Depuis 1998, la Kenya Palaeontology Expedition a découvert des restes de Tragulidae à Mosionin (en 1998) et Tabarin (en 2003), et un paon à Sagatia, toutes des localités de la formation pliocène de Mabaget (5,3–4,5 Ma), situées au pied des collines Tugen, dans le Nord du Kenya. Ces découvertes sont très importantes, car elles prouvent la présence d'une forêt tropicale humide au moment du dépôt et confirment que les hominidés, qui sont connus dans cinq localités de la formation, sont associés à des paléoenvironnements forestiers.

In 1998 and 2003, the Kenya Palaeontology Expedition discovered tragulid remains at Mosionin and Tabarin, Early Pliocene (5.3–4.5 Ma) and a peafowl at Sagatia, all localities in the Mabaget Formation, foothills of the Tugen Hills, northern Kenya. These finds are of great significance as they indicate the presence of tropical forest in the region at the time of deposition. Early hominids, which have been found at six localities in the same formation, are thus associated with forest palaeoenvironments.

The Mabaget Formation (Fig. 1), previously called the Chemeron Formation – Northern Extension) [30], is underlain by trachytes and basalts of the Kaparaina Formation, the top of which is aged ca 5.3 Ma. Ignimbrites and volcanic tuffs within the Mabaget Formation have yielded ages of 5.3 Ma and 4.3 Ma [6] and [30]. These volcanogenic deposits respectively underlie and overlie the main fossiliferous levels, which contain a diverse and abundant fauna, including early hominids and an age of between 5.3 Ma and 4.5 Ma is inferred for them. This estimate is slightly older than the age of 4.49–4.41 Ma published by Deino et al., [6] for the Tabarin hominid mandible [12]. Hominid fossils have now been recovered from six separate localities in the Mabaget Formation, i.e. Mabaget [30] and [41], Pelion, Ngetabkwony, Tabarin [12], Sagatia, and Tabarin North, all in the basal part of the formation. The tragulid specimens were also found at this level, at Mosionin and Tabarin and the peafowl at Sagatia.

Bar 21'98, from Mosionin, is an unworn left upper M1/ lacking its metacone (Fig. 2(1)) while BAR 2236'03, from Tabarin, is an unworn right second incisor (Fig. 2(2)).

The M1/ is slightly more than 8 mm in mesiodistal length and 9.5 mm in buccolingual breadth (Fig. 1(1)). The protocone is brachyselenodont with a prominent, sharp, lingual cingulum, which extends onto the anterior aspect of the cusp. The anterior crest of the protocone extends buccally and reaches the anterior paracone style. The posterior crest of the protocone ends in the midline of the crown but does not join the anterior crest of the hypocone. The paracone is conical, with strong but low anterior and weak posterior styles, and a prominent anterior crest running from the apex of the cusp to the anterior style paralleling an antero-buccal groove that ends at the base of the anterior style on its buccal side. The hypocone is brachyselenodont with a distal cingulum that extends lingually where it forms a low swelling near cervix. The anterior hypocone crest extends across the crown and ends at the base of the metacone, but it does not reach the posterior crest of the paracone. The posterior crest of the hypocone ends at the base of the metacone. The enamel is lightly wrinkled. The two lingual roots are fused throughout their length, the distal root being smaller than the mesial one (Fig. 2(1c)), as is typical of M1/s of the genus, the M3/ often having very short roots, especially the distal one, which may even be absent.

The Mosionin molar differs from those of Dorcatherium species in its more selenodont cusps, but most importantly by the orientation of the posterior crest of the protocone. In European and African Dorcatherium species similar in size to or slightly larger than Hyemoschus aquaticus, this crest is oriented directly distally [28] and [50], whereas in Hyemoschus and the Mosionin tooth it is directed disto-buccally. Furthermore, at its termination there is a tiny swelling that, with wear, gives a curved appearance to the crest, more evident in worn teeth than in unworn ones. Thus, the Mosionin tooth is much closer morphologically to Hyemoschus than to Dorcatherium or other Miocene tragulids such as Dorcabune [33] and Siamotragulus [44] or extant ones such as Tragulus.

Comparison of measurements of the upper molars of Hyemoschus aquaticus with the fossil from Mosionin shows that it falls within the range of metric variation of the upper first molars of the extant species (Fig. 3). Because the Mosionin tooth is close in size and morphology to the upper first molars of Hyemoschus aquaticus, we assign it to this species. This is the first record of fossil Hyemoschus in Africa. Tragulidae are common in Early and Middle Miocene deposits of Africa [27] and [28], but they have all been attributed to the genus Dorcatherium, which is close to Hyemoschus in many features, although the dentition of the former genus seems to be slightly more bunoselenodont.

The second lower incisor from Tabarin has a long, narrow crown with a distinctive bend one third of the distance from cervix to apex (Fig. 2(2a–e)). The crown is slightly compressed labio-lingually with a prominent lingual ridge that emanates from a low basal tubercle and ends near the apex of the tooth (Fig. 2(2e)). On the lingual side of the crown, the mesial and distal edges are marked by sharp crests. The labial surface is smooth and relatively flat. The root is long and circular in section. The crown is 1.4 mm mesio-distally, 1.6 mm labio-lingually, the crown is 4.9 mm high and the root is 5.6 mm long.

Comparisons with extant ruminant incisors reveal that the Tabarin tooth is similar morphologically and metrically to lower second incisors of the water chevrotain Hyemoschus aquaticus (Fig. 2(3)). It differs from the third incisor and canine of tragulids, the crowns of which are more uniformly curved, but not bent. The first incisor of tragulids is spatulate with a re-entrant angle distally into which fits the bent crown of the second incisor (Fig. 3b). Bovids such as Ourebia and Cephalophus possess second incisor crowns, which are not very spatulate, superficially recalling those of tragulids, but the bend in the incisor occurs at the crown/root margin rather than within the crown. In addition, the lingual crests tend to be less well developed than they are in Hyemoschus. Most other bovids have second incisors that are appreciably more spatulate than the Tabarin tooth, and any close relationship between this tooth and those of bovids can be dismissed.

A specimen (1952–1980) of Hyemoschus aquaticus in the Department of Comparative Anatomy of the ‘Muséum national d’histoire naturelle’, Paris, has an i/2 crown that is 5.1 mm high, ca 1.8 mm mesio-distal and 2.1 mm labio-lingual, which is quite similar in dimensions to the Tabarin tooth. Because of the metric and morphological similarities of BAR 2236'03 to lower second incisors of Hyemoschus aquaticus, we conclude that it should be identified as such rather than as any other ruminant.

The ‘water chevrotain’ Hyemoschus aquaticus is a small nocturnal African ruminant that dwells exclusively in rain forest and is never found far from water [13]. It is predominantly a frugivore, eating fruits that have fallen to the forest floor [8], but it is known to devour crabs, molluscs and carrion. It also eats vegetation that it ruminates. It has a highly developed sense of olfaction, searching for food on the forest floor at night. When alarmed, it seeks water that it dives into, hiding under waterweeds or under overhanging banks and roots of trees.

Its present-day distribution is limited to tropical forest in which fruit is available for more than nine months per year [10], [23] and [36]. This kind of forest is of very low seasonality. As such, the discovery of fossils of Hyemoschus aquaticus in Kenya signifies the former presence of rain forest in the vicinity of the basin at the time of sedimentation, between 5.3 Ma and 4.5 Ma. This suggests a significant expansion of tropical forest eastwards of its present day limits (Fig. 1). Extant African tragulids are confined to regions in which the mean annual rainfall is equal to, or greater than, 1500 mm per annum. They do not occur in areas that have less than 40 kg m^–2 yr^–1 of precipitable water in the atmosphere [25] and which are even moderately seasonally arid.

Extant peafowls comprise two large species, Pavo cristatus and Pavo muticus, which live in southern and southeastern Asia, and a small species, Afropavo congensis, which lives in tropical forest of central Africa. Two extinct species have been described in Europe, Pavo aesculapi (Gaudry), from the Late Miocene and Early Pliocene, known only from southwestern and central Europe, and Pavo bravardi (Gervais), from the Pliocene, known from both eastern and western Europe. Pavo bravardi is a large form, somewhat larger than the extant species of the genus. It disappeared at the end of the Pliocene, most probably due to the onset of major cooling at the start of the Pleistocene [2], [21] and [22].

Remains of a large peafowl were recently reported in Africa, at Aramis 1 and Aramis 6, in Ethiopia, dating from the Early Pliocene [19]. This form is close to both the extant species, P. muticus, and the extinct one, P. bravardi, and it has been identified provisionally as Pavo sp.

The specimens from the Mabaget Formation comprise two distal extremities of tibiotarsus, a left and a right, probably from a single individual. They present typical morphology of the genus Pavo and their dimensions are slightly lower than known tibiotarsi of P. bravardi, but there is strong sexual dimorphism in peafowls, and the size of the specimens falls between the ranges of variation of the two extant species (we are not able to make direct comparisons with material from Aramis, as no tibiotarsi have been reported from there). Their dimensions are clearly greater than those of Afropavo congensis the Congo Peafowl. The Mabaget material is thus identified as Pavo sp.

The presence of a peafowl in the Mabaget fauna is important from the point of view of palaeoecology. The Blue Peafowl, Pavo cristatus, lives in open, generally deciduous, forest, with a vegetation understory, along watercourses. The Green Peafowl, Pavo muticus, occurs in Thailand in riverine forests and in Java where it lives in open forest, or forest edges, but not in dense tropical rainforest [7]. From this we can conclude that the depositional environment of the Mabaget Formation was close to or in forest, with watercourses. The environment indicated by the presence of peafowls accords with that suggested by the water chevrotain.

The Mabaget Formation comprises fluvial, lacustrine and volcanic sediments intercalated with several palaeosols. These palaeosols are laterally widespread in the formation and consist of dark brown to black, iron-rich, almost lateritic horizons with abundant ferruginous pisoliths. At present, this kind of deposit forms in humid climates, and is notable for the lack of carbonate nodules. Palaeosols in the underlying Kaparaina Basalt Formation are also lateritic in affinities. In the latter formation there are at least four thick red palaeosols devoid of carbonate nodules intercalated between lava flows. The evidence suggests that the climate was extremely humid at the time of deposition of both the Kaparaina and Mabaget Formations and that dry seasons, if they occurred, did not result in the formation of carbonate nodules. Evidence from the underlying Lukeino Formation suggests the same thing [29] and the even older Mpesida Beds (6.3 Ma) are known for the fossil forests [15] with abundant, 20–30-m-tall trees, some of which possess buttress roots and others of which were draped in large lianas. All this evidence suggests that the Tugen Hills region was considerably more humid and far less seasonal from 6.3 to 4.5 Ma than it is today, and that it was probably covered in tropical rain forest. Compatible results have been obtained for the same stratigraphic sequence using stable isotopes of herbivore enamel, soil carbonate nodules and palaeoflora [14] as well as ecomorphological and stable isotope studies of suid fossils [1]. The available data suggests that the Tugen Hills region was heavily forested through the Late Miocene and basal Pliocene, but this does not exclude the possibility of localised patches of grassland existing within the basin, as occurs for example, in present day Western Uganda and Eastern Congo (Albert-Edward Rift) [31].

The rodent fauna from the Mabaget Formation (Mein, pers. comm.) based on several hundred specimens, has no representation of gerbillids, which are common in the older Lukeino Formation, and are abundant today in woodland, savannah and steppe settings. The apparent absence of gerbillids in the Mabaget Formation accords with the forested palaeoenvironment deduced from other evidence.

For the past 40 years and more, the prevailing hypothesis about hominid origins has been that the family evolved in open country from a quadrupedal precursor, often described as a knuckle-walker [32], [34], [47] and [54]. The East Side Story of Coppens [5] is a variant of the ‘savannah hypothesis’, with chronological, palaeoecological, geographic and geological details added, but without a specified ancestral taxon. The central theme of Coppens’ hypothesis is that opening up of the countryside due to rifting and domal uplift, which caused the aridification of East Africa, led to the emergence of Hominidae at the end of the Miocene, some 8–7 Ma ago. In this and other versions of the ‘savannah hypothesis’, the emergence of the family is visualised as occurring in open country.

However, it was long ago suggested that 3 Ma ‘Lucy’ (Australopithecus antiquus seu afarensis) from Ethiopia, inhabited montane forest rather than savannah [3] and recent discoveries of Late Miocene and Early Pliocene hominids reveal that all the sites older than 4 Ma at which they have been found, represent well-wooded to forested palaeohabitats. Orrorin tugenensis, a 6–5.7 Ma bipedal hominid, is associated with abundant colobine monkey fossils, suggesting an abundance of trees in the basin [29] and [43]. The same applies to the Western Margin sites in the Middle Awash (5.7–5.2 Ma) [11] and the 4.2 Ma site of Aramis, Ethiopia [48], [49] and [53] as well as the 4.5–5.3 Ma sites in the Mabaget Formation, which have yielded Ardipithecus ramidus (or A. praegens [9]). The Lothagam hominid mandible (Apak Member), which was for many years thought to be Late Miocene in age [24] has recently been redated [20] and is more likely to be Early Pliocene in age (older than 4.2 Ma and younger than 5 Ma). It appears to be associated with well-vegetated conditions with some open country habitats nearby, while two slightly older specimens (Upper Nawata Member: 6.5–5 Ma) are also associated with well-vegetated palaeoenvironments [17] and [18].

It now seems increasingly unlikely that protohominids ventured into the savannah as quadrupeds, then went through a knuckle-walking stage before becoming bipedal, a conclusion already suggested nearly four decades ago by Tuttle [45], subsequently by Senut [37], [38], [39] and [40] and re-iterated more recently by Senut [42] and [43], Tuttle [46] and Clarke [4].

Williamson [51] and [52] proposed, on the basis of the freshwater snail genus Potadoma found in Late Pliocene to Early Pleistocene deposits of West Turkana, that there was an extension of the tropical rainforest to the Turkana Basin, Kenya. Today, this snail is confined to rivers in heavily forested areas. On the basis of faunal studies, Pickford [26] concluded that during the Late Miocene and Early Pliocene the Tugen Hills region was considerably more humid and better vegetated than it is today. The recent discovery of tragulids in the Early Pliocene deposits of the region confirms that rainforest was present in Kenya from about 5.3 Ma to 4.5 Ma, while forest and dense woodland has been inferred even earlier in the basin, in the Lukeino Formation (ca 6 Ma) [29] and the Mpesida Beds (6.3 Ma) [14]. This evidence is in general agreement with the stable isotope studies carried out by Kingston [14] and [16] who analysed carbonate nodules from palaeosols in the Tugen Hills succession. Forest palaeosols usually do not develop carbonate nodules, and thus these palaeosols could not be included in the previous studies. By omitting the humid palaeosols, the sampling programme biased the results, highlighting the C4 grassland contribution at the expense of the more prevalent humid ones, but even so, it is evident that there was forest in the region [14]. Basically similar conclusions were reached by Bishop et al. [1], on the basis of the study of stable isotopes of suid enamel and suid ecomorphology, but the authors pointed out that some of their results suggested ‘either seasonality, some aridity, low moisture or a combination of these conditions’. For the base of the Mabaget Formation, it is more likely that seasonality was minimal, even if other horizons in the succession contain evidence of more marked seasonality.

The faunal studies and examination of palaeosols in the Lukeino, Kaparaina and Mabaget Formations indicate that throughout the period represented by these formations (6–4.5 Ma), the Tugen Hills area was close to or within tropical forest, and that signs of semi-arid to arid conditions are rare to non-existent.

The discovery of fossil tragulid remains attributed to the extant species Hyemoschus aquaticus and of a peafowl, Pavo sp., in the 4.5–5.3 Ma Mabaget Formation, Tugen Hills, Kenya, indicates that the region was endowed with tropical rainforest at the time of deposition and that seasonality was minimal. This is important for understanding the early stages of hominid evolution, because it now appears that Late Miocene and Early Pliocene hominids inhabited forested areas, and that the family had perhaps not yet ventured into open country. This inference has important implications for the origins of bipedalism, suggesting not only that the traditional ‘savannah hypothesis’ of human origins is invalid, but also that the idea that protohominids went through a knuckle-walking stage [35] prior to evolving bipedalism is obsolete. It is more likely that early bipeds evolved in wooded to forested environments and that for several million years they combined bipedalism with arboreal climbing [29] and [43]. This combination of locomotor repertoires persisted in Middle and Late Pliocene australopithecines [38], [39] and [40], but climbing adaptations were eventually reduced and bipedal adaptations accentuated in the Praeanthropus and Homo lineages.