form a continuous structure that defines a rounded trigonid with a disto-lingual pointed extension (Fig. 14A). The talonid of p2-4 is also rounded. It consists of a labially convex hypolophid and a shorter lingual entolophid. These two lophids are connected mesially, at the point where the talonid contacts the trigonid. At the disto-lingual corner of the talonid, a small vertical groove delimits the entolophid from the hypolophid in early stages of wear. A central fossa is also present between these lophids in the centre of the talonid, but it does not resist long into wear. The point at which the talonid first contacts the trigonid, in early stages of wear, is labial relative to the second contact, which occurs when additional wear makes the disto-lingual extremity of the trigonid (metalophid) contact the mesio-lingual part of the talonid. As a consequence, this leads to the isolation of a trigonid talonid fossetid and to a flat lingual face (Fig. 14B). Finally, after additional wear, this fossettid disappears. Thus, p3 and p4 present a labially bilobed outline, with a flat to slightly hollow lingual face and a featureless occlusal surface. At this stage, the trigonid is longer than the talonid, but the talonid is wider, extending well beyond the trigonid labially on p3 and p4 (although this trend varies in intensity compare Figs 14C, 15). Simultaneously, p2 loses the bilobation on its labial face and becomes very simple and reduced (Fig. 15). The p3 and p4 reach this morphological state on latest wear stages (SAL 306). LOWER MOLARS (FIGS 13 16) The molars are roughly similar to the posterior premolars but are mesio-distally larger. The original pattern is approximately the same except that the hypolophid is more extended distally. As a consequence, the disto-lingual groove separating the entolophid from the hypolophid is larger and more persistent. The m1 and m2 present the same morphological changes during wear as p3-4. First, they present a lingual aperture between the trigonid and the talonid. Secondly, a trigonid talonid fossetid is isolated which finally disappears and leads to a featureless occlusal surface with a bilobed labial face (Figs 14A C, 15). The m1-2 tend to acquire an occlusal outline, which becomes more similar to that of p3-4 with increasing wear. When the trigonid talonid fossetid has been isolated, a groove of varying intensity (Fig. 16; see also Discussion) and a slight angulation of the lingual face still emphasizes the boundary between the trigonid and the talonid lingually. This produces a hollow lingual face in late wear stages. Slightly worn molars present a trigonid with a partially or entirely lingually isolated fossa, thus isolating a tiny trigonid fossetid. The entolophid and hypolophids connect at the point where the talonid first contacts the trigonid. The entolophid presents a mesio-distal orientation and is mesially extended (Fig. 14A) compared with its configuration in basal typotherians such as Oldfieldthomasia or Notopithecus. This is typical of archaeohyracids, mesotheriids and hegetotheriids (see cladistic analysis). The third molar is more elongated, owing to the greater distal extent of the hypolophid on the talonid. It also differs from the preceding teeth in being labially trilobed. This is due to the presence of a labial vertical groove on the middle of the labial face of the talonid. This labial trilobed appearance is persistent throughout all wear stages. Furthermore, on m3, the isolation of a small entolophid hypolophid fossetid follows the occurrence of the disto-lingual groove between the entolophid and the hypolophid (Fig. 14C). It does not resist long into wear. No roots have been observed on any specimen, but this part of the crown has not been available for observation on the considered specimens. LOWER DECIDUOUS TEETH (FIG. 17) SAL 1010 preserves deciduous i1-2. These teeth are procumbent, incisiform, and only differ from their permanent counterpart in being more slender and in having a lower crown. Unworn specimens of dp2-4 are known. They present a pattern of lophids which is similar to that of permanent premolars and molars (Fig. 17). They are hypsodont but rooted. Their wear stage succession resembles that of permanent premolars and molars. The trigonid and talonid of the deciduous premolars

18 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 475 both present a fossetid in their centre. These fossettids are likely to be homologous to the fossae described on the permanent premolars and molars. The more resistant of the two fossettids is the trigonid one. It only disappears when a third fossettid, the trigonid talonid one, is being isolated. This latter fossetid is close to disappearing as wear approaches the roots but this disappearance has not been observed. The deciduous lower premolars retain a labially bilobate pattern throughout all wear stages. The shape of the superior tooth rows varies with wear. It changes from parabolic (SAL 183) to more linear (SAL 313). No trend is clear for the inferior tooth rows, although it appears to be more linear than parabolic. Both upper and lower cheek teeth are inconsistently covered with a thin layer of cement, especially on old individuals. ONTOGENY AND TOOTH WEAR We recognized 11 and 12 ontogenetic stages for the upper and lower jaws, respectively (Appendix S2). Only the holotype presents associated upper and lower jaws in this study. It shows the correspondence of upper jaw stage 5 with lower jaw stage 5 (see Appendix S2). ARCHAEOHYRAX PATAGONICUS AMEGHINO, 1897 FIGURES 18 22, APPENDIX S4 (TABLES 3, 4B) Archaeohyrax patagonicus Ameghino, 1897, pp , figs Argyrohyrax proavus Ameghino, 1897, pp (partim); (information from Patterson ms; Reguero, 1999). Archaeohyrax propheticus Chaffee, 1952, p Holotype: MACN A52-617: cranium and associated mandible (Figs 18 21). Referred specimens: MACN A52-624: left mandible with p3-m3 (Fig. 22); MACN A52-620: five isolated upper cheek teeth from various individuals; MACN A52-627: right isolated upper first molar (?); MACN A52-623: left mandible with a deciduous premolar (dp3 or dp4); Patterson (ms) also mentioned MACN A52-619: left M1 and right m2; FMNH P13500: left M1; FMNH P14681: portion of right horizontal ramus with m2-3. Reguero (1999) also mentioned MLP 83-I-12-4: right maxillary with P4 (broken) and M1-3; MLP 93-XI-21-39: right and left maxillary fragments both with dp4-m1; AMNH 29610: palate with right di1-3, Figure 18. Archaeohyrax patagonicus, cranium of the holotype, MACN A A, ventral view. B, dorsal view. C, right lateral view. Scale-bar = 1 cm. and right and left dc, P1 and dp2-4; AMNH 29611: right maxillary with dc, P1, dp2-4 and M1 (erupting). Marshall et al. (1986) referred AMNH (unassociated M1 or M2, m2 and m2-3) to this species. As AMNH and come from the same fauna (Scarritt Pocket) as the AMNH specimen, it is likely that, even if they are hardly comparable with other specimens because they are much younger (deciduous dentition), these specimens belong to this species, as suggested by Reguero (1999; see above). Localities: All specimens come from Patagonia, Argentina. MACN A and MACN A have accompanying labels reading Río Chico frente a Malaspina and Río Chico f. Malaspina, respectively.

19 476 G. BILLET ET AL. Figure 19. Schematic representation of the cranium of Archaeohyrax patagonicus, ventral view, from MACN A Scale bar = 1 cm. Malaspina is inland from Bustamante Bay and on the south-east side of the valley of the Río Chico del Chubut. However, we do not have an exact explanation for frente a Malaspina. Patterson (ms) suggested, based on the association of the term Yacimiento grande with both Cabeza Blanca and Río Chico frente a Malaspina, that this latter site may be synonymous with the locality of Cabeza Blanca. The holotype and other MACN specimens of the Ameghino collection have no mention of a locality. Their colour, preservation and such matrix as was still adhering to them suggest that all, and especially the type, could come from Cabeza Blanca. The same is true of the specimens labelled frente a Malaspina. The FMNH specimens also come from Cabeza Blanca (Patterson, ms). The AMNH specimens come from Scaritt Pocket, Chubut. MLP 93-XI comes from Las Cascadas and MLP 83-I-12-4 comes from Rocas Bayas, Maquinchao, Río Negro (Reguero, 1999). Distribution: Deseadan SALMA, late Oligocene. Diagnosis: Differs from A. suniensis in: larger size; a larger facial extent of the premaxillaries, exceeding the level of the canine posteriorly; a greater number of teeth exhibiting simultaneously a fossette or fossetid (i.e. P3-M2 & p3-m2 in A. suniensis, P2-M3 & p2-m3 in A. patagonicus); less hypsodont cheek teeth; second mental foramen situated less anterior than in A. suniensis; stylomastoid foramen less aproximated to the tympanohyal recess. SKULL (FIGS 18 20) Note on the skull of Archaeohyrax patagonicus (Patterson, ms) The holotype of A. patagonicus was extensively illustrated by Ameghino, who devoted no less than five figures to it in his work of His summary description is accurate, except for the statement that p1 is lacking, but the figures unfortunately are not. They are correct with regard to the general shape, although somewhat idealized the arches, for example, are shown as complete, which they almost certainly were not but are incorrect and misleading in much important morphological detail. When one of us (B.P.) first saw it, the skull was partially covered by matrix and mastic. Partial removal of this has permitted a better observation of this specimen. But it is again covered with mastic in many parts. Some

20 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 477 Figure 20. Schematic representation of bone contacts in the orbitotemporal fossa in the typotherians Protypotherium (Interatheriidae), Mesotherium (Mesotheriidae), Hegetotherium (Hegetotheriidae) and Archaeohyrax (Archaeohyracidae). Modified after Patterson (ms). observations, reported here and originally made by B.P., therefore could not be verified precisely. The skull of this species is known from only one specimen, which is the holotype (Fig. 18). In many aspects, the skull of A. patagonicus greatly resembles that of A. suniensis. Thus, it is often referred to the description of the latter when a particular structure is similar in A. patagonicus. Naso-facial region This region is greatly similar to its homologue in A. suniensis. It only differs in small detail, such as the greater facial extent of the premaxillaries. The premaxillary maxillary suture on the rostrum meets the nasals dorsally above the level of C in A. patagonicus. Its most posterior point is above P1. In A. suniensis, the premaxillary maxillary suture meets the nasals above the level of I3. The limits of the lacrymal are not very clear in A. patagonicus, but, even if possibly smaller than in A. suniensis, the facial extent of this bone was important. The maxillo-frontal contact is narrow. Posteriorly, the maxillary reaches the level of the posterior border of the nasals. The medial bulge of the nasals is as strongly marked as in A. suniensis. The relationships between the frontal and nasal bones are the same as in A. suniensis. One of us (B.P.) was engaged in clearing away matrix and mastic from the maxillary of A. patagonicus when it became apparent that the zygomatic root was hollow. Further preparation revealed a very extensive maxillary sinus within the root. This sinus extends medially above and behind the base of M3, and posteriorly for a short distance within the arch (this cavity has been subsequently filled with plaster in order to safeguard the specimen). The rest of the zygomatic arch is similar to that of A. suniensis. Palate The premaxillary maxillary suture on the palate runs slightly more backward than in A. suniensis, in the part immediately medial to I3. The rest of this suture course is the same as in the Bolivian species. As in A. suniensis, the incisive foramina are minute (Fig. 19). The minor palatine foramina and maxillary palatine suture are hardly detectable. It appears that the maxillo-palatine suture, which is hardly distinct, has a different course to that of A. suniensis. The palatal portion of the palatines has a semicircular shape and does not extend anteriorly beyond the level

21 478 G. BILLET ET AL. Figure 21. Archaeohyrax patagonicus, dentition. A, five isolated upper cheek teeth, MACN A Mandible with right i1-i3, p2-m3 and left i2-c, p3-m3: B, occlusal view; C, left lateral view. Scale bar = 1 cm. of the mesial part of M1. The post-palatal platform presents the same morphology as in A. suniensis. Orbitotemporal region and skull roof The antero-medial part of the orbital wall resembles that of A. suniensis, except that no structure resembling an ethmoid foramen has been detected. In the orbitotemporal fossa, the bone contacts more resemble those of Hegetotherium than those of Mesotherium or Protypotherium (Fig. 20). Within the temporal fossa, the alisphenoid extends up as a broad wedge, widening dorsally, with a long contact with the parietal and a shorter one with the frontal. Within this portion of the bone, there is a conspicuous, asymmetrical V-shaped ridge (the anterior arm of the V being longer and more vertical than the posterior arm), which bounds a depressed triangular area. As in Hegetotherium, this part of the alisphenoid is separated from the portion forming the border of the foramen ovale by a triangular, descending wedge of the squamosal. The orbitosphenoid runs upward and forms a long rectangle ending dorsally slightly below the base of the postorbital process. Beneath and behind the postorbital process, a wedge of the frontal

22 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 479 the supra-orbital canal is preserved on both sides. Curiously, only the right side on the frontal plate is pierced by the superior opening of the supra-orbital canal (this is presumably an individual variation). The zygomatic arches are not preserved entirely. The sagittal crest is conspicuous and is likely to be entirely formed by the parietals. Figure 22. Archaeohyrax patagonicus, schematic representation of left p3-m3, MACN A Labial to right and mesial at the top. Scale bar = 1 cm. extends back beyond the orbitosphenoid to form an oblique suture with the alisphenoid. Anterior to the orbitosphenoid, the descending plate of the frontal forms by far the greater part of the wall of the orbit. The optic foramen, opening in the centre of the orbitosphenoid, is well separated from a foramen situated postero-ventrally to it. The latter may be the sphenorbital fissure, as observed on A. suniensis. As in A. suniensis, the squamosal does not extend as forward as in Hegetotherium and does not reach the postorbital process. The parietals form roughly half of the broad dorsal surface of the posterior cranial roof. Rather narrow posteriorly, they progressively widen toward the alisphenoid contact and then narrow quite abruptly. Anteriorly they diverge into two blunt projections that lap the bases of the postorbital processes. Along the posterior part of the squamosal parietal suture are various venous openings as is usual in notoungulates. Beneath the orbital rim, slightly anterior to the post-orbital processes, the large inferior opening of Basicranium and auditory region Most of this region is roughly preserved on the unique known skull of A. patagonicus, and hence precise anatomical details cannot be determined easily. The holotype bears no suggestion of a hamular process of the pterygoid. This is similar to the condition in Hegetotherium, in which Sinclair (1909: 73) believed the pterygoid to have been lost from his material, and in Pachyrukhos. The same also occurs in Trachytherus alloxus from Salla (Billet et al., 2008). The pterygoid is usually closely appressed to the palatine and the alisphenoid, forming the entopterygoid crest and the hamular process on the mesial side of the palatine crest (Billet et al., 2008). The pterygoid may have been very poorly attached to the cranium in these taxa and have rapidly detached after the animal s death. Therefore, nothing can be concluded on the presence of a hamular process. From the base of the buttresses of the postpalatal platform (observation by B.P. before this region of the skull was filled with mastic), the alisphenoid extends posteriorly as a progressively thinning strip that contacts the tympanic portion of the bulla, and forms all or most of the margin of the long, rather slit-like sphenotympanic fissure. As in A. suniensis, the bulla presents a strong constriction on its anterolateral face (Fig. 19). B.P. had been able to examine the inside of the bulla of A. patagonicus. He noted that the constriction corresponds to a vertical septum within the bulla, as in Hegetotherium (Patterson, 1936). However, in hegetotheriids, the constriction of the bulla is less pronounced than in Archaeohyrax. Apparently, the hollow interior of the hypotympanic sinus does not exactly correspond to the medial part of the bulla delimited by the constriction. The hypotympanic sinus is slightly more oval and less tapering. It narrows anteriorly and terminates in a small bony process closely appressed to the basi- and alisphenoid. The walls of this sinus are not cancellous. Unlike previously believed by B.P., the presence of a septum is not evidence for the existence of a compound (ecto- and ento-tympanic) bulla in notoungulates. Gabbert (2004) hypothesized that the septum of toxodontians was formed only by the ectotympanic bone. The same might also be true for archaeohyracids and hegetotheriids.

23 480 G. BILLET ET AL. The stylomastoid foramen opens midway between the external auditory meatus and the tympanohyal recess, less close to the latter than in A. suniensis. The tympanohyal recess is bounded by the same elements as in A. suniensis. The crista meatus is massive, especially above the external auditory meatus. Medially, between the crista meatus and the dorsal part of the lateral wall of the cavum tympanii, B.P. observed a conspicuous opening which probably corresponds to the external aperture of the canal of Huguier. The postglenoid foramen is large and the postglenoid process forms a conspicuous anterior bulge around it. It opens in the same location as in A. suniensis. In lateral view, the region anterior to the external auditory meatus is very different in the two Archaeohyrax species. In A. patagonicus, a depressed area is present antero-dorsally to the external auditory meatus and dorsally to the crista meatus pillar. It is bounded ventrally by a slight crest of bone. In A. suniensis, instead of a depression there is an important opening, separated posteriorly from the external auditory meatus by a thin strip of bone apparently continuous with the crista meatus. A small foramen is present in this opening in A. suniensis. It may represent a suprameatal foramen as is observed in Trachytherus alloxus (Billet et al., 2008) and many toxodontians (Gabbert, 2004). However, this difference between the two Archaeohyrax species may be only due to preservation (this feature is observable on only one specimen for each species), as the wall hiding this opening and defining the depressed area might have been broken away in the holotype of A. suniensis. The paroccipital process, posterior lacerate foramen and hypoglossal foramen are similar to their condition in A. suniensis except for the larger space separating the hypoglossal foramen from the posterior lacerate foramen latero-anteriorly. Occiput All occipital elements are fused. The nearly circular foramen magnum forms an angle of almost 45 with the long axis of the skull. The posterior exposure of the occipital has a rough sandglass shape as in A. suniensis. The mastoid foramen is a large opening on the occipital surface, between the occipital and squamosal bones in the part corresponding to the sandglass bottleneck. LOWER JAW (FIG. 21) The deep and long ascending ramus of the mandible is preserved. The masseteric fossa is large and well defined. The coronoid process and the angular area are lacking (Fig. 21B, C). The anterior mental foramen is situated below the canine (MACN A ), and the posterior one is situated below m1-2 (MACN A ) or below p4 trigonid on a younger individual (MACN A ). No other differences with A. suniensis have been observed. DENTITION Fewer specimens are known for A. patagonicus than for A. suniensis so the description of the teeth of the former and their variation throughout the wear process is not as complete as for the latter. Here we present the original description (with some corrections) of B.P. Some details originally observed by B.P., especially on the anterior lower teeth, are no longer visible. The dentition greatly resembles that of A. suniensis in most details, and mention of the few differences existing is added to the description below. UPPER TEETH (FIGS 18, 19, 21A) The first incisor, in comparison with I2-C, is a markedly enlarged cropping tooth, the length being nearly twice the width (see measurements in Appendix S4, Table 3). The labial face is gently convex, enamel covered and with a poorly defined vertical groove that does not extend over the full height of the enamel covering. How far this covering extends is not known as the total height of the tooth cannot be seen (enamel is visible for 11.5 mm). Mesial and lingual faces are essentially straight, the labial bluntly rounded. There is a small enamel remnant at the mesial extremity of the lingual face. As in hegetotheriids and mesotheriids (Cifelli, 1993), the I1 are implanted obliquely in the sides of the premaxillaries and converge anteriorly, meeting only at their tips. I2-3 are small teeth with oval crowns, oblique to the mesio-distal axis of the tooth row and with enamel only on their labial face. Despite their small size, both are moderately high-crowned, the enamel extending 7 mm above the worn crown in I3 and further in I2 (base not visible). The crown of all incisors and the canine are worn obliquely inward. The canine is slightly larger and longer than I2-3, transversely compressed and with enamel on all but the disto-lingual face. I2-C are separated from each other and from I1 and P1 by a short diastema. The P1 of the type is very low-crowned and has a featureless grinding surface. The ectoloph is gently rounded distally, but anterior to the theoretical position of the paracone, it slopes abruptly inward. This may be homologous to the parastyle paracone groove of the other premolars. The tooth narrows from its widest part distally to an acute angle mesially. The distal face is rather angular. Enamel is lacking on

24 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 481 this face. The anterior root is a stout buttress extending well forward of the crown. The remaining premolars are molariform and very high-crowned. In the stage of wear exhibited by the holotype, the premolar series shows an increasing labio-lingual width and a decreasing mesio-distal length from P2 to P4. Each tooth is subtriangular, with a short, rounded lingual face, a nearly transverse distal one and a strongly inclined mesial one. The ectoloph is slightly convex with a moderate parastyle paracone groove. The only crown feature remaining on the holotype is a circular to oval medial fossette; this may be quite persistent, disappearing after wear approaches the roots, as seen on some isolated premolars (MACN A52-620). Enamel is lacking over the labial portion of both mesial and distal faces. The first and second molars are similar to the posterior premolars, but differ in greater length and trapezoidal shape, a much less prominent parastyle paracone groove and more elongated central fossettes. M3 is the longer tooth. On the latter, the ectoloph extends more posteriorly as a metastyle extension and the posterior face is concave. Unworn M1 (AMNH 29611) is identical to that described for A. suniensis. Another slightly worn M1 (MACN A52-619) has been lost (A. Kramarz, pers. comm.) since B.P. s (ms) original study. He described it as follows: the remnant of a cleft between protocone and hypocone remains, and the medial (central) fossette is not yet isolated; a vestige of the (postero-labial) fossette is still present, but no trace of the antero-labial. Variability in the depth of these accessory fossettes is revealed by a second isolated M1 from a slightly older individual (FMNH P13500) judging from the isolation of the central fossette and the loss by wear of the cleft. In this tooth, there is a remnant of the (antero-labial) fossette but no trace of the postero-labial. It is odd that there is such variability in the accessory fossette depth as it is remarkably stable in A. suniensis in a much larger number of specimens. Unfortunately, the two specimens mentioned have not been seen by us (except B.P.) and we cannot verify these observations, and particularly the nature of the teeth (one might be deciduous for example). MACN A and MACN A (Fig. 21A) present rooted upper cheek teeth whose exact position in the tooth row is barely determinable (posterior premolars and molars). Most are featureless or with disappearing central fossette. On the holotype, the base of M3 is visible, and it has not begun to form roots. According to the rooting of the above-mentioned specimens and the apparent absence of rooted posterior premolars and molars in A. suniensis (only tapering crown for posterior premolars on SAL 313), it appears that there is a clear difference of hypsodonty level between the two species. The Patagonian species is less hypsodont than the Bolivian one. The holotype and MLP 83-I-12-4 exhibit the same wear stage. They respectively present central fossettes from P2 to M3 and from P4 to M3 (more anterior teeth are not preserved on this specimen). The co-occurrence of a central fossette on (P2-3-)- P4-M3 never occurs in A. suniensis. In this latter species, the maximum of co-occurrence of the central fossettes is P3-M2 at stage 7 only (see details of ontogenetic stages in Appendix S2). Some deciduous upper teeth are present in the AMNH collection. The stage exhibited by AMNH and AMNH appears to be slightly younger than all that known in A. suniensis. The central fossette is not isolated on any teeth, and they present labial accessory fossettes and a postcingulum fossette. LOWER TEETH (FIGS 21B, C, 22) The lower incisors and the canine of the type are small, and nearly equal in size. They are separated from each other and from p1 by very short diastemata. Enamel is present on the labial and lingual faces of i1-2 but may have been lacking (at least in the wear stage of the teeth) on the mesial and distal faces. On i3 and on the canine, the enamel appears to be continuous. The i1-2 are cylindrical; the i3 and the canine are more oval and with a suggestion of a posterior heel, perhaps an artefact of wear. Ameghino supposed that p1 was lacking, but the alveolus for it is present on the left side. This area was covered by matrix and mastic when viewed by Ameghino. The right side, on which he obviously relied, is crushed inward, leaving scarcely a trace of the alveolus. Today, the specimen still bears mastic in this area. The second premolar is known only in the type. It is rather deeply worn, with a nearly straight lingual face, the labial groove between trigonid and talonid is shallow but persistent, and the trigonid is more tapering anteriorly than in the succeeding teeth. A slight remnant of the fossettid between trigonid and talonid remains. MACN A (Fig. 22) has p3-m2 slightly worn and m3 quasi unworn. This specimen was mistakenly identified by Ameghino (Ameghino, 1897; Patterson, ms) as Argyrohyrax. At this stage, p3-m2 are similar in structure, and increase progressively in length. They lack enamel over their mesial face. Trigonid and talonid are angulate labially and separated by a deep, persistent vertical labial groove. On the lingual side, there is a slight, short vertical groove in the trigonid (removed by wear in m1). A deep, more persistent groove is present between the trigonid and the talonid

25 482 G. BILLET ET AL. lingually and runs mesio-labially. Labial and lingual grooves approach each other closely, leaving only a narrow isthmus connecting trigonid and talonid. An additional groove is intermediate between the others as regards both depth and persistence, and is situated between the entolophid and the end of the hypolophid. The hypolophid projects posteriorly well beyond the entolophid on these slightly worn teeth. The length of the projecting portion decreases abruptly downward to the level at which the groove between entolophid and hypolophid ends. These slightly worn teeth are larger mesio-distally than their more worn counterpart. The metalophid forms a postero-medial prolongation of the trigonid in both premolars and molars at that stage. The p3-4 preserve a small fossettid in the trigonid and m2 one in both trigonid and talonid. The unworn m3 has trigonid and talonid separated by a shallow transverse horizontal groove that connects the lingual and labial vertical grooves. Protolophid and metalophid are joined by a small crest that continues on around the front of the tooth (paralophid) and on the lingual side, to enclose a small, shallow basin, which, with wear, becomes the evanescent trigonid fossettid. In the talonid, the hypolophid curves labially and reaches forward to the distal face of the trigonid; at this point, it joins the entolophid that runs lingually from the distal face of the trigonid to the lingual groove separating it from the hypolophid. This isolates a second small, shallow basin that becomes the talonid fossettid. The distal portion of the talonid extends far beyond the entolophid and is demarcated by labial and lingual vertical grooves. Thus, the last molar is distinctly trilobate. On the molars (the bottom of the premolars are not visible) of this specimen, the fossettid that develops between the trigonid and the talonid persists along their entire height. The stage exhibited by this specimen (i.e. p3-m3 with the central fossette not yet isolated) is never found in A. suniensis (see Appendix S2). The type specimen shows a later stage in wear (Fig. 21B). The labial groove between trigonid and talonid is deep on all teeth except p2 (p1 is unknown). On the lingual side, all vertical grooves have disappeared and this surface has become a nearly straight face, except on m3 whose lingual groove between the entolophid and the posterior extension of the talonid is still present. No entolophid hypolophid fossettid has been recorded in the position of this groove in A. patagonicus whereas one appears in A. suniensis. The additional labial groove on m3 that demarcates the distal extension of the talonid is prominent. Mesially, the horizontal groove, which is present between the trigonid and talonid on the younger specimen, has been converted by closure into a conspicuous central fossettid as on other cheek teeth. On the molars, this central fossettid is isolated first by the junction of the mesio-labial part of the talonid (hypolophid) to the middle of the distal part (metalophid) of the trigonid as on MACN A52-624, and secondly, by the junction of the disto-lingual part of the trigonid (metalophid) with the mesio-lingual part of the entolophid. Enamel is absent over the mesial faces and at the disto-lingual angles of p3-m3. No statement can be made concerning root formation, the bases of the teeth being concealed. Cement is present, although thin, on both upper and lower cheek teeth. The only deciduous lower tooth known is a very little worn left dp3 or dp4 (MACN A52-623). Apart from the much lower crown, it differs from the permanent teeth only in the larger size of the fossettids in trigonid and talonid. It resembles the deciduous lower teeth of A. suniensis. HEGETOTHERIIDAE AMEGHINO, 1894 Definition: Derived features [the unambiguous synapomorphies that resulted from our cladistic analysis (see discussion of the cladistic analysis)] of hegetotheriids are: reduced M3 (always smaller than M2); long and flat lingual wall on upper molars; flat and straight lingual face on lower molars; M1-2 with a regular shape throughout wear; unique trigonid talonid connection lingual on lower molars, leading to a constant deep labial sulcus separating trigonid and talonid; hypselodont cheekteeth (hypsodonty index undefined). Some cranial characters may be synapomorphies of this taxon but many cranial data are unavailable for the most basal hegetotheriid of our analysis (Sallatherium; see Discussion). SALLATHERIUM REGUERO & CERDEÑO, 2005 SALLATHERIUM ALTIPLANENSE REGUERO & CERDEÑO, 2005 Below we describe a new specimen of S. altiplanense with slightly worn teeth which complements the original description. Diagnosis: Hegetotheriinae with nasals especially long and thin, not enlarged posteriorly. Labial groove on upper cheek teeth more medially placed than in Prohegetotherium. I1 enlarged, more oval in crosssection and more obliquely implanted than in Prohegetotherium, Hegetotherium and Hemihegetotherium. I2-I3-C more reduced than in Prohegetotherium and separated by diastemata, in contrast to Prohegetotherium and Hemihegetotherium, which have a closed dental series. P3-M3 with ovoid occlusal outline, different from the more quadrangular section of other

26 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 483 Figure 23. Sallatherium altiplanense, slightly worn P4- M3, SAL 561. Mesial to right and labial at the top.scale bar = 1 cm. hegetotheriines. Lower canine absent, i3 reduced, and a diastema separates i3 and p1. Lower p3 with anterior lobe longer than posterior lobe, in contrast to Prohegetotherium and Hemihegetotherium (Reguero & Cerdeño, 2005). Holotype: UF 91621, partial skull and associated incomplete mandible. Referred specimen: SAL 561, a right maxillary with P4-M3 (Fig. 23). Description SAL 561 represents a young adult as the teeth are very slightly worn and still present a pattern of fossettes. Referral of this material to S. altiplanense is difficult as young specimens of this species are not known so far. However, it presents cheek teeth that resemble hegetotheriids especially in their occlusal outline, which appear to be more regular than in Archaeohyrax. It also presents an M3 clearly smaller in surface area than M2 (Appendix S4, Table 5). These two characters are characteristic of hegetotheriids as compared with Archaeohyrax suniensis (see cladistic analysis) to which this specimen could have been referred in initial view (as they come from the same deposits and are of approximately the same size). Moreover, the coexistence of the fossettes is very different to that exhibited in Archaeohyrax suniensis, which represents the most distinct difference with this taxon. SAL 561 presents simultaneously an isolated central fossette on M1-3. This never occurs in A. suniensis. Besides the central fossette, an anterior labial accessory fossette is present on M2-3. The co-occurrence of the central fossette and the labial accessory fossettes (here, only the anterior one) never occurs on Archaeohyrax and Protarchaeohyrax molars (Croft et al., 2003; see also Discussion of our cladistic analysis). If the assignment of this specimen to Sallatherium is correct, it demonstrates that the original pattern of hegetotheriid cheek teeth presents the typical face pattern of typotherians. Reguero & Cerdeño (2005: 680, fig. 6) referred to another hegetotheriid, Prohegetotherium schiaffinoi (Kraglievich, 1932), a specimen (MNHN-BOL-V ) from Salla with unworn upper cheek teeth. It also exhibits the co-occurrence of central and accessory fossettes. However, our re-examination of this specimen (quasi-unworn right P2-4) indicates that it is identical to unworn premolars of A. suniensis, a species to which it is referred here (Fig. 10). Patterson (1934b) described juvenile teeth of a hegetotheriid, the pachyrukhine Prosotherium, but they are very different from those of Sallatherium given that they never present the isolation of a central fossette. DISCUSSION ARCHAEOHYRACIDS IN SALLA The recognition of a single archaeohyracid species in Salla is in contradiction to previous studies, which mentioned two species (Reguero & Cifelli, 1997). Of these two supposed species, one was referred to Archaeohyrax sp. nov., and the other was thought to be closely related to Eohegetotherium ( Bryanpattersonia ) sulcidens (= Archaeohyrax gracilis). The latter supposed Salla species has since been considered as a new species of Protarchaeohyrax also known from the Fray Bentos Formation in Uruguay (Bond et al., 1998; Reguero, Ubilla & Perea, 1998, 2003a; Croft et al., 2003; Reguero et al., 2003b). In the present study, we consider that the Salla archaeohyracid remains belong to a single species showing great variation for the cheek teeth in both size and morphology. We observe a continuum that links all extreme and intermediate forms such that no clear distinction between two hypothetical taxa is possible. The variable characters mainly concern tooth size, the shape of the occlusal outline of the cheek teeth, and the lingual grooves on both upper and lower cheek teeth. Concerning the last of these, we illustrate a continuum in Figure 16. The size continuum is shown by measurements taken on cheek teeth on MNHN and YPM-PU specimens (see below). All specimens studied agree well with the ontogenetic series reconstructed for upper and lower cheek teeth. This is also in favour of a certain degree of homogeneity in the variation observed in the Salla archaeohyracid remains. Measurements on cheek teeth of Archaeohyrax suniensis have been reported relative to wear stages

27 484 G. BILLET ET AL. by Croft et al. (2003). These authors pointed out that the M1 & M3 and m1 & m3 present strong dimensional changes during ontogeny. From a larger sample of specimens, we obtained the same results (Appendices S3, S4). Both upper and lower molars widen labio-lingually with wear. M1 shortens mesio-distally with wear while M3 lengthens. M2 does not present a mesio-distal trend of size variation with wear; it is intermediate between M1 and M3. The same characters are present on lower molars. Not enough specimens are known for A. patagonicus to characterize such trends throughout wear for this species. Measurements on the rostrums and skulls (even if scarce) make clearer the difference of size between A. suniensis and A. patagonicus (Appendix S4, Table 4a,b). PHYLOGENETIC RELATIONSHIPS OF ARCHAEOHYRACIDS AND HEGETOTHERIIDS A cladistic analysis has been performed on archaeohyracids and hegetotheriids to formulate hypotheses on relationships among these organisms based on the new data that are provided herein. The analysis was conducted on 15 taxa, with three, Oldfieldthomasia debilitata (Ameghino, 1901), Notopithecus adapinus (Ameghino, 1897) and Mesotheriidae (coded with Trachytherus alloxus from Salla), included as a priori outgroups. The data matrix contains 39 characters, seven of which are with multiple states (five ordered; Appendices S5, S6). The polarization of the characters is based on the outgroup comparison criterion. For the following taxa, the indicated literature was used but was most often completed by direct observations: Protarchaeohyrax (Reguero et al., 2003b), Archaeotypotherium (Croft et al., 2003), Eohyrax (Simpson, 1967), Pseudhyrax (Simpson, 1967; Croft et al., 2003), Oldfieldthomasia and Notopithecus (Simpson, 1967), Hegetotherium (Sinclair, 1909; Patterson, 1936), Hemihegetotherium (Croft & Anaya, 2006) and Propachyrukhos Prosotherium (Loomis, 1914; Simpson, 1945b). Furthermore, the taxa named other Pachyrukhinae in the analysis includes Pachyrukhos, Tremacyllus and Paedotherium and is regarded as monophyletic by Croft & Anaya (2006); it is coded here after this latter study, Patterson (1936) and Sinclair (1909). The taxon Prohegetotherium schiaffinoi was essentially coded after SAL specimens. The holotype of Coelostylodon caroloameghinoi Simpson, 1970 (MACN 10753) is an incomplete skull, which is tentatively referred to here as Eohyrax sp. (M. Bond, pers. comm.). As the genus Eohyrax contains three species (Croft et al., 2003), which only differ in size, and as these species may be synonymous (Simpson, 1967: 105), MACN is not referred to a particular species and the genus Eohyrax (as Pseudhyrax) is coded as monophyletic, pending future studies. Characters included in the analysis come from personal observations (G.B.) and from previous studies (Cifelli, 1993; Cerdeño & Bond, 1998; Croft et al., 2003; Reguero & Castro, 2004; Croft & Anaya, 2006; see Appendix S5). Analysis of the data matrix was performed using PAUP (version 4.0b10, Swofford, 1998) with the exact Branch and Bound algorithm. Parsimony analysis resulted in five most parsimonious trees of 66 steps (CI = 0.79, RI = 0.88). The strict consensus (68 steps, CI = 0.77, RI = 0.86) is presented in Figure 24A. The majority-rule (66 steps, CI = 0.79, RI = 0.88) and Adams consensus (68 steps, CI = 0.77, RI = 0.86) are presented in Figure 24B and C, respectively. The differences between these concern only Hegetotheriinae and particularly the position of Sallatherium within or outside of it. The Adams consensus underlines this situation in branching Sallatherium at the root of Hegetotheriidae (because of its conflicting positions) and in individualizing a clade of hegetotheriines (excluding Sallatherium), which includes Prohege totherium, Hegetotherium and Hemihegetotherium and which is present in all the intersecting set of taxa common to the most parsimonious cladograms. The unambiguous synapomorphies and the synapomorphies with Acctran (Fig. 25A) and Deltran (Fig. 25B) optimizations obtained for the majorityrule consensus cladogram (corresponding to three of the five most parsimonious cladograms) are presented. These synapomorphies differ on the two other most parsimonious cladograms only in the part concerning Hegetotheriinae. The relationships obtained are quite different from previous results (Cifelli, 1993; Croft et al., 2003). In particular, the mesotheriids originate within archaeohyracids. Therefore, the latter are polyphyletic. The Hegetotheria sensu Simpson (1967) (i.e. all archaeohyracids and hegetotheriids) is also polyphyletic. A clade of late archaeohyracids (Archaeotypotherium Protarchaeohyrax Archaeohyrax) is obtained and is the sister taxa to all hegetotheriids. This is different from the study of Croft et al. (2003) in which Archaeohyrax (and/or Protarchaeohyrax) was the sister taxon of hegetotheriids. A unique derived feature diagnoses the clade Archaeotypotherium Protarchaeohyrax Archaeohyrax: character 15, minute incisive foramina (also present in Oldfieldthomasia). The observation of the minute incisive foramina in Archaeotypotherium was made on a published photograph of specimen SGOPV 2900 of A. tinguiriricaense Croft et al., 2003 (Croft et al., 2003: 18). This will need to be confirmed by future studies. A unique synapomorphy supports the clade Protarchaeohyrax Archaeohyrax: no., accessory fossettes on upper molars disappearing before the isolation of the central fossette. This character supported

28 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 485 Figure 24. Cladograms showing the phylogenetic relationships of archaeohyracids plus hegetotheriids (plus mesotheriids) from the cladistic analysis (five most parsimonious trees with 66 steps, CI = 0.79, RI = 0.88). A, strict consensus tree (68 steps, CI = 0.76, RI = 0.86). Nodes that have a Bremer support superior to 2 are indicated as > 2 ; all others have a Bremer support equal to 1. B, majority-rule consensus (66 steps, CI = 0.79, RI = 0.88); 60% refers to the percentage of the most parsimonious cladograms that present this node. C, Adams consensus tree (68 steps, CI = 0.76, RI = 0.86). a clade formed by these taxa and hegetotheriids in the study by Croft et al. (2003). In fact, hegetotheriids were incorrectly coded with respective to this character. Before the juvenile dentition of Sallatherium was described herein, no juvenile teeth of hegetotheriids with a central fossette were known (see above). Therefore, the coding of hegetotheriids for character 5 was previously unknown. Here, we argue that labial and central fossettes co-occur on hegetotheriid upper molars (as shown for Sallatherium) and that the non-coexistence of these fossettes is a synapomorphy of late archaeohyracids.

29 486 G. BILLET ET AL. Figure 25. Majority-rule consensus tree with positioned unambiguous synapomorphy (black font) and synapomorphy depending on optimization (grey font). Underlined characters express homoplasy for them. A, Acctran optimization. B, Deltran optimization. Abbreviation: Prop.-Pros. = Propachyrukhos - Prosotherium. The genus Archaeohyrax is monophyletic and its representatives share the presence of a well-marked labial sulcus on the talonid of m3 (character 13, convergent in hegetotheriids). Adding to this unambiguous synapomorphy, the genus Archaeohyrax might be diagnosed by additional characters (unknown in Protarchaeohyrax at least), as is the case with the Deltran optimization: characters 17, minute premaxillary extent on the palate; 18, nasal bulge; 28, strong antero-lateral constriction of the bulla; and 30, thin tympanic extension forming the anterior and lateral wall of the tympanohyal recess, widely separated from the crista meatus. Aside from the analysis, we suggest that the presence of small incisive foramina might be linked to the reduced extension of the premaxillary on the palate. The incisive foramina of the holotype of Sallatherium altiplanense as figured by Reguero & Cerdeño (2005) are poorly preserved. However, it appears from the small size of the region that must contain them that they might have been as small as in Archaeohyrax. Given the uncertainty of this statement, we have coded the incisive foramina of Sallatherium as unknown condition, as for characters 17 and 18 concerning the premaxillary extent on the palate and the nasal bulge. We hope that future investigations on Sallatherium will help to determine the condition for these characters precisely. For the time being, the

30 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 487 Figure 25. Continued present analysis strongly argues that Sallatherium is an hegetotheriid. The monophyly of hegetotheriids is well supported. Four unambiguous synapomorphies diagnose this clade (characters 1, 9, 11, 14) in all most parsimonious cladograms. More characters may diagnose this clade, depending on the position of Sallatherium and on optimization (see for example Fig. 25). The majority-rule consensus presents the Hegetotheriinae as a monophyletic group (60% of the cases). The Adams consensus (as explained above) shows that the existence of a clade with at least Prohegetotherium, Hegetotherium and Hemihegetotherium is retrieved in all the most parsimonious cladograms. This is supported by characters 31 (vestigial crista meatus) and 38 (short, broad distal fusion of tibia and fibula). This latter character has already been suspected to be an element arguing for the monophyly of the Hegetotheriinae (Croft & Anaya, 2006) as the distal fusion of the tibia and fibula appears to present a very different configuration from that of Pachyrukhinae (Sinclair, 1909; Cerdeño & Contreras, 2000; Croft et al., 2004). However, no published cladistic analysis has demonstrated the monophyly of these two subfamilies. The difference in the development of the crista meatus between Hegetotheriinae and Pachyrukhinae has not, to our knowledge, been previously mentioned. This new observation brings a critical argument supporting more consistently the monophyly of Hegetotheriinae (with or without Sallatherium) and needs to be detailed here. Indeed, the crista meatus appears to be vestigial or absent in hegetotheriines (Fig. 26A, B). In fact, two interpretations are possible for the crista meatus position of

31 488 G. BILLET ET AL. Figure 26. Details of the auditory regions of hegetotheriids. A, right auditory region of Prohegetotherium schiaffinoi, SAL 6. B, right auditory region (reversed) of Hegetotherium sp., SCZ 189. C, right auditory region of Prosotherium garzoni, AMNH D, right auditory region of Paedotherium sp., MNR 45 from Barranca Lobos, Mar del Plata, Argentina. Scale bar = 1 cm. hegetotheriines. (1) The vestige of the crista meatus may comprise a small lateral bulge, the position (on the antero-ventro-lateral surface of the meatus) of which corresponds to that of the crista meatus lateral extremity in Archaeohyrax and Prosotherium (this bulge is barely visible in Fig. 26A). In this case, the tympanohyal recess appears to be bounded anteriorly and laterally only by the semicircular independent tympanic extension. This latter structure has been mentioned above in Archaeohyrax and is also present in many notoungulates in which it is often barely observable because it is appressed against the crista meatus anteriorly. (2) The wall that bounds the tympanohyal recess anteriorly and laterally comprises the crista meatus, which is fused to the tympanic extension. In the two cases, however, the crista meatus appears to have a vestigial extent and the configuration of this region is unique to hegetotheriines (Fig. 26A, B). Indeed, the crista meatus in Prosotherium (Fig. 26C) is strong and remarkably similar to that of Archaeohyrax. In other pachyrukhines such as Paedotherium (Fig. 26D), the crista meatus, although reduced in length (it does not reach the lateral border of the meatus), is still clearly present and is not reduced in height. In Pachyrukhinae, the independent tympanic extension appears to be fused anteriorly with the crista meatus but is distinct from it postero-laterally and thus bounds the tympanohyal recess laterally (Fig. 26C, D). Prohegetotherium sculptum Ameghino, 1897 from Patagonia was not included in the analysis because it is only known by scarce remains that would have been coded like the better-known Prohegetotherium schiaffinoi. No derived features characterizing the

32 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 489 genus Prohegetotherium have been found and it is suggested that it may be paraphyletic in its current conception. Our study does not present a satisfying resolution among Hegetotheriinae, particularly concerning the position of Hegetotherium and Hemihegetotherium, which might be expected to form a clade because they resemble each other closely. Furthermore, the analysis does not resolve the position of Sallatherium within or oustide of the hegetotheriine clade. Numerous data are missing for Sallatherium and further resolution on its position should be provided with additional material. Mesotheriidae is the sister taxon of a clade that groups late archaeohyracids (excluding Eohyrax and Pseudhyrax) and hegetotheriids. This latter clade is supported by a unique synapomorphy: character 29, carotid foramen shifted anteriorly, medial to auditory bulla. This character is unknown in Eohyrax and Pseudhyrax. Mesotheriidae is grouped with this latter clade on the basis of dental characters: 3, diastema between upper anterior teeth; 14, very hypsodont cheek teeth HI > The more derived position of Pseudhyrax relative to Eohyrax is similar to the results of Croft et al. (2003). There is no derived character known so far that supports the monophyly of archaeohyracids (including Eohyrax and Pseudhyrax) plus hegetotheriids without mesotheriids. In the present analysis, the presence of a diastema and strong hypsodonty places Mesotheriidae as the sister taxon of late archaeohyracids and hegetotheriids. Cifelli (1985, 1993) has argued that mesotheriids may have derived from a Pseudhyrax-like archaeohyracid, particularly on the basis of the resemblance of their lower cheek teeth. The results of our analysis do not contradict this idea. However, the above-mentioned characters (3 and 14) that support the new position of the Mesotheriidae might be interpreted as a convergence in future analyses. In fact, these features are known to have occurred independently in many lineages of notoungulates and may be related to particular ways of feeding (Stebbins, 1981; Pascual & Ortiz-Jaureguizar, 1990; MacFadden, 1997, 2000, 2006; Flynn & Wyss, 1998; Jacobs, Kingston & Jacobs, 1999; Kay et al., 1999; Croft et al., 2003; Flynn et al., 2003; Khon et al., 2004; Ortiz-Jaureguizar & Cladera, 2006). In fact, many of the characters in cladistic analyses on notoungulates concern cheek teeth and are often related to feeding and to the increasing resistance of teeth against abrasion (e.g. hypsodonty, cementum, diastema). We suggest that care must be taken in future analyses on notoungulates not to give undue importance to this type of character, all related to teeth abrasion and feeding, and which therefore may represent the same functional complex. However, the message provided by hypsodonty and closely related characters cannot be ignored. Even if these traits are homoplastic at a certain level, they must also be strict synapomorphies of less inclusive groups, and this provides crucial information. The results of our analysis are not particularly robust, as illustrated by the Bremer support values, which are no more than 1 at the majority of nodes (Fig. 24A). Furthermore, removing just one particular character from the analysis may lead to a great loss of resolution: this is the case, for example, if we remove character 14 on the hypsodonty level. This results in 148 most parsimonious cladograms of 63 steps with very weak resolution: only four nodes are obtained: (1) Archaeohyracidae + Mesotheriidae + Hegetotheriidae, (2) the genus Archaeohyrax, (3) the Hegetotheriidae and (4) the Pachyrukhinae. A probable major cause of the weakness of these results may lie in the great number of missing data (especially cranial data) for many taxa (Eohyrax, Pseudhyrax, Archaeotypotherium, Protarchaeohyrax, Sallatherium). Despite the lack of strong support of our new hypotheses of relationships, we argue that these results are reasonably pertinent with regard to the morphology and temporal distribution of taxa (see below). For instance, both Archaeohyrax species appear to have a markedly specialized cranial morphology, which may suggest an early differentiation, before the Deseadan. In addition, the monophyly of the Hegetotheriinae (with or without Sallatherium) highlights the peculiar cranial and post-cranial morphology of its members. Therefore, these new hypotheses certainly provides interesting routes to explore advanced typotherians. NOMENCLATURAL ISSUES How to name the clade regrouping archaeohyracids plus hegetotheriids has been extensively discussed by Croft et al. (2003). As they discuss, the possible names for this clade are numerous and none is ideal. In the present study, mesotheriids are included within the clade that groups all archaeohyracids and hegetotheriids. We propose to explore how to name the clade that contains all archaeohyracids and hegetotheriids (with or without mesotheriids) and the less inclusive clade [late archaeohyracids (Archaeotypotherium Protarchaeohyrax Archaeohyrax) + hegetotheriids] (also present in Croft et al., 2003), which excludes mesotheriids, Eohyrax and Pseudhyrax. For this latter clade, a solution could be the name Hegetotheroidea (Romer, 1966), which was created a year prior to Simpson s proposal of Hegetotheria for grouping archaeohyracids and hegetotheriids (Simpson, 1967) [Simpson (1945a: 130) previously proposed the name Hege-

33 490 G. BILLET ET AL. totheria but at this time it included only Hegetotheriidae]. However, the concept of Hegetotheroidea proposed here differs from the original of Romer, as it does not include all archaeohyracids (Eohyrax and Pseudhyrax are excluded). This clade could be defined as all typotheres (or all hegetotherians, see below) more closely related to Archaeohyrax and Hegetotherium than to Pseudhyrax, Mesotherium or Interatherium. The other, more inclusive, clade could be named Hegetotheria, which includes all archaeohyracids and hegetotheriids. It is important to maintain the possibility of including Mesotheriidae in this group, as is the case from the results of our analysis. Reguero & Castro (2004) introduced the taxon Typotherioidea for the clade that includes the most recent common ancestor of Mesotheriidae and Hegetotheria, i.e. containing all archaeohyracids and hegetotheriids sensu Simpson, Because our results suggest nesting of the Mesotheriidae within the Hegetotheria sensu Simpson (1967) [as considered by Reguero & Castro (2004)], the proposal of Typotherioidea would become unconvenient and the use of an enlarged version of Hegetotheria would be preferable. Nevertheless, even though our proposals are in agreement with Croft et al. (2003), we defer these namings to future studies providing more stabilized phylogenies. Our study also argues for the existence of a clade of late archaeohyracids (Archaeotypotherium Protarchaeohyrax Archaeohyrax) distinct from hegetotheriids. The family Archaeohyracidae sensu Simpson (1967) is polyphyletic because the Mesotheriidae are nested within it. Therefore, it would be preferable to limit the use of Archaeohyracidae to the clade Archaeotypotherium Protarchaeohyrax Archaeohyrax found in the present study. However, our results concerning late archaeohyracids are preliminary and not strongly supported. Future study of the skull of Archaeotypotherium might provide new information that may argue for another phylogenetic pattern. We therefore also defer restriction of the name Archaeohyracidae to future studies. TEMPORAL DISTRIBUTION The phylogenetic tree displayed in Figure 27 juxtaposes a consensus on the previous phylogenies and our phylogeny with the known stratigraphic range for each taxon The new position of mesotheriids within archaeohyracids sensu Simpson (1967) is not incongruent with the temporal distribution of the concerned taxa. According to the results, the sister taxon of mesotheriids is the clade that groups Archaeotypotherium (and allies) and Hegetotheriidae. The first occurrence of the latter clade corresponds to that of Archaeotypotherium and Protarchaeohyrax in the Tinguirirican (Croft et al., 2003; Flynn et al., 2003; Reguero et al., 2003b). The first occurrence of mesotheriids was thought to be from the Divisaderan (Simpson & Minoprio, 1949; Simpson, Minoprio & Patterson, 1962), a doubtful age considered to be pre-deseadan or even pre-tinguirirican (Flynn & Swisher, 1995; Flynn et al., 2003). However, the stratigraphic provenance of the holotype of the mesotheriid Trachytherus? mendocensis Simpson & Minoprio, 1949 from Divisadero Largo has been questioned recently. These remains might not come from the late Eocene early Oligocene but from younger levels (Cerdeño, González Riga & Bordonaro, 2006). Another candidate for the earliest record of Mesotheriidae is provided by Bond, López & Reguero (1997) and Reguero & Castro (2004) who mentioned mesotheriid postcranial remains from the Tinguirirican of Patagonia (Rocas Bayas, Argentina). This would be compatible with our phylogenetic results. The advantage of this hypothesis of a clade Mesotheriidae late archaeohyracids Hegetotheriidae excluding Eohyrax and Pseudhyrax is that it does not imply a gap in the fossil record of mesotheriids, whereas a great gap (down to the Casamayoran) is currently implied when considering mesotheriids as the sister group to all Archaeohyracidae (sensu Simpson, 1967) Hegetotheriidae. The first occurrence of hegetotheriids is from the Divisadero Largo fauna (see above) or from the Tinguirirican of Cañadón Blanco (Flynn et al., 2003) or from other post-mustersan/pre-deseadan sites from Patagonia (Bond, 1991). The first record of late archaeohyracids (Archaeotypotherium Protarchaeohyrax Archaeohyrax) is Tinguirirican (Croft et al., 2003; Flynn et al., 2003; Reguero et al., 2003b). Therefore, the existence of a clade of late archaeohyracids as the sister-taxa of hegetotheriids fits well with the temporal data. The existence of a monophyletic Hegetotheriinae (suggested in the Adams and majority-rule consensuses), with or without Sallatherium, fits well with the temporal data. Indeed, the first certain reference to Hegetotheriinae (regardless of Sallatherium) corresponds to the late Oligocene (Desesadan) Prohegetotherium (Reguero & Cerdeño, 2005) and the first certain reference to Pachyrukhinae corresponds to the late Oligocene Prosotherium and Propachyrukhos (Loomis, 1914; Simpson, 1945b). The previous hypothesis (Croft & Anaya, 2006) with Hemihegetotherium as the sister-taxa of pachyrukhines is not congruent with either the temporal data or the postcranial anatomy. As Hemihegetotherium is not known before the middle Miocene, this relationship with

34 LATE OLIGOCENE ARCHAEOHYRACIDS FROM BOLIVIA AND ARGENTINA 491 Figure 27. Phylogenetic trees of archaeohyracids plus hegetotheriids (plus mesotheriids) according to (1) a consensus on previous phylogenetic hypotheses (Cifelli, 1993; Croft et al., 2003; Simpson, 1967); (2) the present phylogenetic hypothesis. Temporal data are taken from the literature (Simpson, 1945b, 1967; Cifelli, 1985; Flynn & Swisher, 1995; Kay et al., 1998; Reguero et al., 1998; Kay et al., 1999; Croft et al., 2003; Reguero et al., 2003b; Reguero & Castro, 2004; Reguero & Cerdeño, 2005; Cerdeño, Riga & Bordonaro, 2006; Croft, 2007). Pachyrukhinae would thus have implied a gap of approximately 15 Myr (Croft & Anaya, 2006). Finally, our results agree with Croft and Anaya s prediction that finding additional non-dental characters may provide support for a monophyletic subset of hegetotheriines. CONCLUSIONS The description of the dentition and skull of the new species Archaeohyrax suniensis from Salla and that of the already known Archaeohyrax patagonicus provide the bases for future analyses of archaeohyracid and