Carnivore Diet

The Hata Member of the Bouri Formation is defined for Pliocene sedimentary outcrops in the Middle Awash Valley, Ethiopia. The Hata Member is dated to 2.5 million years ago and has produced a new species of Australopithecus and hominid postcranial remains not currently assigned to species. Spatially associated zooarchaeological remains show that hominids acquired meat and marrow by 2.5 million years ago and that they are the near contemporary of Oldowan artifacts at nearby Gona. The combined evidence suggests that behavioral changes associated with lithic technology and enhanced carnivory may have been coincident with the emergence of the Homo clade from Australopithecus afarensis in eastern Africa

We collected 400 vertebrate fossil specimens from the Hata Member (Table 1). Almost all of these come from within 3 m of the MOVT; most were found immediately above this unit. This assemblage largely reflects a mixture of grazers and water-dependent forms, which is broadly typical of later hominid-bearing Plio-Pleistocene occurrences and consistent with the sedimentological interpretation of the deposits as primarily lake marginal. Alcelaphine bovids are abundant and diverse. All indicators point to a broad featureless margin of a shallow freshwater lake. Minor changes in lake level, which were brought about by fluctuating water input, would probably have maintained broad grassy plains leading to the water’s edge. As discussed below, hominids were active on this landscape.

The bone modifications at these two excavated localities and at other localities from the same stratigraphic horizon across .2 km of outcrop demonstrate that stone tool–wielding hominids were active on the lake margin at 2.5 Ma. The bone modifications indicate that large mammals were disarticulated and defleshed and that their long bones were broken open, presumably to extract marrow, a new food in hominid evolution with important physiological, evolutionary, and behavioral effects. Similar patterns of marrow acquisition have been reported for younger sites such as Koobi Fora and Olduvai Gorge (12).

The situation on the Hata lake margin was even more difficult for early toolmakers. Here, raw materials were not readily available because of the absence of streams capable of carrying even pebbles. There were no nearby basalt outcrops. The absence of locally available raw material on the flat featureless Hata lake margin may explain the absence of lithic artifact concentrations. The bone modification evidence demonstrates that early hominids were transporting stone to the site of carcass manipulation. The paucity of evidence for lithic artifact abandonment at these sites suggests that these early hominids may have been curating their tools (cores and flakes) with foresight for subsequent use. Indications of tool curation by later hominids have been found at the more recent Pleistocene sites of Koobi Fora [Karari escarpment versus Ileret (13)] and Swartkrans [polished bone tools in a single repository (16)]. Additional research into the Hata beds may allow a determination of whether the butchery is related to hunting or scavenging. The Bouri discoveries show that the earliest Pliocene archaeological assemblages and their landscape patterning are strongly conditioned by the availability of raw material. They demonstrate that a major function of the earliest known tools was meat and marrow processing of large carcasses. Finally, they extend this pattern of butchery by hominids well into the Pliocene.

The places where animals have been killed or at least butchered by our ancestors represent obviously the best expression of the relation between man and his prey. Isaac (Isaac, 1976; Isaac & Cradeq, 1981), referring to African deposits of Lower Palaeolithic age, defines a simple kind of such sites as containing the skeleton of a single, large animal, associated with lithic artefacts (his type B sites): they represent a unique episode. However such accumulations seem to be very rare: in fact near the carcass of the huge beast almost always other generally much more fragmentary remains of other animals are found. These can represent "background" material without direct relation with hominid activity, but we cannot be sure of this. Evidently, Isaac's definition does not cover the effective variability of all Palaeolithic and Mesolithic kill and/or butchering sites. Therefore, I have tried, in my tesi di laurea, to develop a typolo gy of the possible kinds of bone concentrations reflecting man's animal procurement behaviour. For this aim, I drew information from various authors discussing the topic (Binford, 1984; Clark & Haynes, 1970; Crader, 1983; Meignen & Texieq, 1956) and read a selected number of papers dealing directly, or indirectly through discussions or summaries, with some 30 sites, my reading assignment depending to some extent on the accessibility of the papers included. I am aware that my sampling of sites is limited and perhaps biased and that the evidence as presented by the various authors is often equivocal, but I hope that my attempt will stimulate the development of a site typology which could be a useful tool for classification and research. 

2. Suggested site typology:

a. Butchering sifes: places with animal natural deaths, later utilised by mary such as sites FLK N Lev. 6 (fig.1) and FLK N Deinotherium at Olduvai (Crader, 1983; Leakey, 1971), and site HAS (fig.2) at Koobi Fora (Cradeq, 1983). 

b. Killing and butchering sites 1: a single animal carcass representing a unique hunting episode. This kind of accumulation is similar to Isaac's type B sites. An American example is Pleasant Lake (Fishea 1984; fig. 3). 

c. Killing and butchering sites 2: extensive disarticulation and dispersion of the bones of a few big animals at the most associated with a comparatively small number of stone artefacts. Examples are Windhoek (Clark & Haynes, 1970) and perhaps Mwanganda (Clark & Haynes,1970). 

d. Hunting losses: animals killed but not utilised by man; High Furlong (Hallam et al., 1973) would be an example. 

e. Hunting stations: dense distributions of osseous remains reflecting the reutilization of the locality for a lorig period, often on a seasonal base. Examples of such palimpsests of archaeological remains could be Mauran (Farizy & David, in press; Girard-Farizy & Leclerc, 1981), Stellmoor (Rus! 1937) and La Cotte de Saint-Brelade, lev. 3 and 6 (Scott, 1980; ftg. q. A subtype of hunting stations could be represented by American mass kills, as for example the Casper Site (Frison, 1974). In these sites, not examined here, animals are normally killed with game drive techniques. 

f. Hunting stops: they can be relatively simple or quite complex: sometimes the hunters seek shelter behind a high rock and light a small fire as suggested by Binford (Binford, 1981). An example could be Phase IVA of the Grotte de l'Hortus (de Lumley, 1971). 

g. Sighting sites: they would be characterised by modest bone accumulations in locations with a panoramic position and allowing to detect game and its movements easily. Examples are the Mesolithic sites described by Bagolini and Dalmeri (Bagolini & Dalmeri,

3.1. Lower Palaeolithic Scavenging: exploitation of the carcasses of big animals that died for natural causes; they are often found near lakes or swamps, as the elephant and maybe the Deinotherium at Olduvai (Leakey, 1971), the hippopotamus of Koobi Fora (Isaac, 7976) and the elephants of Kathu Pan (Klein, 1988), Namib IV (Kleirr, 1988) and Mwanganda's Village (Clark & Haynes, 1970). 

Hunting: scanty traces of hunters' action are encountered. At Olorgesailie, occasional killing of some baboons with a head blow seems to have occurred (Shipman, Bosler & Davis, 1981). At Torralba and Ambrona, people may have killed elephants using wooden spears (fragments of wooden artefacts are present) and big stones (Allain" 1952). At Lehringen (Movius, 1950), hominids killed an Elephas antiquus with a wooden spear discovered in the site (see also Weber, this volume). 

Planning: very limited or absent. The exploitation of animals would have been occasional and opportunistic with short and limited occupation of sites by small groups, as at Olduvai (Cradeq, 1983), Koobi Fora (Cradeq, 1983), etc. 

Food transport: Acheulean people are said to have carried away the most useful and meaty parts of animal carcasses at Torralba (Freemary 1975), Ambrona (Freem an, 1975), Elandsfontein (Klein, 1988), etc. In earlier times, people apparently consumed the meat on the find spot. Specialised activities: at the already cited sites of Torralba, Ambrona and at Mwanganda distinct associations between certain bones and tools would occur: they may represent specialised activity areas. 

Butchering tools: hand-axes and hachereaux are sometimes associated with big animals at Olorgesailie (Shipman, Bosler & Davis, 1981), Elandsfontein (Klein, 1988), Kathu Pan (Klein, 1988), Namib IV (Klein, 1988) etc., suggesting that they were used for butchering.

On a high grassy plateau in Algeria, just 100 kilometers from the Mediterranean Sea, early human ancestors butchered extinct horses, antelopes, and other animals with primitive stone tools 2 million to 2.4 million years ago. The dates, reported today, push back the age of the oldest tools in North Africa by as much as a half a million years and provide new insight into how these protohumans spread across the continent.

For decades, east Africa has been considered the birthplace of our genus Homo, and the epicenter of early toolmaking for almost 1 million years. The oldest known Homo fossils date back 2.8 million years in Ethiopia. Nearby, just 200,000 years later, scientists have found simple tools, such as thumb-size stone flakes, and fist-size cores from which such flakes were struck, in the nearby Rift Valley of Ethiopia.

After 25 years of excavations at the Ain Hanech complex—a dry ravine in Algeria—an international team reports the discovery of about 250 primitive tools and 296 bones of animals from a site called Ain Boucherit. About two dozen animal bones have cut marks that show they were skinned, defleshed, or pounded for marrow. Made of limestone and flint, the sharp-edged flakes and round cores—some the size of tennis balls—resemble those found in east Africa. Both represent the earliest known toolkit, the so-called Oldowan technology, named for the site where they were found 80 years ago at Olduvai in Tanzania.

Ain Hanech lacks volcanic minerals, which provide the gold standard for dating sites in eastern Africa. Instead, the researchers used three other dating methods, notably paleomagnetic dating, which detects known reversals in Earth’s magnetic field that are recorded in rock. The tools and cut-marked bones date as far back as 2.4 million years ago, the researchers report today in Science. They also used the identity of large, extinct animals, such as mastodons and ancient horses, to confirm the dates.

The cut-marked bones represent “the oldest substantive evidence for butchery” anywhere, says paleoanthropologist Thomas Plummer of the City University of New York’s Queens College, who was not involved with the study. Although other sites of this age in east Africa have stone tools, the evidence for actual butchery of animals is not as strong, he says.

At Ain Hanech, the dates provide “convincing evidence for stone tools and cut-marked bones at about 2 million years or more,” says geochronologist Warren Sharp of the Berkeley Geochronology Center in California. But he finds the 2.4 million date “less compelling,” because of potential issues with the dating methods.

Whether the tools are 2 million or 2.4 million years old, they suggest toolmakers had spread farther and wider across Africa earlier than previously known. “There must have been a corridor through the Sahara with movement between east Africa and North Africa,” says paleoanthropologist Rick Potts of the Smithsonian Institution’s National Museum of Natural History in Washington, D.C. Alternatively, the new dates suggest hominins in at least two different parts of Africa, separated by 5000 kilometers, were sophisticated enough to independently invent rudimentary stone tools and habitually make them, Potts says.

Either way, the study suggests that by 2 million years ago or so, making stone tools and butchering meat with them was routine for human ancestors in distant corners of the African continent. And this technological revolution may have given them the tools they needed to travel farther and wider across Africa and beyond

The emergence of lithic technology by ∼2.6 million years ago (Ma) is often interpreted as a correlate of increasingly recurrent hominin acquisition and consumption of animal remains. Associated faunal evidence, however, is poorly preserved prior to ∼1.8 Ma, limiting our understanding of early archaeological (Oldowan) hominin carnivory. Here, we detail three large well-preserved zooarchaeological assemblages from Kanjera South, Kenya. The assemblages date to ∼2.0 Ma, pre-dating all previously published archaeofaunas of appreciable size. At Kanjera, there is clear evidence that Oldowan hominins acquired and processed numerous, relatively complete, small ungulate carcasses. Moreover, they had at least occasional access to the fleshed remains of larger, wildebeest-sized animals. The overall record of hominin activities is consistent through the stratified sequence – spanning hundreds to thousands of years – and provides the earliest archaeological evidence of sustained hominin involvement with fleshed animal remains (i.e., persistent carnivory), a foraging adaptation central to many models of hominin evolution.

We report here on the zooarchaeological record of bovid remains. These dominate the assemblages in terms of overall abundances (representing a minimum of 56 individuals), and are amenable to analysis using published protocols and experimental datasets [21]–[30], [56]–[63]. Analytically, we group remains by bed (e.g., ‘KS-1’, ‘KS-3’) rather than by excavation [49]. We further sort specimens by body size class [21], grouping animals into ‘small’ (e.g., Grant’s gazelle, Gazella granti) and ‘medium’ (e.g., Topi, Damaliscus lunatus) sizes. Extinct bovids of intermediate size and weight (e.g., Parmularius sp.) are treated as medium-size animals. Larger bovids (e.g., buffalo, Syncerus caffer) are poorly represented in the assemblages and are not treated in detail here. Following convention, we incorporate taxonomically-unidentifiable long bone fragments in all appropriate analyses.

In our study of bone surface modifications, three investigators (JVF, BLP, and JSO) jointly analyzed specimens, shared observations, and discussed interpretations before providing individual assessments of bone damage [17]. Analysts employed low–power magnification (10×-40×) and strong light sources to identify modifications. They attributed agency (e.g., hominin, carnivore) to modifications only after excluding all possible alternatives (including potential confounds detailed in [32], [64]–[69]).

Values for minimum numbers of skeletal elements (MNE) reflect considerations of animal size and developmental age, extensive refitting efforts, and, for long bones, element identification of shaft portions [21]. High-survival elements (HSE) include the cranium, mandible, humerus, radius, metacarpal, femur, tibia, and metatarsal [61]. Point estimates of Shannon evenness follow published methods [30], [70], whereas interval estimates are constructed using Bayesian models [71].

Bone surface modification frequencies are known to accurately reflect the timing and context of both hominin and carnivore involvement with animal remains. We use them here to assess the identity and sequence of actors and behaviors responsible for forming and modifying the assemblages.

Hominin-modified specimens (i.e., fossil bones bearing cut marks and/or hammerstone percussion damage) are present through the entire KS-1 through KS-3 sequence (Table 2 and Table S1). These specimens provide unambiguous evidence of hominin processing of bovid remains (Figure 2), and indicate a functional relationship between artifactual and faunal materials. When considering the anatomical placement of cut marks, we report bone damage consistent with both defleshing and disarticulation activities [17]. Frequencies of cut-marked limb specimens range from 1.9% to 6.3% in summed (i.e., total bed) assemblages, with similar frequencies observed irrespective of analyst, bed, or animal body size. The overall uniformity of these results suggests a relatively consistent pattern of carcass exploitation through time (within-analyst test for the homogeneity of cut mark frequencies across beds: homogeneity not falsified, all p-values >0.1).

With respect to the timing of hominin access to these smaller-sized individuals, actualistic studies in a modern East African grassland (the Serengeti) show that small bovid carcasses are, almost without exception, completely consumed by lions and/or hyenas within the first few minutes to hours following death [63]. Given the relative abundance of small bovid carcasses at KJS (Table S3), the relative dearth of carnivore tooth marks on their remains (Table S1), and the inferred rarity of such scavenging opportunities in grassland settings, our results strongly suggest that hominins acquired many of these animals very early in their resource lives (i.e., fairly close to the moment of death). At present, the summed evidence that Oldowan foragers acquired, defleshed, and demarrowed numerous, complete, small bovids throughout the formation of all three assemblages plausibly represents the earliest archaeological record of hominin hunting activities.

The skeletal remains of medium-sized bovids reflect a slightly different taphonomic history. Although specimens from all skeletal regions are represented, cranial remains predominate (Figure 5B). Within each assemblage, skeletal element abundances are positively correlated with bone densities (rs range: 0.401 to 0.666; all p-values <0.10) [59], and HSE abundances are not significantly correlated with either standardized food utility values (rs range: −0457 to −0.241; all p-values >0.20) [62] or within-bone nutrient values (rs range: 0.107 to 0.657; all p-values >0.10) [28], [29]. When considering the sum of surface modification data, Shannon evenness values (range: 0.808 to 0.944), and theoretical considerations of transport behaviors [61], [62], the record from KJS most parsimoniously indicates that Oldowan hominins introduced the partial remains of medium-sized carcasses to the site, with specific foraging behaviors varying with respect to body region (e.g., head versus postcrania) and timing of access to carcasses [63].

The overall taphonomic history of medium-sized postcrania is thus fairly equivalent to that of the smaller-sized carcasses. In both cases, remains are present at abundances that far exceed natural landscape accumulation norms (Table 1), and bone surface modification frequencies and skeletal part analyses indicate that hominins had primary access to soft tissues (Table 2, Figure 3, Figure 4). The evidence is consistent with scenarios in which hominins introduced a relative abundance of fleshed medium-sized postcrania to the site. In contrast to the record of smaller-sized bovids, however, skeletal element representation and element evenness scores suggest an increased measure of hominin selectivity in skeletal part choice and transport decisions when dealing with medium-sized remains (Table S3, Table S8). Long bone elements are fairly numerous relative to axial remains (as measured by %MAU) (Figure 5B, Table S3); and the more proximal limb elements (i.e., humerus, radio-ulna, femur, and tibia) are relatively more abundant than metapodials (Figure 5B, Table S3). This patterning differs substantially from that of the smaller-sized bovids. The latter’s remains are more evenly-distributed across the entire postcranial skeleton (HSE’s+low survival elements [LSE’s]), as well as across the six major long bones (Figure 5A, Table S3), and presumably reflects the introduction of numerous, fairly complete small bovids to the site. At issue here: what strategies did hominins follow when selecting and transporting medium-sized remains?

The record is potentially consistent with two main scenarios. In the first, hominins introduce an abundance of compete (or relatively complete) medium-sized carcasses to the site. This follows a ‘food maximizing’ strategy in which hominins face negligible-to-minor transport constraints and transfer most or all of the edible remains from death sites to KJS [75]. As a result, they treat both small and medium-sized bovids in a relatively similar manner when making carcass transport decisions. Observed differences in skeletal element records on-site (smalls vs. mediums) would then presumably reflect systematic differences in post-depositional carnivore scavenging behaviors. In the second scenario, hominins preferentially transport limb remains from medium-sized carcasses, plus some axial elements whenever possible. This follows a ‘weight minimizing’ strategy in which transport constraints (e.g., the number of available carriers, distance to destination, predation risk, etc.) limit hominins to carrying away only a subset of all edible tissues [75]. In this case, carnivore treatment of skeletal remains on-site would be relatively consistent across size groups [25], and observed differences in the skeletal element record (small vs. medium) would instead predominantly reflect systematic size-based differences in hominin transport practices.

Here, comparisons between size groups are particularly informative. For small bovids, LSE values are not grossly disproportionate to those of HSE’s (Figure 5A, Table S3). In fact, their overall skeletal record corresponds fairly well to expectations for dual-patterned hominin-first assemblages, [22], [25], [27], [29]. Note too that skeletal remains of smaller-sized individuals are usually at far greater risk of destruction than those of medium-sized animals, especially in grassland contexts [43], [63].This makes the latter’s record at KJS all the more interesting. In each of the assemblages, medium-sized bovids are fairly depauperate in postcranial axial remains relative to both head and limb elements (Figure 5B, Table S3). As the smaller-sized bovids are more evenly represented across the skeleton (both with and without considerations of cranial remains), we discount the possibility that hominins introduced a substantial amount of medium-sized postcranial axial elements to the assemblages (or, alternatively, that those remains were somehow introduced ‘naturally’; e.g., via mass death). In short, if an abundance of medium-sized axial remains were originally present on-site in substantial numbers, and they were largely deleted by scavenging carnivores, then the overall skeletal record of smaller-sized bovids should reflect a substantially more biased record (both in terms of head remains relative to postcrania, and HSE’s relative to LSE’s). The alternative, a null hypothesis in which all bovids were originally present on-site as similarly-apportioned carcasses, would require that medium-sized postcrania (LSE’s+HSE’s) were preferentially deleted by carnivores relative to all smaller-sized remains. We argue that this is unlikely (especially for the record of HSE’s), and note that tooth-mark frequencies are relatively similar across the remains of both size groups (Table S1). In turn, we argue that the KJS record provides robust evidence that hominins largely – but certainly not exclusively – followed a ‘weight-minimizing’ strategy at KJS when selecting and transporting remains from fleshed medium-sized carcasses.

The record of medium-sized cranial elements requires a bit more explanation. Specifically, these remains are disproportionately abundant within each of the assemblages (Figure 5B, Table S3). If hominins largely followed a ‘weight-minimizing’ strategy, and solely had access to complete medium-sized carcasses, they would not have preferentially transported crania and mandibles to KJS. The reason is clear: head remains are quite heavy given their tissue yields, and will often be ignored at death sites in favor of a slew of higher-ranked remains [75]. These same arguments hold when discussing medium-sized limb HSE’s. The preponderance of head remains on-site (as well as the paucity of long bone remains) is thus unlikely to reflect either simple utility or density-related phenomena. Instead, the record strongly suggests the purposeful introduction of a fair number of isolated heads to the site by Oldowan foragers.

But why acquire, transport, and process an abundance of medium-sized heads? In living animals, these remains contain a wealth of fatty, calorie-packed, nutrient-rich tissues: a rare and valuable food resource in a grassland setting where alternate high-value foodstuffs (fruits, nuts, etc.) are often unavailable [2], [3], [29], [49], [52], [63], [76]–[78]. Medium-sized heads are also relatively dense and durable elements, and their internal contents are generally inaccessible to all but hyenas and tool-wielding hominins [63], [79], [80]. As a result, they are often seasonally-available as scavengable resources in East African grasslands [63], [76], [79]–[83]. Additionally, bone surface modification studies at KJS clearly demonstrate that hominins accessed internal head contents: several cranial vault and mandibular fragments bear evidence of percussion striae. Considered in sum, the presumed availability of these isolated remains across the landscape, the relative abundance of these remains in the KJS assemblages, and unambiguous material evidence that hominins exploited their contents on-site is most parsimoniously interpreted as reflecting very early archaeological evidence of a distinct hominin scavenging strategy – one that included a strong focus on acquiring and exploiting fatty, nutrient-rich, energy-dense within-head food resources (e.g., brain matter, mandibular nerve and marrow, etc.) [e.g., 24,63,76,82,84–86].

The total abundance of remains on site, (Table 1), the number of animals represented (Table 1), the high taxonomic diversity present [17], [50], [52], the relatively low frequency of tooth-marked specimens (Figure 3, Figure 4, Table S1), and a sedimentological record wholly inconsistent with a fluvial accumulation of remains [49], [52] also combine to suggest that the KJS assemblages are unlikely to represent in situ death or ‘background scatter’ accumulations formed by non-hominin agencies. Similarly, the skeletal element record of medium-sized bovids suggests that they were unlikely to have been present on-site as complete carcasses, an expectation of most ‘kill-site’ and/or landscape accumulation models. When limiting discussion to medium-sized postcrania, the high abundance of limb remains (including many isolated epiphyses) relative to axial elements is also the inverse expectation for landscape assemblages (Figure 5B) [63].

Finally, as with many zooarchaeological assemblages, the KJS skeletal inventories are dominated by numerous unidentifiable long bone shaft fragments. At issue: who or what created these fragments from whole bones? The relative rarity of ‘dry bone’ fractures, coupled with abundant evidence of ‘green bone’ breakage, strongly suggests the involvement of behavioral agents of modification, especially given the inferred low-energy depositional setting at KJS [17], [49]–[52]. Bone surface modifications (e.g., percussion marks and notches; tooth marks and notches) indicative of access to within-bone resources, however, are found at relatively low frequencies in each of the assemblages (Figure 3; Figure 4; Table 2; Table S1; Table S2) [17]. This result is surprising as it is inconsistent with known outcomes of both hominin and carnivore bone breakage practices, where surface modification frequencies are, on average, substantially higher [e.g., 22,23,25,57,58]. A similar pattern of an abundance of shattered but largely unmodified long bone specimens is observed in many other Paleolithic assemblages [31,45,72,73; Table S2], suggesting to us that current bone breakage models may not fully account for all relevant variables. Notably, at KJS there is no evidence that post-depositional sediment compaction and/or bone weathering influenced the bone breakage record [17]. Further experimental research may be required to fully explain these observations.

Ancient humans used complex hunting techniques to ambush and kill antelopes, gazelles, wildebeest and other large animals at least two million years ago. The discovery – made by anthropologist Professor Henry Bunn of Wisconsin University – pushes back the definitive date for the beginning of systematic human hunting by hundreds of thousands of years.

Two million years ago, our human ancestors were small-brained apemen and in the past many scientists have assumed the meat they ate had been gathered from animals that had died from natural causes or had been left behind by lions, leopards and other carnivores.

But Bunn argues that our apemen ancestors, although primitive and fairly puny, were capable of ambushing herds of large animals after carefully selecting individuals for slaughter. The appearance of this skill so early in our evolutionary past has key implications for the development of human intellect.

"We know that humans ate meat two million years ago," said Bunn, who was speaking in Bordeaux at the annual meeting of the European Society for the study of Human Evolution (ESHE). "What was not clear was the source of that meat. However, we have compared the type of prey killed by lions and leopards today with the type of prey selected by humans in those days. This has shown that men and women could not have been taking kill from other animals or eating those that had died of natural causes. They were selecting and killing what they wanted."

That finding has major implications, he added. "Until now the oldest, unambiguous evidence of human hunting has come from a 400,000-year-old site in Germany where horses were clearly being speared and their flesh eaten. We have now pushed that date back to around two million years ago."

The hunting instinct of early humans is a controversial subject. In the first half of the 20th century, many scientists argued that our ancestors' urge to hunt and kill drove us to develop spears and axes and to evolve bigger and bigger brains in order to handle these increasingly complex weapons. Extreme violence is in our nature, it was argued by fossil experts such as Raymond Dart and writers like Robert Ardrey, whose book African Genesis on the subject was particularly influential. By the 80s, the idea had run out of favour, and scientists argued that our larger brains evolved mainly to help us co-operate with each other. We developed language and other skills that helped us maintain complex societies.

"I don't disagree with this scenario," said Bunn. "But it has led us to downplay the hunting abilities of our early ancestors. People have dismissed them as mere scavengers and I don't think that looks right any more."

In his study, Bunn and his colleagues looked at a huge butchery site in the Olduvai Gorge in Tanzania. The carcasses of wildebeest, antelopes and gazelles were brought there by ancient humans, most probably members of the species Homo habilis, more than 1.8 million years ago. The meat was then stripped from the animals' bones and eaten.

"We decided to look at the ages of the animals that had been dragged there," said Benn. "By studying the teeth in the skulls that were left, we could get a very precise indication of what type of meat these early humans were consuming. Were they bringing back creatures that were in their prime or were old or young? Then we compared our results with the kinds of animals killed by lions and leopards."

The results for several species of large antelope Bunn analysed showed that humans preferred only adult animals in their prime, for example. Lions and leopards killed old, young and adults indiscriminately. For small antelope species, the picture was slightly different. Humans preferred only older animals, while lions and leopards had a fancy only for adults in their prime.

"For all the animals we looked at, we found a completely different pattern of meat preference between ancient humans and other carnivores, indicating that we were not just scavenging from lions and leopards and taking their leftovers. We were picking what we wanted and were killing it ourselves."

Bunn believes these early humans probably sat in trees and waited until herds of antelopes or gazelles passed below, then speared them at point-blank range. This skill, developed far earlier than suspected, was to have profound implications. Once our species got a taste for meat, it was provided with a dense, protein-rich source of energy. We no longer needed to invest internal resources on huge digestive tracts that were previously required to process vegetation and fruit, which are more difficult to digest. Freed from that task by meat, the new, energy-rich resources were then diverted inside our bodies and used to fuel our growing brains.

As a result, over the next two million years our crania grew, producing species of humans with increasingly large brains – until this carnivorous predilection produced Homo sapiens.