The African hominoid clade consists of humans, chimpanzees, gorillas, and extinct forms more closely related to them than orangutans. This group is also often referred to as the Homininae or hominines (Fig 1C and D). The terms hominines and African hominoid clade will be used interchangeably in this paper. Two different hypotheses have been proposed to explain the geographic origins of this group. These hypotheses can be summarized as follows:

H₁: Modern African hominoids arose in Africa.

This hypothesis proposes that stem catarrhines and hominoids arose in Africa. Some lineages then migrated into Asia giving rise to modern gibbons, orangutans, and Eurasian Miocene apes. The stem African hominine lineage remained in Africa, eventually diversifying into gorillas and the chimpanzee-human lineage, which in turn diverged into the chimpanzee and human lineages. Thus, some crown hominoids persisted in Africa throughout the Miocene, and all modern hominoids have an ultimate African origin.

H₂: Modern African hominoids arose in Eurasia.

This hypothesis also suggests that the stem catarrhines originated in Africa. The common ancestor of the extant large hominoids migrated to Eurasia approximately 17 Ma. This lineage then diversified into the modern and fossil Eurasian apes, and all remaining African Miocene hominoids went extinct. The common ancestor of the Homininae arose in Eurasia in the earliest Late Miocene and migrated back to Africa between 6 and 9 Ma ago, diversifying rapidly into the modern African forms including hominins. Thus, in this hypothesis, hominines have a Eurasian origin.

The first hypothesis (H₁) has been the traditional view for many years (e.g., [4]). Once it was known that the earliest catarrhines and hominoids were found in Africa, it seemed logical that the modern African apes also originated there, and that only the lineages leading to Eurasian hominoids migrated out of Africa (Fig. 1B). Currently accepted molecular divergence dates for the apes suggest that the Pongo-hominine split is at 14 Ma, while gorillas diverge at approximately 8 Ma, and hominins and chimpanzees diverge at around 6.5 Ma [65]. Thus according to the African origin hypothesis, we should expect to find the stem hominine ancestor living in Africa between 14 and 8 Ma. It has been proposed that this ancestor would likely be chimpanzee or gorilla like, in that it would have eaten ripe fruits and lived in tropical forests [85]. Since tropical forests in Africa during this period would have largely been restricted to Central and Western Africa, we might predict that it is most likely that hominine ancestors would have inhabited these areas.

New hominoid discoveries in Eurasia have caused some researchers to support the second hypothesis – that the ancestor of all living hominoids migrated out of Africa into Eurasia, and that the common ancestor of the African apes migrated back into Africa in the Late Miocene. Begun [8], [9], [10], [11] and [12] has been the major advocate of a Eurasian origin for the Homininae. He hypothesizes that Miocene hominoids originated from a thick-enameled common ancestor, such as Afropithecus (17–18 Ma in Kenya) or Heliopithecus (Early Middle Miocene, Saudi Arabia). This lineage migrated to Eurasia, its oldest representative being a molar fragment from Engelswiess Germany, currently dated to 16.5 Ma [11] and [31]. A radiation of Eurasian Middle Miocene hominoids followed, with the ancestor of the hominines eventually migrating back into Africa (Fig. 1A). The implications of this hypothesis are that hominine ancestors should not be found in Africa between 14 and 8 Ma, but only after they return from Eurasia and diverge into gorillas and the hominin-chimpanzee lineage. Under this hypothesis, hominines could not be completely restricted to tropical forest, since they would have needed to expand their range in the early Late Miocene between Eurasia and Africa through non-tropical forest habitats [11]. Consequently, we might expect that they would have a more widespread distribution in Africa than extant apes.

Begun’s reasons for suggesting a Eurasian origin for the Homininae stem mostly from the fact that he, like many researchers, does not recognize any of the few known African Miocene hominoids as good ancestral candidates for modern African apes and humans. Instead, based on cladistic analyses, he believes that Dryopithecus and Ouranopithecus from Europe are most closely related to the ancestor of the hominines [10], [12] and [46] (Fig. 1C). Further, he argues that the absence of fossil remains of likely hominine ancestors in Africa suggests that they were not present in Africa during this period.

Stewart and Disotell [76] have examined the role of intercontinental population range extensions in hominoid evolution as a way of evaluating the Eurasian hypothesis for the origin of hominines. They propose that the most likely scenario for the origin of the Homininae is one that requires the fewest migration events. Their analysis supports Begun’s hypothesis [11] and [12] since they conclude that an Asian origin is more likely because it requires fewer migration events and is therefore more parsimonious. Stewart and Disotell’s conclusion concerning hominoid migrations is based on presumed relationships between fossil hominoid taxa taken directly from Begun’s work [10] and [12] and as such is not an independent test of his hypothesis. In addition, they assume that many Eurasian fossil hominoids have separate African origins, requiring several migrations of hominoids into Eurasia, but it is unclear why this must be the case. As suggested by Moya-Sola et al. [57], it is plausible that the distribution of Eurasian hominoids can be explained by two extensions into Eurasia – one for the lesser apes, and one for the modern and fossil Asian great apes (Fig. 1B).

Many different factors affect the composition of a fossil assemblage, in addition to the composition of the biota from which the fossil assemblage derives. Factors that can affect the probability of finding hominoids at a given site include: 1.

Habitat sampling. Although Miocene hominoids are generally no longer thought to be suspensory arborealists like modern apes, they are still interpreted as maintaining an arboreal lifestyle throughout the Middle and Late Miocene [70]. Some Miocene hominoids such as Griphopithecus and Kenyapithecus have been interpreted as inhabiting non-tropical woodland habitats [2] and [86]. These forms are not closely related to the modern African apes and consequently should not be used to reconstruct the chimpanzee and gorilla ancestral morphotype. Modern hominoids are restricted to tropical forests, and it is likely that this is the case for the last common ancestor of chimpanzees and gorillas as well.

Chimpanzees and gorillas have extremely similar morphologies, and have been suggested to be roughly scaled versions of the same animal [30]. If this is the case, then it seems likely that the common ancestor of chimpanzees and gorillas would have been very similar to the modern forms and, as such, would have inhabited an analogous tropical forest environment and relied heavily on ripe fruits. The location and extent of tropical forest in Africa has varied through time, however there has always been tropical forest habitat in Africa that would be suitable for large bodied apes. Consequently, only fossil localities which sample forested environments are expected to yield hominoid remains.

With these factors in mind, a literature review was conducted for the African localities listed by Begun [9]. For each locality, data was collected for the following parameters: 1.

Species diversity – the faunal list of all mammalian taxa reported at the locality.

Once data was collected for these parameters, sites were evaluated to determine whether it is likely that hominoids would be found at the fossil site if they had been present in the living community. The prediction is that sites that contain fossil hominoids should have relatively large sample sizes and should represent forested environments. Fossil sites that do not have these characteristics should not be expected to contain fossil hominoids. If there are several sites that meet these criteria, but do not contain fossil hominoids, then this might be good evidence that hominoids were absent from Africa during this period. On the other hand, if none of the surveyed sites meet these criteria, then it is premature to assume that hominoids are indeed absent.

Often, complete information was not available for all the data categories for a given site. In particular, information on total sample sizes and the abundance of each fossil taxon at a locality are often not reported in the literature. Also, many research reports do not include the numbers of unidentifiable specimens that were recovered from the site. The total number of specimens is reported whenever it is known, but in many cases I have estimated the total sample size, dividing sites into categories of small (n < 100), medium (100 < n < 1000), and large (n > 1000). Abundance data for specific taxa and for unidentifiable material were not estimated, and are often missing from the data set. In addition, some fossil sites examined consist of a number of separate fossil localities that often sample different time periods or environments. Reported sample sizes are for the entire site because it was generally difficult to get abundance and species data for individual localities.

An additional issue is that taxonomic identifications for many of the localities excavated in the earlier part of the twentieth century are often out of date. This was not a major problem because identifications to the genus level were adequate for the purposes of this project. Faunal lists are as thorough as possible, but it seems likely that at many sites there are additional taxa not described in the literature – particularly micromammals.

After reviewing the available information for the African localities between 5 and 12 Ma old listed by Begun [9], it is apparent that the African Late Miocene fossil record is indeed rather poor (Table 1). First, known fossil localities are unevenly distributed in space (Fig. 2). Of the 25 fossil sites surveyed, most sites are from East (n = 11) and North (n = 13) Africa, with only one site from Southern Africa, and none from Central or Western Africa. Late Miocene deposits in these regions are probably rare, but until more palaeontological exploration is done in these under-sampled areas the African Late Miocene fossil record will be poorly sampled geographically. Most of the known Late Miocene localities in Africa fall outside of the likely distribution of tropical rainforest during the Miocene.

The temporal distribution of African Miocene fossil localities is also uneven (Fig. 3). Most sites date to between 5 and 7 Ma, with the period between 7 and 9 Ma being particularly poorly sampled. This trend is even more exaggerated when one looks at sample sizes of sites through time. The sites with the largest sample sizes are almost always in the 5 to 7 Ma range, such as Lothagam, the Lukeino Formation, and Sahabi. Newly reported fossil finds from the Late Miocene of Chad also fall in this interval [20]. Conversely, only a few hundred specimens are reported for the entire two million year period between 7 and 9 Ma ago. Uneven sampling suggests that much taxonomic diversity is being missed, particularly in poorly sampled time periods. Given such small sample sizes, it is not surprising that fossil apes have not been found in this critical period.

Although hominoid remains are extremely rare in Africa between 5 and 12 Ma ago, they are present. Three of the fossil sites have hominin remains – Lothagam, the Lukeino Formation, and the Middle Awash deposits [47], [73] and [74], in addition to recently reported hominin finds from Chad [20]. An additional three localities have ape remains – the Ngorora Formation in the Tugen Hills (3 specimens), Nkondo, and Samburu [34], [37] and [60]. Fossil hominins and apes are never found at the same locality, reinforcing the hypothesis that apes would have remained in closed forest, while early hominins would have moved into woodland environments.

Sites that contain either apes or hominins have larger than average sample sizes. The site with the smallest sample size that contains a hominoid is Nkondo in Uganda with a total of 832 fossils collected [62] and [63]. Of the nineteen sites that contain neither hominins nor apes, twelve have significantly fewer than 500 collected specimens (Table 1), and therefore have a very low probability of sampling fossil hominoids even if they had been present in the biotic community because the sample size is too small to sample rare taxa. This leaves seven Late Miocene African sites that are large enough to sample rare taxa, and would be the best candidates for containing hominoid fossils. These seven sites are Sahabi, Bled Douarah, Nchorora, the Mpesida Beds in the Baringo Basin, the Sinda Basin localities, Manonga Valley, and Langebaanweg.

Sahabi in Libya and Bled Douarah in Tunisia are both large North African fossil sites. Sahabi has an extremely diverse fauna that contains several species of carnivores and Old World Monkeys [17]. Mammalian species of a wide range of body sizes are represented. The environment of Sahabi is reconstructed as representing a mosaic of mixed-woodland and open-country habitats [15]. Thus, the absence of fossil apes in the Sahabi localities is likely due to the fact that this site samples a non-forested environment unsuitable for apes. Bled Douarah is significantly older than Sahabi (9.5–12.5 Ma), but in other respects is quite similar. Several thousand specimens have been collected and there is a relatively diverse fauna [51] and [68]. This fauna includes early equids and hypsodont bovids [67], suggesting that the environment is relatively open and again is unsuitable for hominoids ancestral to chimpanzees and gorillas.

The Nchorora Formation in Ethiopia is dated to 10.7–10.5 Ma [79]. There is no published information on the number of fossils collected from the site. However, the fauna is only moderately diverse, and fossil remains are heavily fragmented [79]. The faunal list includes rodents, a felid, early equids, suids, and bovids. The environment is reconstructed as representing a lakeside setting, with wooded and more open country environments nearby [39] and [40]. This environment is probably not suitable for hominoids during this period.

The Mpesida Beds in the Tugen Hills of Kenya are dated to 7–6.2 Ma. Outcrops of these beds occur in distinct, unconnected lenses of different ages [35] that probably sample different environments. The number of fossils collected from the Mpesida Beds is not published, but the sample size is likely to be relatively small as the mammalian fauna recovered from the site is not very diverse. The fauna includes colobines, proboscideans, equids, suids, giraffids, and bovids, and is mostly large-bodied. Fossil plants have been recovered from several localities. One locality contains fossil wood indicating the presence of tropical lowland rain forest at 6.3 Ma, while another suggests more seasonal woodlands [44]. The tropical lowland forest habitat could be good ape habitat, but the number of fossils that actually come from this habitat is probably quite small, suggesting that apes would not be sampled even if they were present in the living community.

The Sinda Basin localities are found in the Western Rift Valley in Eastern Uganda [52] and [87]. The localities are not well dated, but biostratigraphy indicates the fossils are Late Miocene in age. Over 500 fossils have been collected, of which only 52 could be identified as mammalian, and only 27 were assigned a more specific taxonomic designation. The fossil assemblage appears to have been subject to a rather rigorous set of taphonomic processes, with only tooth fragments of very large taxa preserved. The Sinda Basin fossiliferous localities have been interpreted as representing a humid tropical forest fauna [87], likely suitable for apes. But since preservation at these localities is biased in favor of large taxa, and the number of fossils collected is relatively small, we would not expect to find hominoid remains at this site.

The Ibole member of the Wembere–Manonga Formation in the Manonga Valley, Tanzania is dated to 5.0–5.5 Ma by biostratigraphic correlation with other East African sites [27]. The Ibole Member faunas are not very diverse and are dominated by large bodied animals, so there may be a size bias in preservation. Fossil and sedimentological evidence suggests that the Ibole deposits represent a lakeside setting. Proboscideans and suids dominate the fauna and suggest that closed habitats surrounded the lake [28]. On the other hand, bovids are also relatively common, representing 25% of the total fauna, and suggest the presence of more open habitats in the area. The four species of rodents present are considered to be characteristic of open or semi-aquatic environments [27]. In addition, the Ibole Member fauna contains no primates and no arboreal taxa. The lack of small arboreal taxa suggests that the Ibole faunas may not represent a closed forest habitat and would therefore not be suitable ape habitat, although it may be suitable for hominins [20].

Most fossils from Langebaanweg in South Africa have come from the ‘E’ quarry locality and the fauna has been roughly dated to 5 Ma by biostratigraphy. Thousands of specimens have been removed from this locality from several fossil horizons in the Varswater Formation [33]. The fauna is extremely diverse with a large range of body sizes represented and many carnivore species. The only primate remains known are two cercopithecid teeth. The presence of rhinocerotids, equids, and several species of hypsodont bovids suggest a relatively open environment [26]. Hendy [33] concludes that the Langebaanweg fauna represents a temperate environment in transition from woodland vegetation into a more open grassland environment with strong seasonal rainfall – a habitat unsuitable for Late Miocene apes ancestral to chimpanzees and gorillas.

Thus none of the Late Miocene African sites surveyed would necessarily be expected to contain hominoids ancestral to modern African apes. Sites that do contain fossil apes share the following features: (1) they sample forested environments, (2) have large sample sizes with at least 500 specimens, and often considerably more, (3) have a high taxonomic diversity, (4) and a high quality of preservation. Few sites combine all of these characteristics, and those that do contain ape remains (Ngorora, Samburu, and Nkondo). None of the other fossil sites examined have all these features (Table 1).

In summary, many Late Miocene African sites have extremely small sample sizes, indicating that rare faunal elements would not be preserved. Others have obvious sampling biases, containing only fragments of very large mammals, or consisting solely of micromammal teeth, and others probably did not sample suitable habitat for apes. Thus it is not unexpected that they do not contain fossil apes and these sites should not be used as evidence to suggest that hominoids were not present in Africa at this time. I conclude therefore that biases in preservation and inadequate exploration of the African continent lead to the poor Middle and Late Miocene hominoid record in Africa, not an actual absence of crown hominoids.

Because the question of where hominines originated is a controversial one, temporal gaps in the record and fragmentary fossils allow much room for interpretation, and result in alternative hypotheses concerning the evolutionary history of Miocene hominoids. I have reviewed some of the evidence associated with Begun’s [9] hypothesis that the Homininae have a Eurasian origin, most particularly the claim that the fossil record is sufficient to conclude that hominoids do not exist in Africa throughout much of the Miocene.

Results of my analysis show that the Late Miocene fossil record in Africa is relatively poor, particularly when compared with the Eurasian fossil record from this period. There are numerous palaeontological sites, but many of these sites contain few fossils and are biased towards either only very large or very small taxa. Four of the twenty-five sites examined contained only micromammals (rodents and insectivores), while five contained only large mammals (Table 1). It is unlikely that hominoids would be found in sites with such preservation patterns.

Sample sizes at most of these African Miocene sites tend to be quite small. Specific data on the number of specimens recovered from a site is frequently not available, but approximate estimates of sample size can be made for most localities. Forty-four percent of the 25 localities examined had fewer than 100 specimens, and none of these contained hominoids. At the three sites in which ape fossils were recovered, sample sizes tend to be significantly larger, with at least 500 specimens recovered (Table 1). Hominoid remains formed less than 0.5% of the total fossil assemblage at each of these sites. This figure is similar to hominoid sites in Eurasia. For example, even though over 300 Sivapithecus fossils have been found in the Siwaliks of Pakistan, they represent only 0.75% of the fossil assemblage collected [7] and [43]. This is presumably due to the fact that hominoids are relatively rare components of a living faunal assemblage because of their low population densities and long life histories [42]. Consequently, a site with less than a hundred fossils has a very low probability of containing a hominoid fossil – even if hominoids were present in the living community. One good indicator that sample size and diversity are high enough to sample rare taxa is when the faunal list also contains several species of other rare taxa such as carnivores.

Some sites do have large sample sizes but still do not contain fossil apes. This can be explained as the result of unsuitable environmental sampling, as these sites typically represent more open country or savannah environments. As discussed previously, it is unlikely that the immediate ancestors of chimpanzees, humans, and gorillas would have lived in such open environments. There is evidence to suggest that Miocene hominoids occupied a broader ecological niche than modern apes, however still lived in largely forested environments [2], [3] and [80]. In general, it would seem that although some Miocene hominoids did move into ‘more open’ environments, these habitats were still forested, but were more seasonally variable than the rainforest habitats favored by modern apes [2] and [3].

There is limited evidence to suggest that some African hominoids, such as Kenyapithecus may have moved into savannah-like environments at sites such as Fort Ternan and Kaimogool (e.g., [86]), however this evidence is controversial, and most environmental reconstructions and the available morphological evidence suggest forested environments [3]. Even if Kenyapithecus had occupied non-forested environments, this should not be used as evidence to reconstruct the environment of the immediate common ancestor of the modern African ape clade because Kenyapithecus is not closely related to this group. The principle of parsimony suggests that if gorillas, bonobos, and chimpanzees all inhabit tropical forest environments, then most likely their common ancestor did as well – accepting that hominins clearly represent a derived condition [64], [65] and [85].

The suggestion that hominoids ancestral to hominines lived in Eurasia and not Africa during the Late Miocene is based on the fact that appropriate ancestral candidates have not been found in Africa [9]. Several authors have commented that this assertion is premature because the fossil record in Africa from this period is poor [57], [64] and [81]. The goal of this paper was to examine whether or not the quality of the African Late Miocene record is sufficient to conclude that hominoids were indeed absent during this period.

A literature review of known African early Late Miocene sites demonstrates that these sites are unevenly distributed in space and time (Fig. 2 and Fig. 3), with little information available from Central and Western Africa – the geographical areas most likely to have been home to apes. Even with this large gap in our knowledge, fossil ape remains are present at three different localities in East Africa, demonstrating that they were present in Africa during this time period. This suggests that further palaeontological exploration could uncover new fossil taxa, perhaps including the ancestors of the Homininae.

Many currently known Miocene fossil localities do not contain hominoids. This can be interpreted to mean either that hominoids were not present in these areas [9] and [76], or that hominoid remains have not been preserved at these sites due to small sample sizes, sampling of non-forested environments, or biased preservation. The fact that fossil chimpanzees and gorillas have not been found in later periods when it is certain that they were present should make us skeptical that we have adequately sampled the fossil faunas of the African Miocene.

Ultimately, the only way to know where the ancestor of the Homininae resided is to look for new hominoid sites, particularly in currently unsampled areas. Large portions of Eurasia as well as Africa are unstudied, leading to serious gaps in our knowledge of hominoid history. By identifying new areas for palaeontological exploration, particularly in areas where modern apes live, we may find new fossils that will help resolve current disputes in hominoid phylogeny and palaeobiology.