Type species: Siamogale thailandica Ginsburg et al., 1983.

Included species: Siamogale thailandica Ginsburg et al., 1983, S. melilutra Wang et al., 2018, S. bounosa, n. sp.

Diagnosis (from Wang et al., 2018: 41–42):

Siamogale has typical lutrine cranial and dental morphologies: a large infraorbital canal, presence of antorbital fossa, uninflated bulla, robust and protruding mastoid process, mastoid process separated by a broad shelf from the paroccipital process, postglenoid foramen positioned anteriorly to the auditory meatus, inion positioned anteriorly relative to the lambdoid crest, stylomastoid foramen separated by a bony ridge from the tympanohyal-bulla connection, masseter muscle attachment area ventrally expanded to beyond the ventral rim of the masseteric fossa, parallel zygomatic arches, shortened angular process, premolars with surrounding cingulum, shortening of P4 metastylar blade, presence of a notch between talonid and trigonid of m1, and widening of m1 talonid. Siamogale differs from Paralutra jaegeri by the presence of a distal ridge of m1 metaconid connected to the entoconid crest, M1 cuspule distal to metacone, and absence of P4 metastylar notch. Siamogale differs from Paludolutra, Tyrrhenolutra and Enhydritherium in having a crestiform protocone, lack of hypocone and presence of parastyle on P4, a distolingually expanded M1 talon, and metaconule placed distally to the metacone.

Geographical and stratigraphical distribution: Q–K and J lignite zones, late middle Miocene of Mae Moh Basin, Thailand (13.4–13.2 Ma and 12.8–12.2 Ma, respectively; Ginsburg et al., 1983; Grohé et al., 2010; this study); Shuitangba fossil site, latest Miocene of Zhaotong Basin, Yunnan Province, China (∼6.2 Ma; Wang et al., 2018); Dapinggou, Pliocene Gaozhuang Formation of Yushe Basin, Shanxi Province, China (∼4.9–5.4 Ma; Teilhard de Chardin and Leroy, 1945; Wang et al., 2018).

Holotype: MM-54, nearly complete skull dorsoventrally compressed; on the left side, the zygomatic arch and the basicranial region are not preserved; on the right side, the region between the palatine and basioccipital is strongly damaged; right tympanic bulla partially preserved; left I1, C, P3–M1 and alveoli for I2, I3, P2; right I1, I3, C, P3, M1 and alveoli for I2, P2, P4; left i2 or i3; cervical vertebra; stored at PALEVOPRIM, Poitiers (France).

Etymology: From Greek bounos (masculine), “hill”; for the bunodont teeth morphology of this species.

Diagnosis: Siamogale with three upper incisors and three upper premolars on each side; P4 lingual shelf extended to the distal part of the metastyle; M1 with equally wide metacone and paracone buccal shelves. Differs from S. thailandica and S. melilutra in having equally developed buccal shelves of paracone and metacone on M1; from S. thailandica in having additionally a wider P3, a P4 with a slightly convex buccal wall, a nearly straight mesial wall, a more-extended lingual shelf, and an enlarged distolingual part of the M1 metacone; from S. melilutra in having additionally a narrower braincase where the mastoid processes do not extend to the level of the zygomatic arches, a more-cusped protocone and an hypoconal crest on P4, a P4 protocone more mesially located, a convex buccal border of P4, and an M1 paraconule running closer to the mesial border of the tooth.

Description: Skull (figs. 2–3; table 1): The skull has suffered from significant dorsoventral crushing leading to distortion of the premaxilla, maxilla, and nasal, and an overlapping of the right parietal and frontal on the left portions of these bones (fig. 2). On the left side, most of the region of the skull behind M1, including the zygomatic arch and the basicranial region, is missing. On the right side, the region between the palatine and basioccipital is partly lacking and otherwise broken into pieces. The auditory bulla is fragmented, but the lateral part of the ectotympanic and the external auditory meatus are still preserved (fig. 3). Due to the crushing, the foramina of the orbital region are not visible. In dorsal view, the nasal and premaxilla are collapsed with respect to the maxilla. Several displaced elements, probably from the same individual, are present on the skull: a cervical vertebra pressed against the buccal border of the right premolar row is visible in ventral view, and a left i2 or i3 is present in dorsal view, on the caudal part of the right zygomatic arch.

On the dorsal surface of the skull, the sutures between bones are not clearly visible. On its ventral surface, the suture between premaxilla and maxilla can be traced from the middle of the base of the canine lingual borders, the suture between maxilla and palatine can be seen especially on the left side of the skull close to the distolingual corner of M1, the suture between maxilla and jugal is observable buccal to the right M1 on the rostral portion of the zygomatic arch, and the suture between jugal and squamosal is distinct on the caudal half of the right zygomatic arch. Siamogale bounosa has a short and wide rostrum, as is typical in otters. Due to the deformation of the skull, neither the development of the postorbital process of the frontal nor the degree of the postorbital constriction can be evaluated. The infraorbital foramen is preserved only on the left side of the skull, but its shape and diameter cannot be evaluated accurately due to the crushing of the skull. In Siamogale thailandica and S. melilutra, the infraorbital foramen is oval, slightly wider dorsoventrally than laterally (TF 6296, Grohé et al., 2010: 1007, fig. 2g for S. thailandica; ZT-10-03-064b, Wang et al., 2018: fig. 4 for S. melilutra). The right zygomatic arch, rather thin and deep, is laterally curved, as opposed to its straighter condition in S. melilutra (ZT-10-03-064b, Wang et al., 2018: figs 2–4). It also bears a strong postorbital process of the zygomatic arch, as in S. melilutra. The dorsal surface of the skull exhibits no apparent sagittal crest, unlike in S. melilutra, but two weakly marked temporal crests. The right temporal crest is better preserved. It extends from the level of the rostral border of the orbit to the supraoccipital, by converging toward the break between the right and left parietal and frontal bones. The supraoccipital is elevated above the dorsal surface of the skull, which could be a result of the dorsoventral crushing. The nuchal crest is rather convex. The braincase is short, although longer than the rostrum, and is much narrower relative to the zygomatic arches. In ventral view, the right squamosal shows a short and wide glenoid fossa. A postglenoid foramen opens at the caudomedial base of the postglenoid process and at the rostrolateral corner of the external auditory meatus (fig. 3). The tympanic bulla is fragmented, but the ectotympanic is partially preserved and the external auditory meatus appears rather elongated. The bulla is inflated, and its surface is perforated by small pits on its rostral portion, as in specimen ZT-10-03-064b of S. melilutra (Wang et al., 2018: fig. 5A). The stylomastoid foramen is visible on the reconstructed digital skull on the caudolateral border of the bulla. Laterally and caudally to the bulla, the mastoid process is poorly developed ventrally and laterally, in contrast to the condition observed in S. melilutra, where the mastoid processes nearly extend to the level of the zygomatic arches and project much more ventrally.

Upper teeth (fig. 2; table 2): The left premaxilla bears I1 and the alveoli for I2–I3 whereas the right premaxilla bears I1 and I3 and the alveolus of I2. I1 shows a convex buccal border and a widened mesiolingual crown base. Its main cusp is relatively blunt. I3 has a convex buccal border, a main cusp displaced buccally relative to the long axis of the tooth and a lingual shelf. It is much larger than I1 and a cingulum is present mesially and lingually. The canines are sharp and weakly curved distally. They show longitudinal crests on their distal and mesiolingual faces. Behind the canines, the maxilla shows two alveoli for the P2, especially clearly visible on the right side. There is no P1. P3, preserved on both sides, is two-rooted, wide and short. It is slightly wider distally than mesially. It exhibits a sharp main cusp, slightly mesially placed with respect to the long axis of the tooth, and it bears two crests, mesial and distal, joining the crown base. A cingulum surrounds the tooth. It is thicker distally.

P4 is only preserved in the right maxilla. It is three-rooted (as seen by alveoli present on the left maxilla), triangular, and is slightly longer than wide. The tooth shows a convex buccal border, a nearly straight mesial border between the parastyle and protocone, and a lingual shelf extending until the distal end of the metastyle. The paracone mesial face is marked by two crests: one runs buccally to reach a weakly developed parastyle; the other runs lingually to join the base of the protocone. The paracone is much higher than the protocone and its tip is slightly distal relative to the protocone tip. Distal to the paracone, the carnassial notch is absent and the metastyle is short. On the lingual shelf, two crests separate the convex lingual face of the protocone from its flattened inner face. A weak crestiform hypocone lies just distal to the protocone. The enamel is wrinkled, particularly on the lingual shelf. A thin cingulum surrounds the tooth.

M1 bears low paracone and metacone. The widths of their buccal shelves are similar. The metacone is slightly shorter than the paracone. A metaconule is present on the distolingual border of the metacone. The paraconule and protocone are crestiform and separated by a distinct notch. Together they form a crescent-shaped blade extending distolingually from the mesial border of the M1 to the middle of the mesiodistal axis of the tooth. The distolingual border of M1 is enlarged. A cingulum surrounds the tooth, and the width of this cingulum is particularly important distolingual to the metacone and on the lingual border of the tooth, especially in its middle. As for P4, the enamel of M1 is wrinkled.

Additional elements appended to the skull (fig. 2): A left i2 or i3 is located on the dorsal side of the caudal part of the right zygomatic arch. At the base of the main cuspid a cingulid is present at the lateral and distal border of the tooth. A cervical vertebra caudal to the axis is tightly pressed against the right premolar region. An indeterminate fragmented tooth is partly embedded in the coal matrix on the ventral side of the skull. It is narrower than the upper canines and could correspond to a different individual or even to a different species.

Comparisons and discussion: MM-54 shows diagnostic morphological features of the genus Siamogale. Compared with the dental morphology of the type species S. thailandica, this new species exhibits a wider P3 relative to its length (table 2), a P4 with a slightly convex buccal border, a less-constricted area between the paracone and protocone, a more-extended lingual shelf, and an M1 with equally developed paracone and metacone buccal shelves and an enlarged distolingual part of the metacone. These morphological traits do not match the range of morphological variation observed among the numerous specimens attributed to S. thailandica from the same basin. Therefore, we attribute MM-54 to a new species. This taxonomic determination is consistent with the more recent age of MM-54 (J lignite zone, dated to 12.8–12.2 Ma) by comparison with the specimens of the type species S. thailandica (Q–K zones, dated to 13.4–13.2 Ma) discovered in the Mae Moh Basin. S. bounosa is a derived chronospecies of S. thailandica. In addition, S. bounosa differs from the more derived Chinese S. melilutra (Wang et al., 2018) in its smaller dimensions (tables 1–2), a narrower braincase where the mastoid processes do not extend to the level of the zygomatic arches, absence of a sagittal crest, a more-cusped protocone and the presence of an hypoconal crest on P4, a P4 protocone that is more mesially located, a buccal border of P4 that is convex rather than concave, equally developed buccal shelves of paracone and metacone on M1 rather than a more-developed paracone buccal shelf, and a paraconule running closer to the mesial border of M1.

The dental morphology of the type species S. thailandica recalls that of Galictinae, Lutrinae, and Melinae (Grohé et al., 2010). However, Siamogale bounosa shows a more-derived dentition compared with the type species, with a lingual shelf of P4 reaching the end of the metastyle associated with equally developed buccal shelves of the paracone and metacone of M1, which are clearly different from the dental features observed in extant and fossil galictines (e.g., Lutravus, Trigonictis, Sminthosinis, Enhydrictis, Pannonictis). Moreover, the reduction in P4 length with an extension of the lingual shelf beyond the metastyle blade is different from the condition observed in the extant badgers Meles and Arctonyx. The shape of the P4 protocone of S. bounosa is also crestiform, as opposed to cusped in Meles and reduced to a bulge in Arctonyx. Finally, increasing morphological evidence of a relationship between Siamogale and other fossil lutrines support its belonging to this subfamily (Wang et al., 2018).

Type species: Vishnuonyx chinjiensis Pilgrim, 1932.

Included species: Vishnuonyx chinjiensis Pilgrim, 1932, V. angololensis Werdelin, 2003, V. maemohensis, n. sp.

Diagnosis (from Pilgrim, 1932: 93–94):

Lutrinae of rather small size; having a P4 of triangular shape; antero-posterior diameter on the outer side much exceeding that on the inner side, and also much exceeding the transverse diameter; with a high, pointed paracone; metacone lower but elongated; parastyle weak; with two internal cusps, protocone and hypocone, both much lower than the paracone; protocone situated rather far forward; with a slight internal cingulum; M1 rather small; mandible with a deep ramus; p4 elongate, only slightly broader posteriorly; with a strong posterior accessory cusp, not situated so much to the outside of the main cusp as in Sivaonyx, without an anterior accessory cusp; with a broad cingulum; m1 with talonid broader and shorter than trigonid, surrounded by a crenulated rim, of which the entoconid is as strong as the hypoconid; m2 elongate oval, rather longer than in Sivaonyx.

Remarks: Because of the inclusion of a new species of Vishnuonyx from Thailand and of the description of some material of V. chinjiensis (type species) from Pakistan further below, we add some remarks to the diagnosis of Pilgrim: a small mesial accessory cuspid is present on the p4 of Vishnuonyx, the m1 talonid is slightly narrower than the trigonid, the entoconid is lower than the hypoconid, and the presence of a crenulated rim surrounding the talonid as well as the depth of the mandibular ramus (i.e., as deep as or deeper than the m1 length) are variable within specimens of the genus.

Geographical and stratigraphical distribution: R and Q–K lignite zones, late middle Miocene of Mae Moh Basin, Thailand (14.2–14.1 and 13.4–13.2 Ma, respectively; this study); Middle Miocene of Chinji Formation, Potwar Plateau, Pakistan (∼14–11.4 Ma; Pilgrim, 1932); Middle Miocene of Ramnagar, India (∼13.8–13.2 Ma; Nanda and Seghal, 1993); Middle Miocene Locality 2/11, Kabarsero, Member D, Bed 3, Ngorora Formation, Kenya (∼12 Ma; Morales and Pickford, 2005); Late Miocene of the lower member of the Nawata Formation, Lothagam, Kenya (7.1±0.1–6.5±0.1 Ma; Werdelin, 2003); Late Miocene of Haradaso Member, Sagantole Formation, Middle Awash, Ethiopia (4.85–5.2 Ma; Haile-Selassie, 2008).

Holotype: MM-30, right mandible with c, p3–m2 and alveoli for p1–p2, and left mandible with c–m2 from the same individual; stored at the DMR, Bangkok.

Referred material: MM-32, left mandible (m1); MM-33, left mandible (p4–m1, alveolus for m2); MM-34, left mandible (p2–m1, alveolus for m2); MM-35, right mandible (p3–m1, alveoli for p2 and m2); MM-36, right P4; MM-37, left M1 (all stored at the DMR, Bangkok); MM-78, rostral part of a skull broken at the level of the glenoid fossae and right zygomatic arch missing; on the right side: I1–I2, alveoli for I3–C poorly visible, P1–M1, on the left side: alveoli for I1–I2 poorly visible, I3, alveolus for C, P1–M1; MM-79, right mandible (partial distal root for p3, broken roots for p4 and m1, and alveolus for m2) (both stored at PALEVOPRIM, Poitiers, France).

Etymology: The species name derives from the locality where it was originally found, Mae Moh.

Diagnosis: Vishnuonyx with p1 present; protoconid of m1 much higher than metaconid, metaconid displaced distally relative to protoconid, talonid narrower than trigonid, high hypoconid, talonid basin poorly developed, talonid distal border oriented distobuccally; m2 with high metaconid, aligned and oblique paraconid-protoconid-hypoconid, talonid distal border oriented distolingually; P4 with protocone distal crest bearing a tiny cusp, thick lingual cingulum at the metastyle base; M1 transversely widened, small metaconule, paraconule-protocone blade, poorly developed distolingual cingulum.

Differential Diagnosis: Differs from V. chinjiensis and V. angololensis in having a P4 with a thinner cingulum at the base of the lingual cusps and a thicker one lingual to the metastyle end; from V. chinjiensis additionally by a shallower mandibular ramus, a larger P4 bearing a notch rather than a U-shaped valley between the protocone and hypocone; from V. angololensis additionally by a smaller P4 having a more distinct parastyle.

Description: Mandible and lower dentition (fig. 4; table 3): The left mandible of the holotype MM-30 bears c–m2. Only the horizontal ramus is preserved. The masseteric fossa is moderately deep and extends below the distal part of m2. In buccal view, two mental foramina open beneath the p1–p2 contact and the distal part of p3. In lingual view, the mandibular symphysis extends caudoventrally until the distal part of p3. The teeth are set close together and are oriented mesiobuccally from the canine to the p3. The p1 and m2 are single-rooted. Two thick longitudinal crests extend from the canine tip and run on the middle of its lingual and distal faces to join the basal cingulid. Another, thinner one, runs on the canine buccal face. There is a lingual and distal cingulid. The p1 is much lower than p2 and p3, which have approximately the same height. The p4 is twice as high as p2 and p3. The p1–p3 are built on the same pattern: the main cuspid is mesial relative to the crown length and two crests extend mesially and distally until reaching the base of the crown. The p4 bears a tiny mesial accessory cuspid and a high and distinct distal accessory cuspid that reaches the mid-height of the main cuspid. The p4 distal accessory cuspid is displaced buccally relative to the main cuspid and is slightly lower than the m1 paraconid in buccal view. From p1 to p4, the main cuspid tends to be more distal and the cingulid surrounding the premolars thickens to form a distal cingular shelf in p4. The m1 trigonid is longer than the talonid. The m1 paraconid, which bends lingually, is lower than the main cuspid of p4. The right mandible of the holotype shows an unbroken m1 protoconid, which allows a more-detailed description of the trigonid morphology. The paraconid has approximately the same length as the protoconid, the latter being the highest cuspid of m1. The metaconid reaches approximately the same height as the paraconid. The paraconid has a distolingual crest that is half as long as the distobuccal crest. This short distolingual crest of the paraconid and the rather distal position of the metaconid relative to the protoconid create a wide trigonid lingual valley. The metaconid has a straight distal face that abruptly meets the lingual talonid, as commonly observed in otters. The protocristid is shorter than the paracristid. The talonid is much lower and slightly narrower than the trigonid. The cristid obliqua extends distolingually from the base of the protoconid to join a sharp and high hypoconid. The distal border of the tooth is straight and distobuccally oriented. The entoconid crest is lower than the hypoconid. The talonid is poorly basined. A cingulid is present on the buccal and mesial side of the tooth. It is thinner on the lingual side where it ends behind the paraconid. The m2 is very small and slightly longer than wide. The paraconid, protoconid, and hypoconid are obliquely aligned. The metaconid is high, located at the mediolingual side of the molar, and is displaced distally relative to the protoconid. The partially broken paraconid is close to the protoconid. The cristid obliqua extends distobuccally from the distal base of the protoconid to the hypoconid. The distal border of the tooth has a distolingual orientation, opposite to that of m1. The cingulid is visible on both the buccal and lingual sides of m2.

Skull (fig. 5; table 4): The skull is heavily compressed and teeth point toward the right side: the left side of the jaws folds medially, so that the upper premolars and molars sit against the maxilla while the right side is displaced laterally. The ventral floor of the skull, from the premaxilla to the pterygoid, is broken into many pieces. On the premaxilla, the incisive foramina are elongated. The maxilla and palatine regions are collapsed inside the skull. The rostroventral and rostrodorsal branches of the zygomatic arches are broken off and sit tightly against each other such that the outline of the infraorbital foramen is not fully visible. On the left side, only the caudal part of the zygomatic arch attached to the glenoid fossa is missing. The zygomatic arch appears thin behind the postorbital process. On the right side, the zygomatic arch is missing behind the infraorbital foramen. The basicranial region is not preserved, but part of the left tympanic bulla has been displaced rostrally to the glenoid fossa. In dorsal view, the rostrum is short, the postorbital processes of the frontal are marked, and a low but distinct sagittal crest is present. From the size and general proportions of the skull, V. maemohensis appears as large as a modern river otter.

Upper dentition (fig. 5; table 5): Most of the upper teeth are preserved on the skull. V. maemohensis has three upper incisors, as indicated by the right I1 and I2 and left I3. I1 and I2 are very similar in size, with I2 slightly larger, and both show a single spatulate main cusp. I3 is larger than I1 and I2. The I3 crown is higher; the tip of the main cusp is located buccally and the crown widens at the distolingual base of the tooth. I3 is surrounded by a cingulum that becomes thinner buccally. No canines are preserved. The alveolus for the left canine is less deformed than the one for the right canine and is oval. P1–P3 show the same general morphology: a main cusp from which two crests extend mesially and distally, a lingual bulge and a basal cingulum surrounding the tooth. The tip of the P1 main cusp is mesially displaced relative to the tooth length; this cusp is less mesially displaced in P2 and lies in the middle of the tooth in P3. The lingual bulge, small in P1, gets wider in P2 and almost creates an extra root mesiolingually to the distal root in P3. The skull teeth showing strong wear, the morphology of P4 and M1 is better preserved on the isolated specimens MM-36 and MM-37. P4 has a long metastyle and a high paracone from which two crests extend mesially. The mesiobuccal one joins a low and sharp parastyle and the mesiolingual one joins the base of a high protocone, forming a wide U-shaped valley in mesial view. The protocone is displaced mesially with respect to the paracone and distally relative to the parastyle. The hypocone lies distolingually to the paracone and distobuccally to the protocone. It is half the height of the protocone and closer to the lingual border of the carnassial. The distal crest going down from the protocone bears a very low cusp near the notch separating the two lingual cusps. These cusps delineate a relatively broad lingual shelf, which does not extend to the distal end of the metastyle, contributing to a triangular P4. The distal crest going down from the hypocone meets the lingual cingulum at the middle of the metastyle lingual face. The lingual cingulum, well marked at this level, is thicker than the buccal one. M1 specimen MM-37 is broken at its distobuccal corner such that the metacone is not preserved. The tooth is wide transversely and the mesial and distal borders are roughly parallel. The cusps are high and sharp. The parastyle and the paracone buccal shelf are moderately developed. Directly lingual to the paracone, a paraconule-protocone blade arises from the mesial border of the tooth to run lingually and gently distally. The paraconule is lower than the protocone. A very shallow notch separates these two cusps. A small and very distal metaconule is present lingual to the metacone region. It is lower than the paraconule. The distolingual cingulum is poorly developed, but thicker than the buccal and mesial ones. On the skull specimen MM-78, the left M1 still preserves the buccal region. The metacone is shorter than the paracone and its buccal shelf is less developed than the buccal shelf of the paracone.

Comparisons and discussion (tables 3–5): Vishnuonyx was represented by two species so far: V. chinjiensis Pilgrim, 1932, and V. angololensis Werdelin, 2003. The type species Vishnuonyx chinjiensis was originally described from a left maxilla with P4 and a fragment of the root of M1 from the middle Miocene of the Chinji Formation, in the Potwar Plateau of northern Pakistan (Pilgrim, 1932). Another P4 attributed to this species is reported from the middle Miocene of the Ngorora Formation, Kenya (Morales and Pickford, 2005). We compared MM-36 with the holotype of V. chinjiensis (GSI D 223, Pilgrim 1932: pl. II, figs. 17–17A; Pickford, 2007: fig. 3A; only the P4 figured) and to a better preserved and yet undescribed P4 from the Potwar Plateau of northern Pakistan displaying similar proportions, morphology, and age as the holotype (GSP-Y2108 from loc.Y53 dated to 12.756 Ma based on the timescale of Gradstein et al., 2012; here attributed to V. chinjiensis, fig. 4L–M). The P4 of V. maemohensis is different from that of V. chinjiensis in being longer relative to its breadth (table 5), in displaying a notch between the protocone and hypocone rather than a U-shaped valley, and in having a lingual cingulum thicker at the level of the metastyle than at the level of the protocone-hypocone compared with a similarly developed cingulum on those regions for V. chinjiensis. The Mae Moh upper carnassial also differs from the Kenyan P4 of V. chinjiensis (Morales and Pickford, 2005: 278, fig. 4A) in being longer and narrower and in exhibiting a deeper separation between the protocone and hypocone. In addition to the holotype, Pilgrim (1932) describes and figures a mandible fragment of V. chinjiensis found in the same beds near Chinji (GSI D 245, Pilgrim, 1932: pl. IV, figs. 6–6a). This mandible preserves only p4, a part of m1 talonid, and a fragment of m2 (ramus with p4 refigured by Pickford, 2007: 88, fig. 3). The p4 bears a strong distal accessory cuspid and the m1 talonid has a narrow basin with a distinct hypoconid, as in the Thai specimens. Although described as absent by Pilgrim (1932), a tiny mesial accessory cuspid merging with the cingulid of p4 is visible on GSI D 245, as in the Thai material. Moreover, Pilgrim (1932) described the m1 talonid of V. chinjiensis as wider than the trigonid based on this specimen. However, the m1 trigonid is lacking and only the mesial alveolus for the trigonid is visible, so that the relative proportions of m1 talonid and trigonid cannot be inferred. In V. maemohensis the m2 is shorter and wider than in GSI D 245, while the m1 and p4 have the same range of dimensions, although the p4 might have been slightly longer than in V. chinjiensis relative to its width (table 3). We also compare V. maemohensis specimens to another yet undescribed specimen from the Chinji Formation of the Potwar Plateau displaying similar proportions, morphology and age as GSI D 245 but with a completely preserved m1 (GSP-Y 40764 from loc. Y828 dated to 13.357 Ma based on the timescale of Gradstein et al., 2012; here attributed to V. chinjiensis, fig. 4I–K). It is very similar to V. maemohensis in morphology, but the m1 is narrower relative to its length and the depth of the mandible below m1 is greater relative to m1 length than in V. maemohensis. The type species of Vishnuonyx is also recorded in the middle Miocene of Ramnagar, India, by a mandibular fragment with p4 (Nanda and Sehgal, 1993). For the same width, the p4 of V. maemohensis is longer than in the Indian specimen (table 3).

Vishnuonyx angololensis is reported by a P4 from the late Miocene of the lower member of the Nawata Formation of Lothagam, in Kenya (Werdelin, 2003). MM-36 is closer to that specimen than to the material of V. chinjiensis by the presence of a notch between the protocone and hypocone. It has, however, smaller dimensions and is narrower relative to its length, and it bears a more distinct parastyle and a thinner lingual cingulum at the base of the lingual cusps compared with the P4 of V. angololensis.

Haile-Selassie (2008) attributed a mandible from the earliest Pliocene of the Haradaso Member, Sagantole Formation, Middle Awash (Ethiopia) to Vishnuonyx sp. He stated that this specimen could belong to V. angololensis based on the temporal and spatial proximity between Lothagam and Middle Awash. V. maemohensis differs from this material in displaying smaller proportions, a higher protoconid, a longer trigonid relative to m1 length, a wider lingual trigonid valley, and in having a shallower mandibular ramus.

We also compare V. maemohensis to some Torolutra specimens, as it has been suggested that V. angololensis might belong to this genus (Haile-Selassie, 2008). Torolutra ougandensis was first reported from mandibular fragments and a lower canine from the lower Pliocene of Nyaburogo, Nkondo and Warwire Formations, Uganda (Petter et al., 1991). Compared with this material, the lower teeth of V. maemohensis are shorter and narrower, the cingulid is thinner, especially on the buccal side of p4, and the m1 has a narrower paraconid, a wider lingual valley of the trigonid, and a narrower and deeper talonid basin. Additional material of Torolutra ougandensis, including a P4, is described from the late Miocene Kapcheberek locality, Lukeino Formation of the Tugen Hills, Kenya (Morales et al., 2005). MM-36 differs from it in being shorter and narrower relative to its length, and in having a notch rather than a valley separating the protocone and hypocone, a thicker lingual cingulum at the region of the metastyle and thinner cingulum surrounding the lingual cusps. Specimens referred to T. ougandensis by Haile-Selassie (2008) from the latest Miocene of Adu-Asa Formation and earliest Pliocene of Sagantole Formation (Asa Koma and Haradaso members, respectively) of the Middle Awash, Ethiopia, bring additional morphological information. The lower canine of V. maemohensis is less robust at its base, and its crown is less convex in its proximal half compared with GAW-VP-3/25 (Haile-Selassie 2008: 560, fig. 3G). The maxillary fragment GAW-VP-1/3 bears an M1, which is described but not figured by the authors: “A strong shelf-like cingulum is present on the buccal side of the paracone and metacone, larger at the former. The crown also has a deep, centrally positioned median valley. The paracone and metacone are of equal height and separated by a V-shaped valley” (Haile-Selassie, 2008: 562). This description, relatively common for otter species, does not allow comparisons with the material of Mae Moh. However, the M1 of V. maemohensis is shorter and wider than this specimen. Moreover, the authors also figured and described an M1 from the Asa Koma Member of the Adu-Asa Formation of the Middle Awash assigned to Lutrinae indet. aff. Torolutra sp. (ASK-VP-3/411). MM-37 differs from it in being wider relative to its length, in having a paraconule (absent in ASK-VP-3/411, according to Haile-Selassie, 2008, maybe because of wear), in having higher and sharper cusps, a protocone more distant to the lingual border of the tooth, a thinner cingulum at the mesiobuccal and distolingual borders of the tooth, and a narrower lingual root.

The difference in morphology and measurements between the upper carnassial from Mae Moh and those of V. chinjiensis and V. angololensis are important enough to erect a new species. Vishnuonyx maemohensis is the oldest and could be the most primitive species of the genus (e.g., high protoconid, narrow talonid basin on m1, narrower P4 with distinct parastyle). The Thai specimens described here highlight the closer morphological resemblance between Vishnuonyx and the African Mio-Pliocene and early Pleistocene Torolutra (see Werdelin and Peigné, 2010, for a list of localities; Werdelin and Lewis, 2013) than to the African and Eurasian Mio-Pleistocene Sivaonyx-Enhydriodon. The primitive morphology of V. maemohensis and the oldest presence of Vishnuonyx in the African continent at ca. 12 Ma (Morales and Pickford, 2005) could confirm a dispersal event of the genus from Asia to Africa during the middle Miocene, probably between ca. 12 and 14 Ma (Grohé et al., 2012).

Holotype: TF 6210, fragment of left hemimandible with m1, and isolated left fragmentary canine, p4, m2 lacking its distalmost part, and right m2. This material belongs to a single, young adult individual and is stored at the DMR, Bangkok.

Referred material: MM-38, left m1 and MM-39, right m2; both stored at the DMR, Bangkok.

Diagnosis (from Peigné et al., 2006: 529):

Mid-sized amphicyonine differing from the typical species of Amphicyon (mainly European and North American species) by teeth that do not display the diagnostic ‘swollen inflated appearance’ (Hunt, 2003, p.82) and by a less reduced p4 relatively to m1 (compared with middle Miocene Amphicyon) and an enlarged m2 relatively to m1 (compared to early Miocene Amphicyon). Compared to Cynelos: m1 talonid with a less developed entoconid and a less pronounced basin; much enlarged (relatively to m1), less rectangular m2 with a narrower talonid that tapers distally and a more prominent hypoconid. Compared to Ysengrinia: much enlarged m2 relatively to m1, m2 broader than m1 (except in Y. depereti). Compared to Pseudocyon: less reduced p4 and enlarged m2 (both relatively to m1), m2 with trigonid cusps not mesially shifted (trigonid longer than talonid) and with a prominent hypoconid crest. Compared to Pliocyon: enlarged m2 relatively to m1. Compared to Ischyrocyon (especially late Miocene specimens): narrower and less reduced (relatively to m1) p4; distal accessory cusp present on p4, m1 retaining a metaconid and having a buccally placed hypoconid.

Description (fig. 4O–Q, S–U): The trigonid of m1 is longer and slightly narrower than the talonid. The paraconid is very mesial, short and narrow. Its tip is nearly aligned with that of the protoconid. The metaconid is approximately as high as the paraconid and lower than the protoconid. It is partly fused to the lingual face of the protoconid and displaced distally relative to the tip of the protoconid. The mesial crest of the metaconid is very short. The distal face of the protoconid and metaconid is oblique. The talonid bears a high and slightly distobuccally oriented cristid obliqua. This crest starts from the protoconid base to join a strong hypoconid at two-thirds of the talonid length. The hypoconid is slightly lower than the metaconid and paraconid. The distal crest of the hypoconid is reduced in height abruptly distolingually to reach half the hypoconid height. It continues into an entoconid crest that gently curves along the distal and lingual borders of the talonid to reach the base of the metaconid distal face. The entoconid crest is crenulated distally. The talonid is unbasined. The lingual face of the hypoconid and cristid obliqua is oblique and weakly concave whereas the lingualmost part of the talonid is roughly horizontal. The cingulid is poorly visible. It is more developed on the buccal side of the talonid. The dorsalmost part of the buccal face of the paraconid, the mesialmost part of that of the protoconid, and the dorsodistal corner of the buccal face of the hypoconid show shearing facets.

The m2 is wider and shorter than m1. The m2 trigonid is wider than the talonid. It shows a very low vestigial paraconid from which two curved crests extend distally. Whereas the distobuccal crest joins the protoconid tip, the distolingual one interrupts its course on the metaconid mesial face, just below its tip. The protoconid is the highest trigonid cuspid. Compared with the m1 metaconid, the m2 metaconid is larger (clearly separated from the protoconid), higher (but still lower than the protoconid) and less distal relative to the protoconid. Protoconid and metaconid are linked by a short V-shaped crest. A distal crest extends from the metaconid tip to below the middle of its distal face. The talonid is lower than the trigonid. The cristid obliqua is oriented slightly distolingually and joins a distal crest going down from the protoconid tip at an approximately right angle. The hypoconid is strong, approximately as high as in m1. Compared with m1, the distal crest of the hypoconid becomes lower less abruptly and the entoconid crest extends only until it reaches the lingual half of the talonid. The m2 talonid is unbasined and the region lingual to the hypoconid is more concave than in m1. The cingulid is developed buccally.

Comparisons and discussion (fig. 4N–U, table 6): The lower carnassial has been recovered from the same lignite zone (K1) as the holotype of Maemohcyon potisati (Peigné et al., 2006) and is of similar size and morphology. The m2 comes from the Q lignite zone, which is only slightly older (Benammi et al., 2002; Coster et al., 2010; fig. 1). We consider the morphological differences between the holotype and these specimens (i.e., less tapered m1 and m2 talonids, crenulated and curved m1 entoconid crest, more oblique m1 cristid obliqua in the new material) as part of the intraspecific variation recognized within amphicyonid species (e.g., Amphicyon giganteus in Ginsburg and Telles Antunes, 1968). This new assessment of the morphological variation in Maemohcyon implies that m2 shape is not a useful character differentiating Maemohcyon from Cynelos (cf. the differential diagnosis of Peigné et al., 2006). In addition, contrary to the initial comparisons of Peigné et al. (2006), the distal transverse reduction of the m2 talonid can no longer be used to differentiate Maemohcyon potisati from several Amphicyon species: A. palaeindicus from the middle and late Miocene of the Potwar Plateau, northern Pakistan (Lydekker, 1876, 1884; Matthew, 1929; Pilgrim, 1932; Colbert, 1935), the middle Miocene Dang Valley, western Nepal (West et al., 1978), and the late Miocene of Yuanmou, southern China (Qi, 2006), A. shahbazi from the Oligo-Miocene deposits of the Bugti Hills and the early Miocene Murree Formation in Pakistan (Pilgrim, 1912; Matthew, 1929; Pilgrim, 1932; Ginsburg and Welcomme, 2002), A. sindiensis from the middle Miocene basal beds of the Manchar Formation, Sindh Province of southern Pakistan (Lydekker, 1884; Pilgrim, 1932; Raza et al., 1984), and A. ulungurensis from the Halamagai Formation of the Junggar Basin, Xinjiang Autonomous Region of China (Qi, 1989; Jiangzuo et al., 2018).

Since the first description of Maemohcyon by Peigné et al. (2006), other amphicyonines have been described and are compared here. Specimens of Amphicyon sp. are reported from the late Miocene Lower and Upper Irrawaddy formations of Myanmar (Takai et al., 2006; Sein and Thein, 2011). Among these specimens, the material from Tebingan, Magway region of central Myanmar, is a right mandible bearing m1–m2 (Sein and Thein, 2011). Compared with this specimen, Maemohcyon potisati exhibits smaller dental dimensions and a more vertical mesial face of the m1 paraconid. Another Amphicyon species, A. zhanxiangi, has been recently described from the lower part of the Zhang'enbao Formation, in Yehuliquanzigou, and from an unknown locality in Tongxin County, Ningxia Autonomous Region of northern China (Jiangzuo et al., 2019). The material belongs to the late Shanwangian Dingjiaergou local fauna (late early Miocene to early middle Miocene). It comprises an m1 trigonid and a partial maxilla with P4–M3. The morphology of the m1 fragment of A. zhanxiangi is rather common for an amphicyonine and does not allow relevant comparisons with Maemohcyon. The Chinese species is, however, a larger form, with a width across the protoconid equal to 17.12 mm, according to Jiangzuo et al. (2019), while the m1 width of Maemohcyon is 12.3 mm for the holotype and 13.4 mm for the new specimen described here. Moreover, Jiangzuo et al. (2018) described several species of amphicyonids from the middle Miocene Halamagai Formation, northwestern China, including lower molars of the amphicyonines Cynelos cf. bohemicus and Amphicyon cf. ulungurensis. Peigné et al. (2006) already noted relevant differences between Maemohcyon and Cynelos genera, and between Maemohcyon potisati and A. ulungurensis. Lastly, Ysengrinia sp. has been reported from the early Miocene Shanwang Basin, Shandong Province of eastern China (Qiu and Qiu, 2013). Previously described as a new unnamed genus of Thaumastocyoninae (Qiu et al., 1986: 24, figs. 1–2), the Chinese form, represented by isolated lower m1 and m2 and a maxilla, has been reassigned to Ysengrinia. On both lower molars, only the m1 is illustrated (Qiu et al., 1986: 24, fig. 2) and shows a less vertical mesial face of the paraconid, a more centrally placed protoconid, a higher metaconid, and a shorter talonid as compared to Maemohcyon. The description of the m2 does not allow more useful comparisons with Maemohcyon. Maemohcyon has a longer m1 and shorter m2 than this Chinese form.

Peigné et al. (2006) provide more comparisons with North American, European and African amphicyonines (i.e., Cynelos, Amphicyon, Ysengrinia, Pseudocyon, Ictiocyon, Pseudarctos, Pliocyon, Ischyrocyon, Afrocyon and Myacyon), although they were not able to suggest a contemporaneous close relative to Maemohcyon. Due to the poor state of preservation and fragmentary nature of Asian amphicyonid fossils (notably in China and Pakistan), most of the remains are attributed by default to the genus Amphicyon. Peigné et al. (2006) distinguished the Mae Moh amphicyonid from Amphicyon lineages notably because of the nearly vertical buccal faces of the lower teeth (see Peigné et al., 2006 for more comparative details).

Material: MM-40, right M1 lacking its buccalmost part, MM-77, right M2; both stored at PALEVOPRIM, Poitiers (France).

Description (fig. 4V–W, table 6): Only the lingual part of the right M1 is preserved, bearing the protocone, the pre- and postprotocristae and the lingual cingulum. The mesial and distal borders are roughly parallel; only the buccalmost part of the distal border is gently concave. The preprotocrista is higher than the postprotocrista. The protocone is located slightly mesially relative to the mesiodistal axis of the tooth. The cingulum surrounds the lingual border symmetrically. A broad, flat, and oblique valley separates the protocone region from the lingual cingulum.

The M2 has a shorter lingual border relative to the buccal one and its distal border is concave. This tooth shows a higher and wider paracone than the metacone. The crest connecting the paracone and metacone creates a shallow but wide U-shaped valley. The buccal cingulum bears a small parastyle at the mesiobuccal base of the paracone. Between the distobuccal face of the paracone and the mesiobuccal face of the metacone, there is a small shelf separating both cusps from the buccal cingulum, where a metastyle is located. A short crest extends down from the metacone tip distolingually, then slightly mesiolingually, where the tiny metaconule rises. The protocone is situated in the midregion of the tooth, mesiodistally, and is closer to the lingual cingulum than to the paracone and metacone. The metaconule and protocone are not connected. The postprotocrista is short and extends only to a very low and small cusp just distobuccal to the protocone. The preprotocrista is long and splits into two crests at the base of the paracone: one crest reaches the middle of the paracone lingual face, the other one, slightly crenulated, extends around the paracone to the middle of its mesial base. There is no paraconule. The lingual cingulum is thicker than the buccal one and extends from the level of the metaconule to the middle of the mesiobuccal crest extending from the preprotocrista. Its thickest part is distolingual to the protocone and its edge is slightly crenulated.

Comparisons (table 6): These upper molars correspond in morphology and size to Maemohcyon potisati. We here compare this material to the remains of Miocene Asian amphicyonines.

The only upper dentition material of Miocene amphicyonines recovered in Southeast Asia belongs to Amphicyon sp., from the late Miocene Upper Irrawaddy Formation of Myanmar, near Magway (Takai et al., 2006). MM-40 differs from the M1 of the maxilla fragment of this species by having a wider valley between the protocone and the lingual cingulum, and a postprotocrista that is not connected to the lingual cingulum. MM-77 differs from the M2 in being shorter lingually, and in having a more-developed metacone, a less-marked postprotocrista, and a thicker lingual cingulum. The upper molars from Mae Moh are also smaller than in the Magway specimen as estimated from the figure (Takai et al., 2006: fig. 7).

Several species from the Siwaliks have been created based on isolated upper teeth. This is the case of Amphicyon palaeindicus, which was first reported from Kushalghar, Punjab Province of northern Pakistan from a single M2 (Lydekker, 1876). The holotype (GSI D 26) is an M2, illustrations of which appearing divergent according to publications (figured by Lydekker, 1876: pl. VII, fig. 5; Lydekker, 1884: pl. XXXII, fig. 8; Matthew, 1929: 482, fig. 19 right). Compared with the holotype cast BMNH M 1558, we observed that MM-77 has a smaller breadth relative to its length, and a metastyle is present. Another M2 is reported from the late Miocene of Yuanmou, in the Yunnan Province of China (Qi, 2006). However, this tooth (PDYV1090, Qi, 2006: 149, figs. 3.16.1a–c) does not resemble the holotype of A. palaeindicus, and could belong to a different species of amphicyonid. Compared with this specimen, MM-77 is more asymmetrical, with a less-developed metacone and postprotocrista, and the tooth bears a metastyle on the buccal cingulum, unlike the Yunnan M2. MM-77 is much smaller than these two specimens from China and Pakistan.

Amphicyon pithecophilus is known from the middle Miocene Chinji Formation of northern Pakistan and from an indeterminate stratigraphic level in the locality of Nurpur, Himachal Pradesh of India (Lydekker, 1884; Pilgrim, 1932). The M2 of ?Maemohcyon potisati differs from the M2 of A. pithecophilus (holotype GSI D 129, Pilgrim, 1932: pl. II, figs. 3–3a) in being much shorter lingually relative to the buccal side, in having a poorly marked postprotocrista, and in lacking a ridge extending down buccally from the protocone tip. MM-40 differs from the M1 of A. cf. pithecophilus (GSI D 155, Pilgrim, 1932: pl. II, figs. 2–2a) in having a rounded, not triangular, lingual outline, a more internal and less mesially placed protocone, and uncrenulated pre- and postprotocristae. The upper molars from Mae Moh also have smaller proportions compared with those of A. pithecophilus.

Arctamphicyon lydekkeri is reported from the late Miocene Dhok Pathan Formation near Padhri, in the Punjab Province of northern Pakistan (Pilgrim, 1910). The holotype (GSI D 133) is an M2 according to Matthew (1929: 482, fig. 18 left). Pilgrim (1932) identified this tooth as an M1 (Pilgrim, 1932: pl. II, fig. 7) because he described an additional M2 (GSI D 134, paratype, Pilgrim, 1932: pl. II, figs. 8–8a) which probably came from the same spot, and showed “identical colour and state of fossilisation” but a different wear stage (Pilgrim, 1932: 25). For Pilgrim, this could indicate a more distal position of this tooth through the dental row as compared to the holotype. However, Peigné et al. (2006) argued that the differences in dimensions and morphology between these specimens were not sufficient to justify the attribution of the material to a first and second molar, and in consequence, they both are M2s. MM-77 differs from the holotype and the paratype of Arctamphicyon lydekkeri by being smaller, lingually narrower, and by having a thicker lingual cingulum, a lower paracone and a metastyle on the buccal cingulum.

Amphicyon shahbazi was initially described from lower dentition from the Oligocene-Miocene deposits of the Bugti Hills, Balochistan Province of Pakistan (Pilgrim, 1912). Later, Pilgrim (1932) described two additional upper dentition specimens from the Bugti Hills, including an upper molar. MM-40 differs from the M1 of A. cf. shahbazi (GSI D 113, Pilgrim, 1932: pl. II, figs. 6–6a) by being smaller and by showing a more symmetrical morphology and a longer valley between the protocone and the lingual cingulum. Ginsburg and Welcomme (2002) also reported an M2 of A. shahbazi (PAK 2424) from the same region. MM-77 differs from it by having a metastyle on the buccal cingulum and a shorter postprotocrista, and by being smaller and shorter relative to its breadth.

Upper molars of Amphicyoninae from the Miocene of Kazakhstan (Bonis et al., 1997; Kordikova, 2001), Japan (Kohno, 1997) and China (Qiu and Qiu, 2013; Jiangzuo et al., 2018, 2019) can also be compared here. Askazansoria mavrini from the early Miocene of Askazansor Formation, Betpakdala Steppes of southern Kazakhstan, initially attributed to Ysengrinia (Bonis et al., 1997), is represented by a maxilla bearing P4–M2. MM-40 differs from the M1 by bearing more mesially placed protocone and preprotocrista. MM-77 differs from the M2 by being larger and by having a less-reduced metacone and a different tooth outline, with a narrower and more-rounded lingual border. Ysengrinia sp. is recorded by an M1 from the early middle Miocene Korematsu Formation of Shobara, Hiroshima Prefecture of southwestern Japan (Kohno, 1997). The M1 of ?Maemohcyon potisati differs from that specimen in having a more symmetrical lingual cingulum. As mentioned before, Ysengrinia sp. also occurs in the early Miocene Shanwang Basin, Shandong Province of eastern China (Qiu and Qiu, 2013). On the maxilla bearing P4–M3, the M1 lingual part of this Chinese form is not sufficiently well preserved to allow comparisons with MM-40. However, MM-77 differs from the M2 in having a narrower lingual border relative to the buccal one, a longer preprotocrista subdivided into two crests when approaching the paracone, a postprotocrista developed distolingually and not just distally, and in the presence of a metastyle on the buccal cingulum. Jiangzuo et al. (2018) described several species of amphicyonids from the middle Miocene Halamagai Formation of the Junggar Basin, northwestern China, including upper molars of Cynelos aff. helbingi. The upper molars of Maemohcyon are larger than those of this Chinese form, and their protocones are more centrally placed. Compared with the M2 of Cynelos aff. helbingi, MM-77 is longer relative to its width, it has a more prominent paracone, a metastyle on the buccal cingulum, a shorter postprotocrista, and a lingual cingulum bending distally. Lastly, Amphicyon zhanxiangi, from the early-middle Miocene of Tongxin County, northern China (Jiangzuo et al., 2019), is represented by two specimens including a maxillary fragment with P4–M3, chosen as the holotype of the species. MM-40 differs from the M1 of A. zhanxiangi in being more symmetrical. MM-77 differs from the M2 of this species in being triangular rather than rectangular in shape, with a lingual cingulum thickening distally, a prominent paracone, a more-elongated preprotocrista, a poorly distinct postprotocrista and a metaconule. The tooth from Mae Moh is also smaller and shorter relative to its width.

Material: MM-41, left M1; stored at PALEVOPRIM, Poitiers (France).

Description (fig. 4X): MM-41 is relatively symmetric and slightly compressed mesiodistally relative to its breadth, recalling M1 rather than M2 amphicyonid morphology. It is subtriangular and wider than long (L = 8.3 mm, W = 11.9 mm). The lingual root and one of the buccal ones are partially preserved. The lingual root is larger than the buccal one. The paracone and metacone are crestiform and relatively low. The paracone is clearly higher and slightly longer than the metacone. The crest joining these two buccal cusps is marked by a shallow V-shaped notch. The buccal rim displays a developed constriction between the paracone and metacone. The protocone and the pre- and postprotocristae are worn. The protocone is located slightly mesially with respect to the mesiodistal axis of the tooth, the preprotocrista being shorter than the postprotocrista. No conules are visible. The protocone is separated from the lingual cingulum by a narrow valley. Both the protocone and the lingual cingulum are very low. The cingulum is also slightly marked buccally.

Comparisons (fig. 4W): This specimen belongs to a very small amphicyonid species. In the Miocene of Asia, undescribed specimens listed as cf. Amphicyon sp. from the middle Miocene Ban San Klang Basin of northern Thailand and from the early Miocene Basal Manchar Formation (locality 2) at Bhagothoro, near Sehwan, Sindh Province of southern Pakistan (Raza et al., 1984; Ducrocq et al., 1995) display the closest similarity in morphology and size to MM-41 (Grohé, personal obs.). The assignment to the wastebasket genus Amphicyon is, however, erroneous. The small size and morphology of these specimens rather recall the Eurasian monospecific genera Ictiocyon and Pseudarctos (Ginsburg, 1992).

Ictiocyon socialis is known from the early Miocene of Europe (MN3–MN4a). In Asia, the species is reported from a mandible coming from the early Miocene beds of the Xianshuihe Formation of Lanzhou Basin, Gansu Province of China (Wang et al., 2005). The larger tooth proportions (M1 L = 9.6 mm, W = 14 mm, MNHN Ar 2554 in Ginsburg, 1992) and the squarer morphology of M1 in I. socialis (Ginsburg, 1992) differ notably from the Mae Moh specimen.

Pseudarctos was initially reported in the form of the species P. bavaricus from the early and middle Miocene of Europe (MN4–MN9). According to Ginsburg (1992), P. bavaricus represents an anagenetic lineage composed of three subspecies, among which size progressively increases through time: P. bavaricus beaucensis from the early Miocene (MN4b), P. bavaricus pontignensis from the early and middle Miocene (MN5), and P. bavaricus bavaricus from the middle and late Miocene (MN6–MN9). This last subspecies is also described from the late Miocene of Yuanmou, Yunnan Province of China (Qi, 2006), but upper molars are only known from Europe. According to the diagnosis of P. bavaricus bavaricus provided by Ginsburg (1992), the M1 buccal cusps are low, which could correspond to the morphology observed in MM-41. The subtriangular outline of the Mae Moh specimen associated with a marked constriction between the paracone and metacone are different from the specimens of P. bavaricus pontignensis and reminiscent of those of P. bavaricus beaucensis figured by Ginsburg (1992). A second occurrence of the genus is reported from the middle Miocene upper bed of Aletexire locality of Tunggur Basin, Inner Mongolia (China) with Pseudarctos sp. (Wang et al., 2003). This material is not figured and cannot be compared here. Until additional specimens are found, we attribute the single upper tooth of Mae Moh to cf. Pseudarctos sp.

Holotype: MM-45, right mandible with p2–m2 and alveolus of p1; stored at the DMR, Bangkok.

Referred material: MM-43, left mandible with p3–m2 and alveolus of p2; MM-44, left mandible with p2–m1 and alveolus of m2; stored at the DMR, Bangkok.

Etymology: Generic name derived from Siam, referring to Thailand, and ictis, meaning “marten” in Greek. Species name derived from carbo, meaning “coal” in Latin, in reference to the open coal mine of Mae Moh.

Diagnosis: Small Paradoxurinae with low-crowned teeth; diastema between c–p1 and p1–p2; single-rooted p1; developed mesial and distal accessory cuspids on p2–p4; vertical central pillar on the lingual side of the main cuspid preceded and followed by a depression on p2–p4; from p2 to p4, increase of: the separation of the accessory cuspids from the main cuspid, the distal development of the talonid, and the size of the lingual vertical central pillar, becoming an incipient metaconid on p4; p4 with obliquely aligned accessory cuspids and main cuspid, distolingual basined shelf with crenulated border; m1 protoconid higher than paraconid, metaconid smaller and distally placed relative to protoconid, high, equally developed entoconid and hypoconid, hypoconid slightly mesial or directly buccal to entoconid, low distal rim of the talonid with multiple small cuspids, basined talonid; small single-rooted m2, protoconid very buccal, poorly differentiated lingual and distal rims.

Differential diagnosis: Differs from Mioparadoxurus, Kichechia, Orangictis, and Pseudocivetta by higher accessory cuspids and molarization of premolars (especially on p4 where an incipient metaconid is present); from Ketketictis, Kichechia, Orangictis, Tugenictis, and Kanuites by a cristid obliqua parallel to the axis of m1; from Mioparadoxurus, Kichechia, Orangictis, Tugenictis, and Pseudocivetta by longer trigonid and less lingually positioned paraconid; from Mioparadoxurus, Kichechia, Orangictis, and Tugenictis by m2 much shorter relative to m1; from Kichechia and Orangictis by a lower m2 trigonid and m2 metaconid placed distally relative to the protoconid; from Mioparadoxurus and Pseudocivetta by a narrower m1 talonid; from Kichechia and Orangictis by the lack of a cuspid between the metaconid and entoconid; from Orangictis by a shallower mandible relative to m1 length; from Kanuites by a lower m1 protoconid.

Description (fig. 6A–C, table 7): MM-45 is the best-preserved specimen. The mandible is long and thin. It widens and becomes deeper from the level of p1 alveolus to the level of p4–m1. Behind the m2 and rostrally to the broken angular process, the ventral part of the mandible is concave. The masseteric fossa is relatively deep and its rostral edge reaches below the m2 distal end. The ascending ramus does not rise abruptly distal to m2. Two mental foramina are present on the buccal side: one beneath p1–p2 and another one beneath p3. The lingual mandibular foramen is located far from the dental row (8.8 mm behind). At the rostral end, the symphysis extends to the level of p2. The position of the p1 alveolus reveals two diastemata, between c–p1 and p1–p2. The premolars are asymmetrical. They have mesial and distal accessory cuspids, the positions and proportions of which vary from p2 to p4: the mesial accessory cuspid moves away from the main cuspid by becoming longer and wider, and is shifted lingually to the main cuspid, resulting in a distobuccal alignment of the accessory and main cuspids on p4; the distal accessory cuspids also become lower, further separated from the main cuspid, and displaced buccally relative to the main cuspid tip; the distal cingulid broadens distally and lingually to create a small basin with crenulated borders on p3. This basin is even more developed on p4. A vertical central pillar runs along the lingual face of the main cuspid on p2 and p3. It is preceded mesially by a small depression and distally joins the lingual border of the distal basined shelf. This pillar is more developed on p4 and becomes an incipient metaconid. Together with the main cuspid and the mesial accessory cuspid, it forms a small trigonid. The main cuspids of p2–p4 are approximately as high as the m1 paraconid. The m1 trigonid is rather low-crowned. It is slightly longer and wider than the talonid. The paraconid tip is as lingually located as the metaconid tip. The protoconid is higher than the paraconid. The metaconid is lower than the paraconid (unworn metaconid on MM-44) and is displaced distally with respect to the protoconid, resulting in a lingually opened trigonid. The talonid is well-basined and bears high and equally developed entoconid and hypoconid. Depending on the specimens, the hypoconid is shifted mesially with respect to the entoconid (in MM-45) or is directly buccal to it (in MM-43). Between the two cuspids, multiple small cuspids shape the distal rim of the talonid. They are half the height of the entoconid and hypoconid. The m2 is single-rooted and rounded. It bears a reduced paraconid, a high protoconid, and a lower hypoconid. The protoconid is displaced buccally relative to the hypoconid. A lower crest without differentiated cuspids surrounds the lingual and distal rims of the tooth. Cingulids are mainly present at the base of the premolars and at the buccal base of the m1 paraconid.

Comparisons and discussion (table 7): Among viverrids, the structure of the lower carnassial (i.e., wide talonid, individualized entoconid and hypoconid) is reminiscent of the hypo- and mesocarnivorous subfamilies Hemigalinae, Paradoxurinae, and Viverrinae. Compared with extant Viverrinae genera (Civettictis, Viverra and Viverricula), Siamictis, n. gen., has a narrower and shorter protoconid, resulting in similar proportions of the trigonid and talonid. We thus compared Siamictis with non-hypercarnivorous viverrid taxa morphologically close to extant hemigalines and to fossil paradoxurines (Morales and Pickford, 2011; Adrian et al., 2018).

In Asia, two species show morphological similarities with extant hemigalines and paradoxurines. Lufengictis peii from the late Miocene of Lufeng, Yunnan Province of China (Qi, 2004), is very similar to the extant Cynogale bennettii (otter civet). Siamictis differs from Lufengictis by having much smaller dimensions, a narrower m1, a longer m1 paraconid, m1 trigonid cuspids higher than the talonid ones (as opposed to similarly low trigonid and talonid in Lufengictis), and a much shorter m2 relative to m1. Compared with Mioparadoxurus meini from the late Miocene of Haritalyangar, northern India (Morales and Pickford, 2011), Siamictis displays much smaller dimensions, much higher accessory cuspids on p4 that is also molarized, a m1 protoconid higher than the p4 main cuspid, a narrower m1 with a longer paraconid, and a talonid as wide as the trigonid (wider than the trigonid in Mioparadoxurus).

Several other Miocene viverrid genera with hypo- or mesocarnivorous forms were originally described from the Siwalik Hills of Pakistan and India: Vishnuictis, Viverra, and an unnamed large viverrid genus (Lydekker, 1884; Pilgrim, 1932; Pilbeam et al., 1979). Three species of Vishnuictis are reported from the Miocene-Pliocene of the Indian Subcontinent and China. The type species, V. durandi, is represented by two skulls (Lydekker, 1884; Pilgrim, 1932) from uncertain horizons of the Siwaliks, which cannot be compared with Siamictis except for their size, which is larger. V. salmontanus has been recovered from the late Miocene of the Dhok Pathan Formation of Pakistan and from Yuanmou, Yunnan Province of China (Pilgrim, 1932; Qi et al., 2006; Dong and Qi, 2013). According to the description of Pilgrim (1932: 103), the holotype GSI D 160 shows diastemata between c–p1 and p1–p2, the m1 trigonid is longer than the talonid, the m1 paraconid points lingually, and the m1 metaconid is smaller than the protoconid, as in Siamictis. The m2 outline (Pilgrim, 1932: pl. IV, fig. 8a) is, moreover, similar to that in Siamictis. However, in V. salmontanus, tooth proportions are larger, the m1 protoconid is broader, and the m2 bears two distolingual distinct cuspids. Vishnuictis hariensis, known from the Indian late Miocene locality of Haritalyangar (Prasad, 1968), does not show any diastema between p1 and p2, unlike Siamictis. A fourth species of this genus, V. africana, is recorded in Africa, in the middle Miocene Muruyur Formation of Kipsaraman, Baringo District, Kenya (Morales and Pickford, 2008). Siamictis differs from V. africana in its smaller dimensions, the presence of a metaconid ridge on p4, and the lack of an internal cuspid on the m1 talonid basin.

Viverra chinjiensis has been recovered from the middle Miocene of the Chinji Formation (Pilgrim, 1932) and the early late Miocene of the Nagri Formation, northern Pakistan (Pilbeam et al., 1979), as well as from the middle Miocene of Ramnagar (Basu, 2004) and the late Miocene of Kalagarh (Tiwari, 1983; Patnaik, 2013), both in northern India, and from the middle Miocene of Chaungtha, Myanmar (Chavasseau et al., 2013). Morales and Pickford (2008) noted that this species should be provisionally referred to Vishnuictis. The only figured material of the species is the holotype, a mandible and an associated upper molar (GSI D 214, Pilgrim, 1932: pl. IV, figs. 9–9a and 10). Siamictis differs from V. chinjiensis in having smaller dimensions, higher and longer mesial accessory cuspids, a metaconid ridge on premolars (especially on p4), a cristid obliqua parallel to the axis of the tooth rather than distobuccally oriented, m2 cuspids lower than the m1 talonid ones, a more mesially placed protoconid on m2, and a shorter m2 relative to m1.

Viverra nagrii is described by Prasad (1968) on the basis of mandible fragments from the late Miocene of Haritalyangar. Siamictis differs from the holotype of V. nagrii (GSI 18076, Prasad, 1968: pl. IV, fig. 5) by having much higher mesial accessory cuspids and more-developed distal ones.

Lutrinae gen. indet. hasnoti (= “large, probably new viverrid genus” according to Pilbeam et al., 1979) is known from the late Miocene of the Nagri and Dhok Pathan formations of northern Pakistan (Pilgrim, 1932; Pilbeam et al., 1979). Siamictis differs from the holotype of this species (GSI D 135, Pilgrim 1932: pl. II, figs. 18, 18a, 18b) in having much smaller dimensions, a p4 with a metaconid ridge, a much higher mesial accessory cuspid and a more buccally placed distal one so that the distal shelf is developed lingually to the distal accessory cuspid (and not both lingually and buccally to it as in the Pakistani viverrid), and an m1 with a broader talonid and a narrower protoconid.

Outside Asia, the meso- and hypocarnivorous fossil viverrids are mainly restricted to Africa, with six genera attributed to the paradoxurine subfamily: Ketketictis, Kichechia, Orangictis, Tugenictis, Pseudocivetta, and Kanuites (Morales and Pickford, 2011; Adrian et al., 2018; Werdelin, 2019). Ketketictis solida is represented by an isolated m1 discovered in the early Miocene site of Wadi Moghra (= Wadi Moghara), in Egypt (Morlo et al., 2007). Siamictis differs from Ketketictis by displaying smaller dimensions, a broader talonid (as wide as the trigonid), higher hypoconid and entoconid, and a lower distal rim of the talonid. Kichechia is represented by two species from the early Miocene of Rusinga and Mfwanganu Islands, Songhor, and the Kalodirr Member of the Lothidok Formation of West Turkana, Kenya and possibly Napak, Uganda (Savage, 1965; Werdelin and Peigné, 2010; Adrian et al., 2018). Siamictis and Kichechia have similar m1s length but the premolars of Siamictis are narrower relative to their lengths. Moreover, Siamictis differs from Kichechia by having well-developed accessory cuspids on premolars, a much higher mesial accessory cuspid and an incipient metaconid on p4, a m1 with a less lingually oriented paraconid, a more distal metaconid, no cuspid between the metaconid and entoconid, a cristid obliqua parallel to the axis of the tooth, a lower m2 trigonid with a metaconid placed distally relative to the protoconid (opposite condition in Kichechia), and a shorter m2 relative to m1. Orangictis gariepensis was initially recorded from the middle Miocene of Arrisdrift, southern Namibia (Morales et al., 2001), by several mandibles. Additional isolated lower and upper molars from the middle Miocene of Fort Ternan, western Kenya, are referred to Orangictis sp. (Werdelin, 2019). Siamictis differs from Orangictis by exhibiting smaller dimensions, a tendency to the molarization of premolars with an incipient metaconid on p4, a shallower mandible relative to m1 length, narrower premolars relative to their lengths, a longer m1 paraconid, lack of a small cuspid between the m1 entoconid and metaconid, a cristid obliqua parallel to the axis of the tooth on m1, a shorter m2 relative to m1, m2 cuspids lower than those of the m1 talonid, and an m2 metaconid placed distally relative to the protoconid (opposite to the condition in Orangictis). Tugenictis ngororaensis remains come from sediments of the middle Miocene Ngorora Formation exposed in western Kenya (Morales and Pickford, 2005). This species is known by only two isolated lower molars of much larger dimensions compared with Siamictis. Relative to Tugenictis, the m1 of Siamictis displays a longer paraconid, a more distal metaconid, much lower talonid cuspids, and a cristid obliqua parallel to the axis of the tooth. Even if the m2 of Tugenictis does not belong to the same individual as the m1, it was certainly longer relative to m1 as compared to Siamictis (calculated m2/m1 L = 0.7 from the measurements of Morales and Pickford (2005) compared with 0.4–0.5 in Siamictis). Pseudocivetta ingens has been recovered from the early Pleistocene of the Shungura and Koobi Fora formations, in the Omo-Turkana basin, from Melka Kunture in Ethiopia, and from Olduvai in Tanzania (Petter, 1967, 1973; Petter and Howell, 1977; Geraads et al., 2004; Werdelin and Lewis, 2013). It is a much larger form as compared to Siamictis. The m1 of Siamictis differs from those of Pseudocivetta (O 33-354, Petter and Howell, 1977: 284; L39-13, Werdelin and Lewis, 2013: 91, fig. 6.8) by having a higher and longer trigonid and more distal metaconid and entoconid. Moreover, Siamictis also differs in p4 morphology by displaying higher cuspids and an incipient metaconid (p4s of P. ingens: 758 FLK NNI, Petter, 1973: pl. IV, fig. 12; 768 FLK NNI, Petter and Howell, 1977: 284). Kanuites lewisae is recorded in middle Miocene deposits of Fort Ternan, Kenya (Dehghani and Werdelin, 2008; Werdelin, 2019). Siamictis differs from Kanuites by being slightly smaller and displaying an m1 with a less lingually placed paraconid, a lower protoconid, a cristid obliqua parallel to the axis of the tooth, and multiple small cuspids between the hypoconid and entoconid (instead of a single hypoconulid in Kanuites).

Due to the scarcity of material attributed to putative paradoxurines, their phylogenetic relationships and their geographical origin remain unclear. Morales and Pickford (2011) proposed hypotheses of phylogenetic relationships among fossil and extant paradoxurines, placing Pseudocivetta and Mioparadoxurus as the closest relatives to the extant Paradoxurus, while other cladistic analysis by Adrian et al. (2018) and Werdelin (2019) suggested different inter-generic relationships. Regarding viverrid diversity, Siamictis is a very small form with a dental morphology close to that of extant paradoxurines. Siamictis, however, exhibits a more primitive morphology, with a longer and more-opened trigonid on m1, and less bunodont teeth, and a more-derived morphology of its premolars, with notably high accessory cuspids and an incipient metaconid on p4.

Material: TF 2659, right mandible with p3 and m1, and alveoli of p4 and m2; stored at the DMR, Bangkok.

Description (fig. 6G–I): The mandible is partially crushed, and the middle of the horizontal ramus caved in. Its lingual part sits on the sediments. It is also broken at the level of the ascending ramus and mesially to p3. In buccal view, the masseteric fossa is relatively deep and no foramina are visible. The p3 is approximately as high as the m1 paraconid. It bears a very small mesial accessory cuspid. The distal accessory cuspid is weak, merging with the distinct distal crest, and extending down from the main cuspid to about half its height. The distalmost part of p3 is broken off buccally, but indicates that the premolar is wider distally than mesially. Behind p3, the presence of small alveoli suggests a double-rooted p4 (L = 6.8 mm; W = 3.4 mm, estimated measurements), with a mesial root that is smaller than the distal one, and a longer crown length than that of p3. The m1 (L = 10.5 mm; W = 5.3 mm) has a trigonid as wide as and longer than the talonid. The protoconid is higher than the paraconid. The metaconid is slightly lower than the paraconid and half as long as the protoconid. The metaconid tip is slightly distal to the level of the protoconid tip, creating an open trigonid. The talonid is weakly basined. The cristid obliqua is slightly distobuccally oriented. The hypoconid is the highest cuspid of the talonid. The entoconid is displaced slightly mesially relative to the hypoconid. A low hypoconulid is present, distobuccally to the entoconid, at the lingualmost part of the distal border of the talonid. The m2 was single-rooted and small (probably as long as the m1 talonid), as indicated by the alveolus. The mesial edge of m2 alveolus extends mesiobuccally.

Comparisons: This species has a rather wide m1 talonid bearing a well-developed entoconid, and a p4 with a distal accessory cuspid, recalling the morphology of viverrids from the subfamily Viverrinae. It is much larger than Siamictis and its dental morphology is very different (e.g., very small mesial accessory cuspid and much smaller distal accessory cuspid on p3, distobuccally oriented cristid obliqua, hypoconid higher than entoconid, and hypoconid displaced distally relative to entoconid). The size and dental morphology rather remind those of Viverra spp. (e.g., trigonid cuspids of different height, talonid approximately as wide as the trigonid, three distinct talonid cuspids). Among Asian Viverra species, the Thai specimen differs from V. chinjiensis Pilgrim, 1932, from the middle to late Miocene of the Siwaliks and Myanmar (Pilbeam et al., 1979; Tiwari, 1983; Basu, 2004; Patnaik, 2013; Chavasseau et al., 2013) by being much larger and having a less distobuccally oriented cristid obliqua, and by a narrower protoconid on m1. It differs from V. nagrii Prasad, 1968, from the late Miocene of India, by a reduced distal accessory cuspid on p4. It differs from V. bakerii Bose, 1879, possibly from the late Pliocene–early Pleistocene Pinjor Formation of the Siwaliks, by its smaller dimensions (Lydekker, 1884; Pilgrim, 1932). It differs from V. peii Qiu, 1980, from the late Pliocene–early Pleistocene of Zhoukoudian, near Beijing, by being much smaller and bearing a narrower protoconid, a broader and higher metaconid, and a less distobuccally oriented cristid obliqua on m1. Outside Asia, V. leakeyi Petter, 1963, is a giant form of civet recorded from the late Miocene to the early Pleistocene in several localities of northern, eastern, and southern Africa (e.g., Petter, 1963; Hendey, 1974; Petter and Howell, 1977; Geraads, 1997; Werdelin, 2003; Werdelin and Peigné, 2010). One of the best-illustrated m1s comes from the Shungura Formation, Omo Basin of Ethiopia, and is figured by Peigné et al. (2008). The Thai viverrine differs from that specimen of V. leakeyi by a narrower protoconid, a more mesial entoconid and the lack of an additional cuspid mesial to the entoconid. Viverra howelli Rook and Martinez-Navarro, 2004, is recorded from the late Miocene deposits of Bacinello-Cinigiano Basin, in the Grosseto Province of Italy, as well as from Sahabi, in Libya (Howell, 1987) and as well as Lothagam, in Ethiopia (Werdelin, 2003) according to Rook and Martinez-Navarro (2004). The Mae Moh viverrine differs from the holotype of V. howelli by a narrower protoconid, a shallower trigonid lingual valley and a more mesial and lower entoconid. Among European fossils, it differs from Viverra pepratxi (Depéret, 1890) from the early Pliocene of France by a larger metaconid, the lack of a cuspid at the base of the protoconid, and a more distal hypoconid as compared to the entoconid. In summary, the Mae Moh viverrine could belong to a new species of Viverra, but further determination would be premature.

Type species: Herpestes filholi Gaillard, 1899.

Other species included: L. aurelianensis (Schlosser, 1888), L. atavus Beaumont, 1973, L. rangwai Schmidt-Kittler, 1987, L. mbitensis Schmidt-Kittler, 1987, L. senutae Morales et al., 2008, L. namibiensis Morales et al., 2008, L. peignei, n. sp.

Diagnosis (from Werdelin and Peigné, 2010: 636):

Small-sized carnivoran; dental formula (lower dentition only) I 3, C 1, P 4, M 2; premolars with tall cusps; p4 posterior accessory cusp very large; m1 postvallid notch less deep than in Herpestes; m2 trigonid and talonid distinct.

Remarks: Major (1903) proposed the name Leptoplesictis for Herpestes filholi Gaillard, 1899, from La Grive Saint-Alban, Isère, France, because of the slenderness of the teeth relative to the extant genus Herpestes. Herpestes filholi is de facto the type species of the new genus. He also created another species, Leptoplesictis minor, based on a small size difference with L. filholi (the length of p4–m1 is “21.5 mm in L. filholi, against 18 mm in L. minor”; Major, 1903: 536). Leptoplesictis minor was later synonymized with L. filholi by Roth (1988).

Geographical and stratigraphical distribution: Early Miocene of Artenay, France (MN4; Ginsburg, 2002), Petersbuch 2, Germany (MN4; Roth, 1988), Rusinga Island, Kenya (17–15 Ma; Schmidt-Kittler, 1987), Grillental and Langental, Sperrgebiet, Namibia (20–19 Ma; Morales et al., 2018); Middle Miocene of Pontlevoy, France (MN5; Gervais, 1967–1969; Stehlin and Helbing, 1925), Vieux-Collonges, France (MN5; Beaumont, 1973), Sansan, France (MN6; Ginsburg, 1961; Roth, 1988; Peigné, 2013), Stein am Rhein, Germany (MN6; Roth, 1988), lower Hostalets de Pierola, Spain (MN7/8; Fraile et al., 1997; Morales et al., 1999), La Grive-Saint-Alban, France (MN7/8; Gaillard, 1899); late middle Miocene of Q–K lignite zones of Mae Moh Basin, Thailand (13.4–13.2 Ma; this study).

Holotype: MM-51, left mandible with p2–m1 and alveolus of p1; stored at the DMR, Bangkok.

Referred material: MM-46, left mandible (p3–m1 with p3 and m1 trigonid broken, alveoli of m2, p2 and p1), stored at the DMR, Bangkok; MM-47, left mandible (p4–m1, alveolus of m2), stored at PALEVOPRIM, Poitiers; MM-48, right mandible (p3–m1 with p4 and m1 trigonid horizontally broken, distal alveolus of p2 and alveolus of m2), stored at PALEVOPRIM, Poitiers; MM-49, right mandible (p3), stored at PALEVOPRIM, Poitiers; MM-50, right mandible (m1, alveolus of m2), stored at PALEVOPRIM, Poitiers; MM-52, right mandible (m1, alveoli of p4 and m2), stored at the DMR, Bangkok; MM-53, right mandible (m1, broken p4 and p3), stored at the DMR, Bangkok; MM-103, left mandible (broken p3, p4, mesial root of m1) and associated lower canine, stored at PALEVOPRIM, Poitiers.

Etymology: Species name in memory of Stéphane Peigné, who greatly contributed to carnivoran systematics and evolution.

Diagnosis: Medium-sized Leptoplesictis; m1 with open trigonid and basined tricusped talonid with a dominant hypoconid; single-rooted p1; p2 without accessory cuspids; p3 and p4 with a tiny and a small mesial accessory cuspid, respectively; p3 and p4 with a distal accessory cuspid, very large in the latter.

Differential diagnosis: Differs from L. filholi, L. aurelianensis, L. atavus, L. mbitensis, and L. rangwai by a larger size; from L. filholi by a lower mesial accessory cuspid on p3 and p4, the lack of a basal cingulid mesial to the mesial accessory cuspids, and a relatively shorter m2 (from the alveolus); from L. aurelianensis by a shorter m1 trigonid and a longer m2 (from the alveolus); from L. atavus by the presence of a distal accessory cuspid on p3, a higher distal accessory cuspid on p4, a more open m1 trigonid with metaconid tip bending backward, and lower entoconid and hypoconid; from L. rangwai by a lower mesial accessory cuspid on p4, a higher m1 metaconid, and a shorter m2 (from the alveolus); from L. mbitensis by the lack of a lingual cuspid on p4 talonid; from L. senutae by a shorter and narrower p2, and the presence of an hypoconulid on the m1 talonid; from L. namibiensis by a smaller size, and a lower-crowned m1 talonid.

Description (fig. 7, table 8): The mandible MM-46 preserves the longest ramus in the sample. The ventral border of the specimen is convex from the canine level to m2, and concave at the level of the ascending ramus. The angular process is broken off but its distal end is located at 17 mm from the rostralmost part of the masseteric fossa. In buccal view, the most distal part of the ramus shows a flat triangular surface, just ventral and caudal to a deep masseteric fossa. In lingual view, the angular process curves lingually to delineate the ventral outline of the medial pterygoid muscle insertion area. The ascending ramus is broken off dorsally, so that the coronoid process and articular condyle are not preserved and the masseteric fossa is not complete. Mental foramina are present under the p1–p2 contact and under p3 (under the midpart of p3 for MM-46 and MM-49 while under the mesial root of p3 for MM-48 and MM-51). The lingual mandibular foramen is located far from the dental row (approximately 8.5 mm behind the distal edge of the m2 alveolus). The mandibular symphysis extends to the distal root of p2. The p1 and m2 are single-rooted. Specimen MM-51 allows observation of more details on the premolars. The p2 is lower than the p3–p4, which have approximately the same height, similar to that of the m1 paraconid. The p2 shows a small distal unbasined shelf, and the main cuspid is displaced mesially relative to the long axis of the tooth. The p3 has a tiny mesial accessory cuspid at the mesial base of the main cuspid, a distal accessory cuspid slightly buccal to the main cuspid and merging with its distal face (accessory cuspid poorly developed in MM-51, more distinct in MM-49 and even more so in MM-48, although it remains small), and a small cingular shelf oriented distolingually and unbasined. This talonid is marked buccally by a crest and closed lingually by a cingulid. It appears higher at the distal end of the tooth than at the base of the main cuspid distal face. The p4 bears a small mesial accessory cuspid that is more developed than in p3 and bends lingually. It also displays a prominent distal accessory cuspid located midheight to the main cuspid and buccal to its tip, so that the three cuspids of p4 are obliquely aligned. A lingual cingular shelf extends on the distal part of this premolar. This talonid is larger than in p3 and the difference in height between the mesial and distal part of the talonid is even more marked. The m1 has a longer, wider, and higher trigonid than talonid. The protoconid is much higher than the paraconid. The metaconid is slightly lower and narrower than the paraconid in lingual view, and its tip is displaced distally relative to the protoconid tip. The mesial faces of the protoconid and metaconid curve backward. The distal crest of the paraconid and the mesial crest of the metaconid are not connected lingually and a wide and deep trigonid valley opens between these cuspids, well separated from each other. The talonid basin is very low and shallow. The hypoconid is the highest talonid cuspid. A hypoconulid and an entoconid are present on the distal and distolingual borders of the talonid, respectively. The entoconid is slightly distal relative to the hypoconid. The cristid obliqua is poorly oriented distobuccally and reaches mesially the middle of the protoconid distal face. The m2 alveolus is small and approaches the length of the m1 talonid. Cingulids are present at the mesiobuccal base of m1 and on the lingual side of the premolars, at the level of the accessory cuspids and distal shelves.

Comparisons and discussion (table 8): The high trigonid and narrow talonid of m1 in association with a relatively short m2 and accessory cuspids on p3–p4 resemble the morphologies observed in some extant Herpestidae (particularly Herpestes). The only herpestid of the Miocene Asian fossil record is the extant genus Herpestes. Barry (1983) reported some material from the late Miocene of the Potwar Plateau, in the Punjab Province of Pakistan. Barry (1983) recognized three species of Herpestes representing the oldest fossil record of the family in Asia. Only ?Herpestes large species is represented by lower dentition, a mandible with p4 (GSP 217, Barry, 1983: 153, fig. 3). This specimen is much larger than the Mae Moh material and shows a p4 morphology that is relatively typical for viverrids and herpestids (i.e., small mesial accessory cuspid, distobuccal accessory cuspid and distolingual extension of the cingulid into a shelf). Besides, our Thai form differs from Herpestes by the talonid borders of m1: the lingual border at the base of the metaconid is higher in the Thai material than in Herpestes specimens, so that the lingual and buccal borders of the talonid at the base of the trigonid are of equal height. Moreover, the m1s of our Thai form are also more slender than the m1s of Herpestes (index m1 W/m1 L = 0.43 to 0.52 for Leptoplesictis peignei against a mean ratio of 0.56 for Herpestes sanguineus and 0.57 for H. pulverulentus from measurements in Petter, 1987).

Very close to Herpestes in morphology, Leptoplesictis Major, 1903, includes seven species recovered from early and middle Miocene localities of Europe and Africa. The type species, L. filholi (Gaillard, 1899), recorded from the middle Miocene of Europe (MN6–MN7/8; Beaumont, 1973; Roth, 1988; Peigné, 2013, and references therein) is smaller and has a higher and stronger mesial accessory cuspid with a basal cingulid on p3 and p4, and the m2 alveolus is elongated and pear shaped, unlike in our specimens (Gaillard, 1899; Beaumont, 1973). Leptoplesictis aurelianensis (Schlosser, 1888) is known from the early and possibly middle Miocene of Europe (MN4–MN5; see Roth, 1988; Peigné, 2013, and references therein). L. aurelianensis has similar developments as the Thai form in the accessory cuspids on p3 and p4, the presence of a distolingual talonid on premolars, m1 trigonid wider than talonid and with a distally placed metaconid, and a low talonid with poorly differentiated cuspids. However, this species is smaller than the Thai taxon, the m1 trigonid is more elongated and the m2 is shorter (Roth, 1988). Leptoplesictis atavus Beaumont, 1973, from the middle Miocene of Europe (MN5–MN6; Beaumont, 1973; Ginsburg, 1961; Roth, 1988; Peigné, 2013), is a smaller form than the Thai species. The m1 metaconid is straight and less distal (not visible in buccal view). The paraconid and metaconid are therefore close to each other, with the distal crest of the paraconid and the mesial crest of the metaconid forming a narrow V-shaped valley, while those cuspids are more separated from each other and their crests are not connected in the Thai material (Roth, 1988). The talonid cuspids, entoconid and hypoconid, are higher than in L. aurelianensis and the Thai form (Roth, 1988). A mandible from Sansan attributed to L. atavus by Peigné (2013) allow us to compare the morphology and proportions of its premolars with our taxon: L. atavus bears a weaker distal accessory cuspid on p4 relative to the Thai form, the distal accessory cuspid is absent on p3, and this tooth is only slightly shorter than p4, unlike in the Thai form. Four additional species are described from the early Miocene of Africa: two from Rusinga Island, Kenya (Schmidt-Kittler, 1987), and two from Sperrgebiet, Namibia (Morales et al., 2008). The holotype of the Kenyan species L. rangwai shows a p4 with a considerably higher mesial accessory cuspid, approximating the height of the distal one, a lower metaconid on m1, and a relatively longer m2 as compared to the Thai species (Schmidt-Kittler, 1987). Most cuspids of p4 and m1, such as the distal accessory cuspid of p4, appear somewhat bulbous, which is different from the sharp tooth morphology of the Thai species. The second Kenyan species, L. mbitensis, differs from L. rangwai and from the Thai species by having a lingual cuspid on its p4 distal shelf (Schmidt-Kittler, 1987). Our form differs from L. senutae from Grillental (Morales et al., 2008) by a shorter and narrower p2 and by the presence of a hypoconulid on the m1 talonid. It differs from L. namibiensis from Langental (Morales et al., 2008) by its smaller size and by the presence of a lower-crowned m1 talonid. Our Thai form therefore pertains to a new species of Leptoplesictis, and represents the earliest representative of Herpestidae in the Asian fossil record.

Material: MM-42, left broken P4; stored at the DMR, Bangkok.

Description and comments (fig. 6D–F): The P4 (L = 13.7 mm; W = 5.8 mm) shows a small parastyle and a relatively high paracone (Hpara = 7.5 mm), which is separated from the metastyle by a carnassial notch. A crest extends mesially and slightly buccally from the paracone tip to join the parastyle. Although the mesiolingual part of the tooth is broken (protocone missing), it seems that the tooth is triangular, with a development of the protocone mesial to the paracone tip and close to the parastyle. A crest runs distolingually from the parastyle in the direction of the original location of the protocone. The cingulum is well marked lingually to the metastyle and buccally to the parastyle.

The presence of a carnassial notch, a distinct parastyle, a long metastyle, a poorly developed lingual region, and sharp cusps suggest that this tooth fragment belongs in Feliformia. This specimen appears morphologically closer to the upper carnassial of some viverrids (e.g., Viverra, Viverricula), prionodontids (extant Prionodon species), and felids. The tooth could correspond in size to the lower dental material of cf. Viverra sp. from Mae Moh. However, we prefer to maintain the attribution of Feliformia indet. until a more complete upper carnassial is recovered.