Pisces Lates niloticus (Nile Perch) is represented at the Niger site by the posterior basal part of the skull. The distal surface of the bone is ovoid with a conical depression much like those developed on vertebrae. Well developed growth ridges are developed inside the conical depression. The ventral part of the specimen is scored by longitudinal grooves and the dorsal surface shows an articular surface for the rest of the skull.

Crocodilia Crocodylus cf niloticus (Nile Crocodile) is represented at locality N 885 by a brevirostrine mandible with buccal depressions between the teeth, two dermal scutes, two articulated vertebrae, the occipital part of a skull and diverse fragments of skull. The mandible is the most diagnostic specimen, and preserves the canine alveolus opposite the rear of the symphysis. The anterior part of the symphysis is broken, but two complete and one damaged precanine alveoli are present. The form of the symphysis and the disposition of the precanine and canine alveoli reveal that the mandible belongs to a brevirostrine crocodile. There are six complete and one broken postcanine alveoli preserved. The alveolar process descends behind the canine, levels out at the third and fourth postcanine teeth and ascends towards the seventh alveolus. The buccal margin of the alveolar process is indented between the teeth, such that in lateral view, the bases of the alveoli stand proud of the rest of the jaw, but lingually the alveoli are confluent with the rest of the jaw. These buccal depressions would receive the upper teeth when the jaws are closed. Immediately to the lingual side of the postcanine alveoli there is a line of emmissary pits. Mesial to the canine there is a large foramen in the dorsal surface of the jaw. On the lingual side of the specimen, there is a deep groove leading anteriorly from the zone where the splenial terminates, revealing that the splenial was not implicated in the symphysis. This groove leads into a large fossa in the symphysis.

Hominoidea is represented at site N 885 by a right mandible fragment (Fig. 3, Fig. 4 and Fig. 5) containing the roots of the first molar. After comparing the Niger specimen with other mammals, including cercopithecids, hominoids, many ungulates, and carnivores, we consider that the best match for the Niger hominoid fossil is with Pan, but because the fossil is fragmentary, we hesitate to attribute it to this or to any other hominoid genus. Particular points of resemblance to African apes are the overall morphology of the jaw, the length and orientation of the roots, and the buccolingual fusion of the roots, a combination that rules out most mammals except for primates.

The mandible is slender and moderately deep. The jaw beneath m/1 is 13.2 mm thick, and its depth from the alveolar process to the ventral margin is 31.6 mm. The roots of the molar are ca 13 mm long and they curve gently distally. They are fused buccolingually (Fig. 5). The preserved part of the tooth measured at cervix is 9.4 mm × 9.7 mm, but judging from the position of the alveoli mesial and distal to the preserved roots, the crown length would originally have been ca 11 mm and the breadth about 9.9 mm. The buccal roots which have been exposed by the removal of alveolar bone by sand blasting (and perhaps by alveolar dehiscence prior to death) show an appreciable amount of flare. The sublingual fossa is shallow and low down.

This mandible fragment is compatible in size and proportions to those of medium-sized hominoids with slender mandibular rami. It is too gracile to belong to an australopithecine, even a small individual such as AL 288-1 (Lucy) [14] from Hadar, Ethiopia. Comparisons with several hominoids, both extinct and extant, reveal overall similarities to both Pan troglodytes (Fig. 4) and Homo rather than to any other genera. A low but palpable ridge low down on the lingual side of the jaw beneath m/1 could represent the distal extremity of a simian shelf where it fades into the body of the mandible. In Homo, the homologous zone of the jaw is generally hollow and is thus devoid of a ridge. The jaw is slightly less deep than that of Orrorin from the Late Miocene of Kenya [30] and the sublingual fossa is not as deeply excavated. The mandible of Orrorin is slightly more robust than that of the Niger hominoid.

The scientific value of the specimen resides in its discovery locus far from any other known fossil hominoids, its Late Miocene age and the attention that it will focus on the Neogene fossil record of West Africa, currently almost unknown, except for the Chadian specimens [3].

Anthracotheriidae Libycosaurus is represented at the site by a distal metapodial attributed to a medium-sized species, larger than L. petrocchii from Beglia and smaller than the large species from Sahabi and Chad [17]. The distal articular facet has a ventral ridge or keel (eroded away by sand-blasting). The dorsal surface of the articulation is smoothly cylindrical without grooves (unlike some hippopotamid specimens which do have shallow grooves that extend onto the dorsal surface from the deep grooves found on the plantar surface), and terminates proximally in a shallow curved depression which separates it from the diaphysis. A shallow ligamentary groove leads proximoaxially from this depression (Fig. 8). The medial and lateral sides of the epiphysis show deep ligamentary pits. The plantar aspect of the diaphysis is flat, with low marginal ridges. The axial side is relatively flat with two longitudinal grooves near the distal end, without any swelling distally. The dorsal surface is gently curved, sloping more laterally than medially. The section of the bone at the break reveals a compressed dorsoplantar morphology, 22.6 mm dorsoplantar diameter by 42.4 mm mediolateral diameter. The distal end is 44.5 mm broad (Fig. 6). Comparison with a sample of Brachyodus onoideus metapodials from the Faluns de la Touraine, France, suggests that it is a IVth metatarsal, in which case the Niger fossil would be from the right hind limb.

The Niger metapodial differs from those of hippopotamids in a number of features. Apart from its more dorsoventrally compressed section (Fig. 7) the diaphysis is perpendicular to the distal articular facet, whereas in hippopotamids the abaxial margin of the diaphysis is inclined and not vertical. In hippopotamids the abaxial part of the dorsal surface of the diaphysis usually slopes quite strongly towards the plantar side, making the cross section of the diaphysis sensibly narrower (and with a curved profile) on the abaxial side than the axial one. In anthracotheres, in contrast, the section narrows only slightly towards the abaxial side of the bone, as in the Niger specimen (Fig. 6) and the abaxial edge is almost flat. In hippos, the distal end of the metapodial diaphysis sports prominent axial and abaxial swellings close to its zone of fusion with the distal epiphysis (Fig. 8). As a result of this swelling, the diaphysis is wider distally than it is near its centre. In anthracotheres, in contrast, the abaxial swelling is minor and the axial one is nonexistent, and as a consequence, the diaphysis is almost constant in breadth from its centre towards its distal end. This morphology is related to the structure of the foot, which in anthracotheres has the axial metapodials in close contact with each other throughout their length, whereas in hippopotamids, the metapodials splay outwards distally, thereby broadening the feet.

The dimensions of this metapodial are greater than those of Libycosaurus petrocchii from Beglia, Tunisia, and belong to the smaller of the two unnamed anthracotheres that occur in Chad and Libya [17], [24] and [25] (Fig. 9). Of these the smaller species ranges in age from about 10 to 8 Ma and the larger from ca 8 Ma to perhaps as young as 6–5 Ma [24]. On this basis the Niger collection is expected to be between 10 and 8 Ma but we take a cautious view and correlate it to the Late Miocene (11–5 Ma).

The value of this anthracothere specimen, even in its fragmentary state, is that it yields evidence of the age of the deposits from which it came. Secondly, it indicates the former presence of shallow swamps or lake margins in what is now an extremely arid part of the world.

Bovidae. The family Bovidae is represented at point N 885 by a left frontal with horn core (Fig. 10) a distal left radius and a left calcaneum lacking the talar process and part of the sustentaculum tali. The base of the horn core is 34 mm mediolateral diameter by 29 mm anteroposterior. It is inserted above the orbit, and its base on the lateral side is lower than the dorsal edge of the frontal. The frontal bears a precornual fossa. The horn core itself has no obvious sign of transverse ridges. The frontal is thick, the pedicle short, and there is no penetration of a sinus into the base of horn core, only a foramen in the roof of the orbit. It preserves the interfrontal suture and a part of the parietal suture. The postcornual fossa is deep and relatively large (8 mm × 5 mm) and is located in a posterolateral position. The medial keel is well marked, although displaced posteriorly and the lateral one is weakly marked. The vertical grooves are exceptionally deep where associated with the medial keel. The horn core slopes backwards at a marked angle and diverges at about 30° from the sagittal plane. The apex of the horn core is broken off, but as preserved it is 77 mm long measured anteriorly from the top of the pedicle.

The morphological features of this fossil frontal and horn core are only encountered in representatives of the bovid subfamily Hippotraginae, and more particularly in members of the tribe Reduncini. In overall aspect it differs from species of Kobus (which is well represented in the African fossil record) by the transverse section of the horn core, which is more circular in Kobus, with the mediolateral diameter slightly greater than the anteroposterior one. Lehmann and Thomas [16] used the more circular outline of the section of the horn core as the basis for attributing two frontals and a horn core from the Late Miocene of Sahabi, Libya, to Redunca aff. darti, differentiating them from Kobus subdolus from the Latest Miocene of Langebaanweg, South Africa [9], and the reduncines from the Late Miocene of Mpesida and Lukeino, Kenya [32]. For the same reasons, the Niger specimen differs from the reduncines from the Late Miocene of Lothagam, Kenya [12].

The Niger bovid species is smaller than Redunca darti from Middle Pliocene of Makapansgat, South Africa (type locality of the species), and from Redunca aff. darti from Sahabi, Libya [16], although it shares morphological affinities with the latter specimens.

Lehman and Thomas [16] described a bovid frontlet from the Late Miocene of Sahabi, Libya, which they attributed to R. darti, which differs from the Niger specimen by the greater divergence of its horn cores, but in other respects the two specimens are concordant, save in their dimensions, the Niger specimen being appreciably smaller than the Sahabi ones. Gentry [11] in contrast, considered that the Sahabi specimens previously attributed to R. darti, belong instead to Kobus aff subdolus. The Niger fossil shows some similarities to a horn core from Lukeino, Kenya, attributed to Kobus porrecticornis by Thomas [32] although the latter specimen does not show the posteromedial flattening at the base, although it does show flattening near the apex. K. porrecticornis was first described from the Middle Siwaliks of Pakistan (Dhok Pathan Zone, ca 7 Ma) and similar fossils occur at Sahabi, Lothagam, Langebaanweg, Lukeino [8] and other Late Miocene sites in Africa.

Zephyreduncinus oundagaisus from the Late Miocene of Ethiopia [33] is markedly divergent from the Niger species. The Ethiopian species has strong horn core compression, the opposite of the Niger specimen and, in addition, the disposition of the keels is different. In Zephyreduncinus (Vrba and Haile-Selassie [33], figure 2) the keel is anterior with slight torsion outwards, whereas in the Niger specimen the keels of the horn are medial and lateral.

Both Redunca and Kobus are aquaphile bovids [10] and [32] living close to water, or in tall grass plains, although they can inhabit quite open country but usually not far from riparian woodland to forests.

Despite the restricted quantity of fossils from the Niger locality, two specimens are useful for biochronology. The most evocative evidence is that of an anthracothere, Libycosaurus sp. equivalent to the medium-sized species from Toros-Ménalla, Chad and Sahabi, Libya [17] and [24]. Secondly, the bovid horn core has some resemblances to material from Sahabi (Libya), Lukeino, Mpesida and Lothagam (Kenya) and Langebaanweg (South Africa). Combined, the two fossils suggest that an age between 11 and 5 Ma is possible for the deposits that yielded these fossils.

The presence of Nile Perch, Lates niloticus, in the Niger Late Miocene deposits indicates the former presence of a freshwater lake or large river in the country. It is not known whether this would have been part of Late Miocene Palaeolake Chad (not to be confused with Holocene and Late Pleistocene Megachad) [21], or whether it was a separate lake. If the latitude and longitude of the locality are close to those indicated by Bergounioux and Crouzel [2] for the type skull of Nigeremys gigantea as suggested by the field label (Fig. 2), then it would in all probability have been a separate lake. The anthracothere Libycosaurus was likely a denizen of shallow swampy parts of the palaeolake or river, and the reduncine probably lived close to water. The palaeoenvironment was possibly rather like that of the extant Lechwe (Kobus leche) which is common in the Okavango Swamp, Botswana. Unlike the Okavango, however, the Niger palaeolake had vast and deep fully lacustrine zones in which Nile Perch could grow to large dimensions.

Nonhuman hominoids are today confined to forests and woodlands, but the common chimpanzee can survive in subhumid environments as long as there are enough trees to supply adequate food and security [18]. This means that the longest dry period should not exceed two months, and that the area should be endowed with riparian forest so that there is always a supply of fresh vegetation in the form of fruit, leaves, pith and bark, even during the driest months of the year. In general however, chimpanzees and gorillas flourish best in tropical rainforests of Guineo-Congolian affinities [34]. At present, the most arid environment inhabited by chimpanzees is wooded savannah in Liberia and Miombo woodland in Tanzania [18], [34] and [35]. If the sort of “prechimp” represented by the Niger fossil can be considered as having the same habits as the extant species, then it may indicate that the country was considerably more humid then than it is today, and was at least covered in wooded savannah if not denser vegetation categories.

The inferred former presence of a Late Miocene palaeolake or large river in Niger inhabited by Nile Perch, surrounded by countryside covered in wooded savannah if not denser vegetation, suggests that the mean annual rainfall was probably greater than 750 mm, and possibly as much as 1200 mm. At present, Niger is one of the driest countries in Africa, much of it, such as Tenéré, being total desert.

The discovery of a Late Miocene aquatic and terrestrial vertebrate fauna in Niger confirms the view that the Sahara is a relatively young desert, postdating the Late Miocene. Recent work in Egypt has revealed that central Egypt was covered in woodland to forests some 11–10 Ma [28] and [29] and Chad is long known to have been more humid in the past than it is today [5] and [6] as was Libya [7] and Tunisia [22] and [23].

At present, the chimpanzees living closest to the Niger fossil site occur 940 km to the south in Ghana and Nigeria and 1700 km to the west in Mali (Fig. 11) [4], [34] and [35]. Niger thus joins an increasing list of African countries that have yielded remains of Late Miocene (ca 11–5.5 Ma) hominoids including, Kenya (Lukeino, 6 Ma, two taxa [20] and [29]; Samburu Hills, 9.5 Ma [13]; Nakali, ca 10 Ma) [15], Ethiopia (Ch’orora, ca 10.5 Ma) [31] and Chad (Toros-Menalla, ca 10–6 Ma) [24] and it indicates that further research in West Africa will undoubtedly add fuel to the debate about hominid origins.

It is clear that during the Late Miocene, hominoids were widespread in Africa (Fig. 1) and although the quantity of fossils is restricted, the diversity is large, with at least eight lineages already recognised (in the order in which they were reported: Otavipithecus, Samburupithecus, Orrorin, Ardipithecus, Sahelanthropus, an unnamed protogorilla-like form, Chororapithecus, and Nakalipithecus) [26]. Under the circumstances, it is necessary to reconsider the “Back to Africa” hypothesis which invokes a Late Miocene reintroduction of hominoids into Africa from Eurasia in order to give rise to extant African apes and hominids [1]. The European genera Dryopithecus (sensu lato), Pierolapithecus [19] “Austriacopithecus” and eventually Ouranopithecus, upon which the « Back to Africa » hypothesis is based, could all represent lineages that migrated from Africa towards Europe, rather than the other way round. Otavipithecus from 13–12 Ma deposits in Namibia, for example, possessed an atlas vertebra morphometrically close to that of Pan paniscus, and its frontal bone [27] shows similar frontal sinus morphology to that of Pan and Gorilla, which indicates that a late Middle Miocene fossil ape from Africa possessed postcranial and cranial anatomy foreshadowing that of extant African apes, and differing from that of Eurasian apes, both fossil and extant. Nakalipithecus [15] from the Late Miocene of Kenya is similar in many ways to Ouranopithecus, but it is older than it, suggesting an African origin for this particular lineage. When additional postcranial remains of African Late Miocene hominoids are found, it will be possible to test the « Back to Africa » hypothesis in a more rigorous way, as the hypothesis relies largely on postcranial resemblances between extant African apes and European Late Miocene hominoids.

A vertebrate fauna of Late Miocene age from Niger is of interest on account of its age and the presence within it of a hominoid primate and an anthracothere. The discovery helps to fill what was a vast gap in the geographic coverage of fossil hominoids; the nearest known fossil specimen of comparable age being from Chad, over 1000 km to the east and 3500 km from sites in Ethiopia and Kenya that have yielded Late Miocene hominoids. The site is closer to the Spanish Late Miocene hominoid sites (3000 km) [19] than it is to the East African ones (3500 km)! The find site, the position of which is not precisely known at present, but lies within the area of outcrop of Mesozoic strata in central Niger, is about 940 km north of the closest extant hominoids (Pan troglodytes) in Ghana and Nigeria. The fauna as a whole indicates the former presence of a freshwater lake or large river in Niger, and a Late Miocene palaeoclimate considerably more humid than that of today.