Haile-Selassie and Howell [3] assigned ALA-VP-2/106, ALA-VP-2/214, and STD-VP-1/2 to Lutrinae gen et sp indet. Full descriptions are also presented there and will not be repeated here. However, detailed comparison subsequent to the completion of Haile-Selassie and Howell [3] has revealed that these specimens belong to an early species of Torolutra. ALA-VP-2/106 and ALA-VP-2/214 possibly belong to the same individual largely due to the proximity of their discovery and similar coloration. ALA-VP-2/106 is metrically and morphologically similar to M₁s of specimens assigned here to Torolutra ougandensis and also to M₁s of the species from Nkondo [21] and Lukeino [19]. Morphology of the humerus (STD-VP-1/2) shows aquatic adaptation and its size is similar to younger specimens assigned to the species and also to extant large size lutrines.

GAW-VP-1/5 is a left mandible anteriorly broken at the P₄ root level. The ascending ramus is missing. It is morphologically similar to all the mandibles assigned here to cf. Torolutra ougandensis. The teeth are preserved from P₄-M₁. The P₄ is double-rooted with a tall main cusp. There is an accessory cusp distal to, and attached to, the main cusp. The accessory cusp is half as long as the main cusp. Its overall dental morphology and metrics are very similar to the Siwaliks lutrine, Vishnuonyx chinjiensis [22]. The distal tubercle is low but well-developed. The P₄ has a strong buccal cingulum and a weaker lingual cingulum. The M₁ is complete. The paraconid is the largest cusp, followed by the paraconid, whereas the metaconid is the smallest. The paraconid is apically worn. The paraconid and protoconid are separated by a deep, slit-like notch, and a buccal wear facet is present on both cusps. At its base, the M₁ is surrounded by a cingulum. Talonid cusps are not distinguishable. The flat and relatively low talonid slopes lingually. The M₁ is morphologically identical with specimens such as GAW-VP-1/146 and GAW-VP-3/25. The cusp proportions of these specimens are also similar.

GAW-VP-1/6 is the posterior half of a right edentulous mandible, morphologically similar to the other mandibular specimens described here but slightly shallower. This specimen appears to belong to a young individual.

GAW-VP-3/25 is a complete mandible with left and right canines and a distal right M₁. The symphysis is extremely receding and hence, distally elongate. Preserved root sockets indicate the presence of P₂ and M₂. The M₂ is single-rooted and sloping. The left C₁ is unworn and the apical half of the right C₁ is missing. The mandible is very shallow, with multiple mental foramina. The talonid of the M₁ slopes lingually. A strong cingulum is present on the buccal faces of the protoconid continuing to the talonid. The canine crown is highly convex anteriorly and its base is circled by cingulum except above the labial CEJ. Judging from the sockets, the incisors were small, two on each side. The ascending rami are missing. It has a very deep masseteric fossa extending medially as far as the distal M₂. The left C¹ is an unworn crown with a complete and inflated root. It has incipient vertical wear on its mesial crest. The crown is anteriorly less convex compared to the lower canines. The cingulum is manifested in the same way as in the lowers. Associated postcranial elements include a pelvic fragment, a caudal vertebra, a left calcaneum, and a proximal MTP fragment.

GAW-VP-1/146 is a right mandible with M₁. The root socket for the P₂ is entirely obliterated and the ascending ramus is missing. It has a strongly sloping symphysis. It is morphologically identical with GAW-VP-1/5 and GAW-VP-3/25. WKH-VP-1/32 is a left edentulous mandible with the distal root of the M₁ and M₂ preserved. The preserved morphology of the mandible is similar to the other mandibles referred to Torolutra. GAW-VP-1/173 is a complete right C₁ and GAW-VP-1/177 is a right C₁ with a complete root, but missing the apical half of the crown. GAW-VP-1/173 is stouter and lower crowned compared to the other lower canines described above. It also has a strong distal cingulum. However, its morphology is similar in having a distally curved crown relative to the root.

GAW-VP-1/3 comprises maxillary fragments with associated left P³ and a right M¹ fragment. The P³ is mesially narrower and distally wider. The main cusp is tall relative to the length of the crown. It is two-rooted and the main cusp is anteriorly slightly convex. The M¹ mesiolingual portion is broken. 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. GAW-VP-1/165 is a left maxillary fragment with the root sockets of the canine and P². Its size and morphology best fit with the specimens described above.

Haile-Selassie and Howell [3] tentatively assigned specimens ALA-VP-2/106, ALA-VP-2/214, and STD-VP-1/2 to Lutrinae gen. et sp. indet. However, further comparative analysis shows that these specimens belong to Torolutra ougandensis. This species was named by Petter et al. [21] to refer to lutrine specimens from the Nkondo and Nyaburogo Formations of Uganda. More specimens were referred to this species from Plio-Pleistocene sites such as Koobi Fora [30]. The earliest record of this species in Africa possibly comes from 6-7 Ma deposits in Chad. Peigné et al. [20] assigned TM 266-01-194, a mandible with a carnassial, from Toros Menalla of Chad to aff. Torolutra sp. The carnassial and mandibular morphology of this specimen is similar to the specimens from the Middle Awash and Lukeino referred to Torolutra ougandensis. This might be an indication that the Toros Menalla specimen belongs to the same species. However, the M₁ length (16.0 mm) is slightly larger in the Toros Menalla specimen compared to other Torolutra M₁s from eastern Africa. This species appears to have been widespread in central and eastern Africa during the Late Miocene and Early Pliocene. It persisted to the Early Pleistocene deposits of eastern Africa [30]. Although the origin of this species is currently unknown, it probably descended from a Eurasian species similar to Paralutra or Paludolutra, or from a species similar to the Siwaliks Vishnuonyx chinjiensis, particularly based on the P₄ morphology and dimensions. Moreover, the holotype of Vishnuonyx angololensis matches, metrically and morphologically, a Torolutra ougandensis P⁴ from Lukeino [19]. However, it should be noted that Late Miocene Eurasian and African lutrines showed a mosaic of dental features shared among different species of the subfamily [28].

AMW-VP-1/61 is the lingual two-thirds of a left M¹ and it is bunodont. The parastyle, paracone and metacone are missing. The paraconule is as large as the protocone and they are almost fused, separated only by a shallow groove. The metaconule is small and positioned close to the metacone position. The area between the preserved cusps and cuspules is a wide, flat basin. The lingual cingulum is strong mesial to the protocone and continues distally, rising to merge with the hypocone. The lingual length at the protocone is 14.1 mm.

GAW-VP-1/58 is a complete bunodont right M₁ crown lacking the roots. The trigonid cusps are separated from each other by V-shaped valleys. The paraconid is the largest cusp, followed in size by the protoconid and metaconid, respectively. The paraconid and protoconid are of equal height and the metaconid is slightly shorter. The protoconid is apically worn and dentine is exposed on its apex. The talonid is wide and the hypoconid and entoconid are well-developed. The accessory cusplet distal to the protoconid is larger than the cusplet behind the entoconid. Additional small cusplets are present on the distal occlusal rim of the talonid. The latter is lingually expanded, particularly at the entoconid level. It has a strong basal cingulum on the buccal side and a weaker one on the lingual side. The talonid basin is positioned low relative to the trigonid, and is deep at its center. Crown morphology is typical of Enhydrini. The crown is 20.4 long and 12.5 wide. The trigonid and talonid lengths are 11.6 and 8.7, respectively.

GAW-VP-1/162 is a left P⁴ fragment. Most of the lingual, mesial, and mesiobuccal parts of the crown are missing. The paracone is extremely worn, with a large dentine exposure on its cusp apex. This dentine exposure extends distally to the apex of the metastyle. The crista obliqua of the protocone is preserved and connects to the lingual crest of the paracone to form the distal wall of the anterior valley. The entire metastyle and most of the hypocone are preserved. The hypocone is relatively large. A strong distal cingulum is present and continues lingually around the hypocone and buccally around the metastyle. Although much of its morphology remains absent due to its fragmentary nature, this specimen best matches BAR 566’05 from Lukeino [17].

Specimen ASK-VP-3/62 is a left P⁴ fragment from Asa Koma Vertebrate Locality 3 (ASK-VP-3) located in the Asa Koma Member of the Adu-Asa Formation, dated to 5.54-5.77 Ma. ASK-VP-3/62 (Fig. 5) is fragmentary and most of the diagnostic occlusal morphology is not preserved. Although Haile-Selassie and Howell [3] assigned ASK-VP-3/62 to Sivaonyx cf. africanus, they indicated that it is similar in its preserved parts to specimen BAR 1720’00 from the Lukeino Formation assigned to S. senutae [17]. The size of ASK-VP-3/62 is large for the latter species and appears to best fit within the size of Sivaonyx soriae [17], the larger Sivaonyx species from Lukeino, although its P⁴ is unknown. Given that size is the only difference described by Morales and Pickford [17] between S. soriae and S. senutae, and considering the contemporaneity of ASK-VP-3/62 with Sivaonyx soriae, it is more plausible to tentatively assign this specimen to Sivaonyx sp. cf. S. soriae. Discovery of more complete specimens from the Asa Koma Member will refine this tentative identification.

Sivaonyx species have always been difficult to distinguish from Enhydriodon species in the Late Miocene of Eurasia and Africa [19]. In some cases, upper and lower dentitions of the same taxa were assigned to different species (details discussed in [19]). While the distinction between Sivaonyx and Enhydriodon species from the Late Miocene/Early Pliocene remains elusive, new discoveries in Africa have increased the sample size of Enhydrini. These new fossil discoveries resulted in the naming of Sivaonyx beyi, a new species from Toros Menalla (Chad) [20], as well as the taxa Sivaonyx soriae, S. senutae, S. kamuhangirei, S. ekecaman, and discovery of more S. africanus material, all from Plio-Miocene sites in eastern and South Africa [3], [17], [18] and [29]. The Langebaanweg material initially identified as Enhydriodon africanus [9] was renamed Sivaonyx hendeyi sp. nov. [19]. The new Sivaonyx species from Chad, S. beyi Peigné et al., 2008, appears to be different from the Lukeino and Middle Awash Sivaonyx both in terms of its size and morphology of the upper and lower first molars. GAW-VP-1/58 from the Middle Awash is morphologically similar to the M₁s of KNM LU 337 and BAR 1984’05 from the Kapcheberek Member of Lukeino Formation. The former specimen was initially assigned to S. africana [19] but later reassigned as the holotype of S. soriae [17]. AMW-VP-1/61 (M¹) is slightly larger in its size and close to Sivaonyx ekecaman from Mabaget (Fig. 6A). It also appears to be larger than BAR 1082’01, a specimen from Lukeino initially assigned to S. africana [19] but later reassigned to S. senutae [17] (Fig. 6B). However, morphology of the lingual cusps and the talon of AMW-VP-1/61 and BAR 1082’01 is very similar.

The difference between S. soriae and S. senutae is merely size, with the former being larger [17]. However, their dental morphological similarities or differences have not been documented due to lack of molars from the homologous position. Sivaonyx beyi has a M¹ morphologically different from the Lukeino, Middle Awash, and other African Sivaonyx M¹ specimens. AMW-VP-1/61 has a lingual length (14.1 mm) larger than both TM 90-00-066 (11.7 mm) [20] and BAR 1082’01 (12.2 mm) [19]. However, it is smaller than the M¹ of S. ekecaman from Kanapoi (15.8 mm), which is much younger [29]. One of the evolutionary trends hypothesized for Sivaonyx species is an increase in body size [17], [18] and [22] and the lingual breadth of the M¹ tends to support this observation (Fig. 6B). However, the fact that M¹s of S. beyi are larger than all other Sivaonyx species falsifies this hypothesis. Alternatively, if there was indeed an increase in body size through time, available evidence then challenges the geological age of S. beyi. On a more general observation, given the temporal distribution and morphological overlap seen in the currently named Late Miocene/Early Pliocene African Sivaonyx species, it appears that they are taxonomically oversplit. Hence, a thorough investigation with increased hypodigm is now required.

GAW-VP-1/1 is a left mandible with M₁. It is anteriorly broken at the level of the M₁ mesial root. Posteriorly, a small portion of the root of the ascending ramus is preserved. The masseteric fossa is deep and anteriorly extends to the level of the M₂. The mandible is relatively deep, measuring 15.8 mm at mid-M₁ level. The M₁ is relatively unworn and very strong shearing carnassial. The length and breadth of the M₁ are 13.5 and 7.2 mm, respectively, with a trigonid length of 8.2 mm and talonid length of 5.3 mm. The protoconid is the tallest cusp, followed by the metaconid, although most of the trigonid surface area is occupied by the paraconid. The talonid is set very low and horizontally flat with its buccal occlusal rim slightly higher than its lingual occlusal rim. The talonid is lower than the trigonid, the paraconid and the protoconid are separated by a deep incision. The buccal cingulum is stronger than the lingual cingulum. The root socket preserved behind the M₁ indicates that the M₂ was single-rooted. The mandibular foramen is positioned much farther below the tooth row and is anteroposteriorly long. The insertion for the digastric muscle is broad anteriorly.

The earliest Vishnuonyx in Africa, which is also contemporaneous with the Siwalik type species of the genus, was documented from the Ngorora Formation of Kenya from deposits dated to ca. 12 Ma [18]. A second species was identified from the latest Miocene Lower Nawata Member of Lothagam, Kenya [28]. These two taxa were recognized based on a single tooth from each species. GAW-VP-1/1 is the most complete mandible of Vishnuonyx found in Africa thus far. Its similarity with the Siwaliks Vishnuonyx chinjiensis, particularly in the morphology and metrics of the M₁, is striking. The presence of a well-developed shearing surface on the buccal side of the carnassial cusps are characters shared between GAW-VP-1/1 and Vishnuonyx chinjiensis Stromer, 1932 as seen in specimen NHM G 4 described from the Siwalik of Pakistan [22]. However, GAW-VP-1/1 is larger than NHM G 4 and further differs from this specimen by having a deeper masseteric fossa and deeper mandible. The relationship between the Vishnuonyx from the Middle Awash and Lothagam (V. angololensis Werdelin, 2003a) is currently unknown because there is no mandible and/or M₁ known for the latter, or a P⁴ known from the Middle Awash. However, Morales et al. [19] indicated that BAR 270’03, a left P⁴ that they assigned to Torolutra ougandensis, is close in size and morphology to the holotype of Vishnuonyx angololensis. This raises two issues: the holotype of V. angololensis is either a misidentified T. ougandensis specimen, or the genera Vishnuonyx and Torolutra have dental morphological similarities. However, the Middle Awash Vishnuonyx and Torolutra specimens appear to have significant differences in their M₁ morphology. Given the similarities in the mandibular and M₁ morphology of V. chinjiensis and Vishnuonyx sp. from the Middle Awash, it appears parsimonious to infer that the African Vishnuonyx are closely related to the Asian V. chinjiensis indicating biogeographic relationships between Africa and Eurasia during the Late Miocene. The lack of similar elements to compare GAW-VP-1/1 with the Lothagam V. angololensis renders it difficult to assign the Middle Awash specimen to the latter species. However, the temporal and spatial proximity of the two areas would suggest that they may be conspecific.

ASK-VP-3/411 is a complete right M¹ crown with most of the lingual and mesiobuccal roots, but lacks the distobuccal root. The lingual root is compressed mesiodistally and long buccolingually. The crown is unworn and low. The parastyle is very large. It has a well-developed metastyle buccal to the metacone. The paracone is slightly smaller than the metacone, a bulbous cusp. Immediately lingual to the metacone is a well-developed, rounded metaconule slightly smaller than the metacone. There is no sign of a paraconule and a wide basin separates the paracone and protocone. The rounded protocone is separated from the mesiodistally elongate hypocone by a narrow groove. The mesial cingulum is well-developed particularly at the mesiolingual corner where it ascends to merge with the hypocone.

ALA-VP-3/411 (Fig. 4) is morphologically different from, and smaller than, M¹s of most Sivaonyx species. It differs from BAR 1082’01, a specimen from Lukeino assigned to S. africana [19], later re-assigned to S. senutae [17], in the lack of a paraconule, presence of a metastyle, and overall size. It belongs to a lutrine the size of Aonyx or modern Lutra lutra. This molar is buccolingually long and mesiodistally compressed compared to Sivaonyx, Aonyx, and Procyon. Its proportions are more like Lutra in being buccolingually elongate. Most M¹s of Lutra species also lack the paraconule and have large parastyles. Unfortunately, upper first molars are unknown for Torolutra ougandensis from the Late Miocene or Early Pliocene, and this makes it difficult to assign ASK-VP-3/411 to the species with confidence. The assignment of ASK-VP-3/411 to aff. Torolutra sp. is largely based on the presence of this genus in the Asa Koma Member and the fact that it is much smaller than all Sivaonyx species known from the Middle Awash. On the other hand, it could also represent the earliest Lutra from eastern Africa because the buccolingual elongation and lack of the paraconule appear to be characters seen in extant Lutra species. Further discoveries might refine the taxonomic affinity of this specimen.