Sahelanthropus, Orrorin and Ardipithecus

In the late Miocene (up to 5·3 Mya), the African continent became more arid, which resulted in fragmentation of the (sub)tropical forests and the appearance of more open environments⁽Reference Bernor¹⁵¹⁾. The widespread dispersal of some of the earliest hominins such as Sahelanthropus ⁽Reference Brunet, Guy and Pilbeam^65, Reference Lebatard, Bourles and Duringer¹⁵²⁾, and Australopithecus bahrelghazali from Chad, might be explained by the presence of the relatively low-lying humid East–West corridor constituted by the remnants of the Cretaceous Central African and Sudan Rifts between Western and Eastern Africa⁽Reference Bosworth and Morley^153, Reference Joordens¹⁵⁴⁾. The reconstructed environment of Sahelanthropus (about 7 Mya) suggests a mosaic of gallery forest at the edge of a deep, well-oxygenated lake, swampy and vegetated areas, and extensive grasslands⁽Reference Vignaud, Duringer and Mackaye¹⁵⁵⁾. Since there is no indication of carnivore modification or fluvial transport of its bones, Sahelanthropus chadensis probably lived in this area⁽Reference Stewart, Cunnane and Stewart¹⁵⁶⁾. The palaeo-environment of Orrorin (about 6 Mya) was probably characterised by open woodland, with dense stands of trees in the vicinity and possibly fringing the lake margin and/or streams that drained into the lake⁽Reference Pickford and Senut¹⁵⁷⁾. Ardipithecus kadabba (5·6 Mya) remains are associated with wet and closed, grassy woodland and forest habitats around lake or river margins⁽Reference WoldeGabriel, Haile-Selassie and Renne¹⁵⁸⁾. Ardipithicus ramidus (4·4 Mya) lived in or near a groundwater-supported grassy woodland to forest⁽Reference WoldeGabriel, Ambrose and Barboni¹⁵⁹⁾. Additionally, the abundance of fossilised shallow-water aquatic species such as catfish, barbus, cichlidae and crocodiles additionally suggests an episodically present flood-plain environment⁽Reference WoldeGabriel, Ambrose and Barboni¹⁵⁹⁾.

The diet of our closest relatives

Field studies on our closest relatives, the extant apes, show that their preferred food items are primarily fruits and/or leaves and stems from terrestrial forests. Lowland gorillas, for example, derive 57 % of their metabolisable energy (en%) from SCFA derived from colonic fermentation of fibre, 2·5 en% from fat, 24 en% from protein and 16 en% from carbohydrate⁽Reference Popovich, Jenkins and Kendall¹⁷⁶⁾. ‘Fallback foods’ are consumed when preferred foods are unavailable⁽Reference Marshall and Wrangham¹⁷⁷⁾ and are generally composed of herbaceous plants and high-fibre fruits from aquatic and terrestrial environments⁽Reference Stewart, Cunnane and Stewart¹⁵⁶⁾. Like our closest relatives⁽Reference Stewart, Cunnane and Stewart^156, Reference Nishida^178–Reference Wrangham, Cheney and Seyfarth¹⁸¹⁾, hominins might have used foods from the aquatic environment as fallback foods, while, although speculative, this niche might eventually have proven favourable with regard to subsequent encephalisation.

Teeth morphology and dental microwear

Comparative anatomy (Fig. 4) of the hominins might confer some information about these fallback and preferred foods of our ancestors. Dental studies of Sahelanthropus (Fig. 1) describe that the teeth had thick enamel⁽Reference Brunet, Guy and Pilbeam⁶⁵⁾, similar to orangutans, suggesting that it could eat hard and tough foods⁽Reference Vogel, van Woerden and Lucas¹⁸²⁾, such as available from the lakeshore vegetation⁽Reference Stewart, Cunnane and Stewart¹⁵⁶⁾. Ardipithecus ramidus, however, had thin molar enamel and smaller teeth compared with later hominins. This dental morphology is consistent with a partially terrestrial, omnivorous/frugivorous niche⁽Reference Suwa, Kono and Simpson¹⁸³⁾. Studies on cranio-dental changes such as tooth size, tooth shape, enamel structure and jaw biomechanics indicate that Australopithecus and Paranthropus had prominent jaws, relatively flat molar teeth, small incisors and thick enamel, suitable for breaking and crushing small hard, brittle foods such as fruits, nuts and underground storage organs⁽Reference Ungar and Sponheimer¹⁵⁰⁾, but unsuitable for breaking down tough plant foods or tearing meat. Together, this would allow early hominins to eat both hard and soft, abrasive and non-abrasive foods, which suits well for life in a variety of habitats⁽Reference Teaford and Ungar¹⁸⁴⁾.

Fig. 4 Lower jaw of a chimpanzee (Pan troglodytes), Australopithecus africanus and Homo sapiens. Note the somewhat human-like shape of the teeth, but ape-like axis in the jaw of Australopithecus. © Australian Museum.

In addition to studies on teeth morphology, microwear studies are essential. Dental microwear studies analyse tooth-wear, showing evidence for where teeth were actually used for and thus what an animal in reality ate⁽Reference Walker, Hoeck and Perez¹⁸⁵⁾. While adaptive morphology will give important clues about what a species was capable of eating, microwear studies reflect what an animal ate during some point in its lifetime. In these studies, ‘complexity’ is used as an indicator for hard and brittle items, while ‘anisotrophy’ is an indicator for tough foods⁽Reference Ungar, Grine and Teaford¹⁸⁶⁾.

Most primates show either low complexity combined with high anisotrophy, indicative of consumption of tough foods such as leaves, stems and meat, or high complexity with low anisotrophy associated with hard-brittle foods, such as nuts and seeds⁽Reference Ungar and Sponheimer^150, Reference Ungar, Scott and Grine¹⁸⁷⁾. Ardipithecus ramidus's preference for an omnivorous/frugivorous diet was confirmed by microwear studies ⁽Reference Suwa, Kono and Simpson¹⁸³⁾, suggesting a diet of fleshy fruits and soft young leaves⁽Reference Ungar and Sponheimer¹⁵⁰⁾. Conversely, microwear textures of Australopithicus afarensis and anamensis show striations rather than pits (low complexity and low anisotrophy), i.e. patterns similar to those of grass-eating and folivorous monkeys instead of the predicted diets predominated by hard and brittle foods⁽Reference Ungar, Scott and Grine^187, Reference Grine, Ungar and Teaford¹⁸⁸⁾. Australopithecus africanus showed microwear patterns that were more anisotropic, suggestive for consumption of tough leaves, grasses and stems⁽Reference Scott, Ungar and Bergstrom¹⁸⁹⁾. Paranthropus robustus, also know as the ‘Nutcracker Man’, has enormous, flat, thickly enamelled teeth that are combined with a robust cranium, mandible and powerful chewing muscles, suggestive of breaking hard and brittle foods⁽Reference Ungar, Grine and Teaford¹⁸⁶⁾. After microwear analysis of its teeth, however, Ungar et al. ⁽Reference Ungar and Sponheimer^150, Reference Ungar, Grine and Teaford¹⁸⁶⁾ showed that P. robustus had low complexity and anisotrophy; thus Paranthropus might only have consumed mechanically challenging items as fallback foods when preferred foods were unavailable. Similarly, microwear studies support the notion that the diet of P. boisei contained large quantities of low-quality vegetation, rather than hard objects⁽Reference Cerling, Mbua and Kirera¹⁹⁰⁾.

Generally, microwear studies confirm earlier dental topography studies⁽Reference Ungar¹⁹¹⁾, which revealed the incorporation of more fracture-resistant foods, i.e. tougher foods as leaves, woody plants, underground storage organs and animal tissues, in the diet of Australopithecus africanus compared with Australopithecus afarensis and for P. robustus compared with Australopithecus africanus ⁽Reference Ungar, Scott and Grine¹⁸⁷⁾. Dental topographic analysis suggested that successive Homo species emphasised more on tougher and elastic foods, perhaps including meat⁽Reference Ungar¹⁹¹⁾. The latter suggestion is in line with the optimal foraging theory, which states that humans prefer foods with high energy density over those with low energy density⁽Reference Ulijaszek^192, Reference Goldstone, de Hernandez and Beaver¹⁹³⁾.

Microwear studies confirmed that early Homo, such as H. erectus and H. habilis, did not prefer fracture-resistant foods, although some H. erectus specimens showed more small pits than H. habilis members, suggesting that none of the early Homo specialised on very hard-brittle or tough foods, but rather could consume a varied diet⁽Reference Ungar, Grine and Teaford¹⁹⁴⁾. This does not imply that early Homo had very broad diets, but rather that early Homo was adapted to subsist in a range of different environments, providing evolutionary advantage in the climatic fluctuations and the mosaic of habitats in Africa during the late Pliocene⁽Reference Ungar, Grine and Teaford¹⁹⁵⁾. A study on dental microwear of 300 000-year-old H. heidelbergensis teeth from Sima de los Huesos in Spain showed striation patterns that indicated a highly abrasive diet, with substantial dependence on poorly processed plant foods such as roots, stems and seeds⁽Reference Perez-Perez, De Castro and Arsuaga¹⁹⁶⁾. Lalueza et al. ⁽Reference Lalueza, PerezPerez and Turbon¹⁹⁷⁾ compared teeth of very recent hunter–gatherers (Inuit, Fueguians, Bushmen, Aborigines, Andamanese, Indians, Veddahs, Tasmanians, Laps and Hindus) with Middle and Upper Pleistocene fossils. Their results indicate that some Neanderthals resemble carnivorous groups, while archaic H. sapiens show a more abrasive diet, partly dependent on vegetable materials.

Overall, remarkably few studies have related microwear patterns to hominin diet for the period between 1·5 Mya and 50 Kya (PS Ungar, personal communication). Studies in more recent hominins, such as from an Upper Palaeolithic site in the Levant (22 500–23 500 before present; BP) showed a high frequency of long narrow scratches and few small pits, suggesting a tough abrasive diet of aquatic foods rather than a diet with hard foods that needed compressive force⁽Reference Mahoney^198, Reference Mahoney¹⁹⁹⁾. A study of subsequent local hunter–gatherer (12 500–10 250 BP) and farmers (10 250–7500 BP) living in the Levant showed larger dental pits and wider scratches among the farmers compared with the hunter–gatherers, suggesting that the implementation of agriculture led to a more fracture-resistant diet⁽Reference Mahoney¹⁹⁹⁾.

Gut morphology, energy expenditure and muscularity

Gut morphology studies⁽Reference Milton^200, Reference Milton²⁰¹⁾ support the introduction of animal foods, at the expense of vegetable foods, in the diet of early Homo. The dominance of the colon (>45 %) in apes indicates adaptation to a diet rich in bulky plant material, such as plant fibre and woody seeds. In contrast, the proportion of the human gut dominated by the small intestine (>56 %) suggests adaptation to a diet that is highly digestible, indicating a closer structural analogy with carnivores than to folivorous or frugivorous mammals. Importantly, the shorter gut in Homo, as compared with primates, might have had some other advantage. During the evolution from Ardipithecus and Australopithecines to early Homo, the improvement of dietary quality coincided with an increase in height, the size of our brain and its metabolic activity. However, an increase in body size coincides with increased daily energy demands, notably during gestation⁽Reference Gittleman and Thompson²⁰²⁾ and lactation⁽Reference Oftedal²⁰³⁾. It was, for example, calculated that daily energy expenditure for a Homo erectus female is about 66 % higher compared with an Australopithicine female, while being almost 100 % higher in a lactating Homo erectus female compared with a non-lactating, non-pregnant Australopithicine ⁽Reference Aiello and Key²⁰⁴⁾. These high energy demands might have been met by increased female fat reserves⁽Reference Aiello and Wells^205, Reference Cunnane and Crawford²⁰⁶⁾, such as demonstrated by the presence of female steatopygia in some traditional human populations, such as the Khoisan of southern Africa.

Apart from the extra energy need for reproduction and increasing height, the human brain of a modern adult uses 20–25 en% of the total RMR, while this value is 8–9 en% for a primate⁽Reference Leonard, Robertson and Snodgrass²⁰⁷⁾. It has been postulated that the extra energy needs did not derive from a general increase in RMR, but partly from a concomitant reduction of the gastrointestinal tract⁽Reference Aiello and Wheeler²⁰⁸⁾ and a reduction in muscularity⁽Reference Leonard, Robertson and Snodgrass²⁰⁷⁾. These features are combined in the ‘expensive tissue hypothesis’ of Aiello & Wheeler⁽Reference Aiello and Wheeler²⁰⁸⁾ that points at the observation that the mass of the human gastrointestinal tract is only 60 % of that expected for a similar-sized primate and that humans are relatively under-muscled compared with other primates⁽Reference Leonard, Robertson and Snodgrass²⁰⁷⁾. Interestingly, the negative relationship between gut and brain size across anthropoid primates⁽Reference Aiello and Roebroeks²⁰⁹⁾ was confirmed in a study with highly encephalised fish⁽Reference Kaufman, Hladik and Pasquet²¹⁰⁾, whereas it became falsified in mammals⁽Reference Navarrete, van Schaik and Isler¹⁰²⁾. Unfortunately, however, the latter study did not include marine mammals⁽Reference Navarrete, van Schaik and Isler¹⁰²⁾, while it is questionable whether with respect to brain development humans adhere to general mammalian rules⁽Reference Potts¹⁰³⁾. Other adaptations might have saved energy expenditure as well. The short human inter-birth interval⁽Reference Aiello and Key²⁰⁴⁾, compared with inter-birth intervals of 4–8 years in the gorilla, chimpanzee and orangutang⁽Reference Galdikas and Wood²¹¹⁾, reduces the most expensive part of reproduction, i.e. lactation⁽Reference Isler and van Schaik^212, Reference Isler²¹³⁾, while the shift from quadrupedal to bipedal locomotion also reduced daily energy expenditure⁽Reference Pontzer, Raichlen and Sockol²¹⁴⁾. Together, all of these adaptations allow for an increased daily energy expenditure, including the reallocation of energy to the metabolically active brain, but were only possible after Homo included more energy-dense foods into its diet. Brain mass in primates is positively related to dietary quality and inversely to body weight⁽Reference Leonard, Robertson and Snodgrass²⁰⁷⁾. Generally, a shift towards more energy-dense foods includes a shift from primarily carbohydrate-rich vegetables to fat- and protein-rich animal foods. However, it has also been suggested that a shift from the complex carbohydrates in leafy vegetables towards underground storage organs, such as tubers, might have provided easy-to-digest carbohydrates⁽Reference Milton^200, Reference Wrangham, Jones and Laden²¹⁵⁾ to support a larger hominin body mass.

A phenotypic specialisation on non-preferred resources, without compromising the ability to use preferred resources, is also known as Liem's paradox⁽Reference Robinson and Wilson²¹⁶⁾. This paradox is important to keep in mind during attempts to reconstruct the preferred diet from the available evidence. Interestingly, Stewart⁽Reference Stewart, Cunnane and Stewart¹⁵⁶⁾ recently noted that in the case of Paranthropus's phenotype with regard to its fallback food, there is just as much evidence to talk about a ‘Nutcracker Man’ as there is to talk about a ‘Shellcracker Man’. Thus, a phenotypic characterisation needs support from other studies to confirm adaptation to preferred rather than fallback foods. In other words, it should be noted that not all physical characteristics might be taken as unambiguous proof for the preferred diet of early hominins (Liem's paradox). The increasing absolute body size and brain size and the reduction in gut size, however, do indicate a shift from low- to high-dietary-quality foods for more recent hominins.

Evidence from the strontium:calcium ratio

Based upon the principle ‘you are what you eat’⁽Reference Harris²¹⁷⁾ several techniques have been developed to study early hominin diets⁽Reference Sponheimer, Dufour, Hublin and Richards²¹⁸⁾. Trace element studies first started with Sr:Ca ratios, which decrease as an animal moves up the food chain, secondary to the biological discrimination against strontium⁽Reference Sillen and Kavanagh²¹⁹⁾. A first study⁽Reference Sillen²²⁰⁾ suggested that Paranthropus robustus (Fig. 1) had lower Sr:Ca ratios compared with contemporaneous Papio (baboon) and Procavia (hyrax), suggesting that Paranthropus was not an exclusive herbivore. However, a subsequent study showed that two Homo specimens had a higher Sr:Ca ratio than Paranthropus ⁽Reference Sillen, Hall and Armstrong²²¹⁾. Since Homo had been assumed to consume more animal foods than Paranthropus and thus have a lower Sr:Ca ratio, these higher Sr:Ca ratios needed explanation. For example, consumption of specific foods with a high Sr:Ca ratio might have increased the ratio in Homo compared with Paranthropus. Foods in the area of research with elevated Sr:Ca ratios were mainly geophytes. These are notably perennial plants with underground food storage organs, such as roots, bulbs, tubers, corms and rhizomes. Consequently, this discrepancy has been attributed to the consumption of underground storage organs by early Homo ⁽Reference Wrangham, Jones and Laden^215, Reference Sillen, Hall and Armstrong²²¹⁾. The use of these underground storage organs may have become necessary from the start of a first period of aridity about 2·8 Mya, when forest was replaced by drier woodlands, forcing hominins to search for available resources around water margins⁽Reference Conklin-Brittain, Wrangham and Hunt²²²⁾.

Although promising at first, the use of Sr:Ca ratios was found to suffer from several limitations. For example, a problem is that hominin Sr:Ca ratios in fossilised bones alter with time⁽Reference Lee-Thorp and Sponheimer²²³⁾, a process that is known as diagenesis. This problem can be circumvented by the use of tooth enamel, which is less susceptible to diagenesis than bone⁽Reference Lee-Thorp and Sponheimer²²³⁾. Subsequent studies showed that Australopithecus africanus had a higher Sr:Ca ratio in its enamel compared with Paranthropus and contemporaneous browsers, grazers, baboons and carnivores⁽Reference Sponheimer, de Ruiter and Lee-Thorp^224–Reference Balter and Simon²²⁶⁾. The interpretation of these findings, however, remains a subject of debate. For example, a group of brown seaweeds and some aquatic plants discriminate against Ca, resulting in an increase in the Sr:Ca ratio, and likewise in the Sr:Ca ratio of the fish feeding on them⁽Reference Balter and Simon^226, Reference Ophel and Fraser²²⁷⁾. Similarly, leaves from trees have lower Sr:Ca ratios compared with grasses, which becomes subsequently reflected in the Sr:Ca ratios of browsers and grazers, respectively⁽Reference Ambrose and Deniro²²⁸⁾. Thus, it might be argued that the relatively high Sr:Ca ratios in Australopithicines and Homo reflect their consumption of aquatic resources or of animals such as insects or other small animals feeding on grasses. For now, the exploration of especially Sr:Ca ratios as a dietary proxy method has largely been stalled⁽Reference Lee-Thorp, Sponheimer and Passey²²⁹⁾.

Evidence from the barium:calcium ratio

Another trace element ratio that might provide information on the composition of the early hominin diet is the Ba:Ca ratio, particularly when used in a multiple-element analysis with Sr:Ca and Sr:Ba ratios⁽Reference Sponheimer, de Ruiter and Lee-Thorp^224, Reference Sponheimer and Lee-Thorp²³⁰⁾. Combined Ba:Ca and Sr:Ba ratios clearly differentiate grazers from browsers and carnivores in both modern and fossil mammals⁽Reference Sponheimer and Lee-Thorp²³⁰⁾. Hominins have a lower Ba:Ca and higher Sr:Ba ratio compared with grazers and browsers. Paranthropus shows considerable similarity for both ratios with both carnivores and Papionins (baboons), while Australopithecus shows an even higher Sr:Ba ratio. The unusual combination of a high Sr:Ca and low Ba:Ca ratio in hominins (and baboons) has been further observed for some animals such as warthogs and mole rats that make extensive use of underground resources⁽Reference Sponheimer, Lee-Thorp and de Ruiter²³¹⁾. Finally, the high Sr:Ba ratio, as observed in Australopithecus, might derive from the consumption of grass seeds, which have Sr:Ba ratios three to four times higher than grass straw, while consumption of these grasses is also consistent with stable-isotope evidence (see below) showing that Australopithecus derived a substantial part of its diet from C₄ resources. Although not included in any study so far, Sr:Ba ratios in aquatic foods might add to the understanding of the unusual combination of low Ba:Ca and high Sr:Ba ratios.

Evidence from the ¹³C:¹²C ratio

Fractionation studies of carbon isotopes differentiate between different routes of photosynthesis. While most tropical African woody plants from forests (like fruits, leaves, trees, roots, bushes, shrubs and forbs) use the C₃ photosynthetic pathway, some South African⁽Reference Vogel²³²⁾ and East African⁽Reference Cerling, Harris and MacFadden²³³⁾ grasses and sedges use the C₄ photosynthetic pathway. C₄ plants such as sedges (for example, Cyperus papyrus) typically occur in a mosaic of extensive seasonal and perennial shallow freshwater wetlands that can also be found in savanna and ‘bushvelds’ receiving summer rainfall⁽Reference Peters and Vogel²³⁴⁾. The real impact of C₄ plants occurred with their spread into Eastern and Southern Africa during the Pliocene⁽Reference Segalen, Lee-Thorp and Cerling²³⁵⁾. Tissues of plants that utilise the C₄ pathway have a relatively high content of the stable carbon isotope ¹³C (about 1·1 % of carbon), since C₃ plants discriminate more strongly against ¹³CO₂ during photosynthesis. As a result C₃ and C₄ plants have quite different ¹³C:¹²C ratios in their tissues, as have the herbivorous animals that feed on these plants⁽Reference Ambrose and Deniro^228, Reference Vogel²³²⁾ and the carnivores that prey on these herbivores⁽Reference Goldstone, de Hernandez and Beaver^193, Reference Cunnane and Crawford²⁰⁶⁾. Differences are expressed as δ¹³C values (δ¹³C =  ((¹³C:¹²C)_sample − (¹³C:¹²C)_standard) × 1000/(¹³C:¹²C)_standard) in parts per thousand (‰) relative to the ¹³C:¹²C ratio in the reference standard (named Pee Dee Belemnite; PDB), i.e. the carbonate obtained from the fossil of a marine Cretaceous cephalopod (Belemnitella americana) which is highly enriched in ¹³C (¹³C:¹²C ratio = 0·0112372). Consequently, most animals have negative δ¹³C values (Fig. 5). The δ¹³C ranges from − 35 to − 21 ‰ (mean − 26 ‰) in C₃ plants and from − 14 to − 10 ‰ (mean − 12 ‰) in C₄ plants⁽Reference Lee-Thorp and Sponheimer²²⁵⁾. Studies that attempted to reconstruct mammalian food webs indicated that carbon is slightly enriched (1–2 ‰) with each trophic step⁽Reference Goldstone, de Hernandez and Beaver^193, Reference Cunnane and Crawford²⁰⁶⁾. To facilitate comparison of current and historical animals, δ¹³C analyses are predominantly performed in hard tissues, such as bone collagen and enamel (notably apatite), since these constitute the majority of the fossil record. It was shown that enamel mineral is enriched by about 13 ‰ compared with dietary δ¹³C⁽Reference Sponheimer, Lee-Thorp, de Ruiter and Ungar²³⁸⁾, while collagen is enriched by about 5 ‰ compared with dietary δ¹³C⁽Reference Lee-Thorp and Sponheimer²²⁵⁾. Collagen from terrestrial mammal C₃ herbivores shows a value of − 21 ‰ (range − 22 to − 14 ‰), while in collagen of C₄ herbivores this value is − 7 ‰ ( − 12 to − 6 ‰). Their respective carnivores show collagen values of − 19 ‰ ( − 21 to − 14 ‰) and − 5 ‰ ( − 8 to − 2 ‰), respectively⁽Reference Lee-Thorp and Sponheimer²²⁵⁾. Marine phytoplankton, which uses the C₃ pathway, shows an average value of − 22 ‰⁽Reference Kelly²³⁹⁾. Collagen of marine fish shows a range from − 15 ‰ to − 10 ‰, while the values in collagen of reef fish range from − 8 to − 4 ‰⁽Reference Schoeninger and Deniro²⁴⁰⁾. Marine carnivores have, similar to terrestrial carnivores, intermediate values of − 14 ‰, ranging from − 10 ‰ in collagen of sea otters to − 15 ‰ in collagen of the common dolphin⁽Reference Kelly^239, Reference Schoeninger and Deniro²⁴⁰⁾. Thus, also the δ¹³C values of marine carnivores compare well with those of their prey. Finally, δ¹³C values for the muscle of freshwater fish species range from − 24 to − 13 ‰, but considerable variation may exist between different lakes⁽Reference Mbabazi, Makanga and Orach-Meza²⁴¹⁾. Unfortunately, no δ¹³C data are available for terrestrial piscivorous carnivores, but the consistency of the other data suggests that these might be between − 24 and − 13 ‰, i.e. comparable with their prey (Fig. 5).

Fig. 5 Normalised collagen δ¹³C values (mean and range; in per thousand (‰)) in plankton, crustaceans, sea grasses, C₃ and C₄ plants; of marine crustaceans, fish and freshwater fish and their respective carnivores; of terrestrial C₃ and C₄ herbivores and their carnivores; and of human groups in historic and prehistoric times. * Corrected⁽Reference Lee-Thorp, Sponheimer and Passey²²⁹⁾ for collagen ( − 5 ‰). † Corrected⁽Reference Sponheimer, Lee-Thorp, de Ruiter and Ungar²³⁸⁾ for enamel ( − 13 ‰). ‡ Arbitrary range of ±  1 ‰ due to a lack of data. § As predicted from other predator–prey relationships and after correction⁽Reference Kelly²³⁹⁾ for tropic level (+1 ‰). Adapted from Ambrose & Deniro⁽Reference Ambrose and Deniro²²⁸⁾, Sponheimer et al. ⁽Reference Sponheimer, Lee-Thorp and de Ruiter^231, Reference Sponheimer, Loudon and Codron²⁴⁷⁾, Peters & Vogel⁽Reference Peters and Vogel²³⁴⁾, Lee-Thorp et al. ⁽Reference Lee-Thorp, Thackeray and van der Merwe^237, Reference Lee-Thorp, Sponheimer and Luyt²⁴⁴⁾, Kelly⁽Reference Kelly²³⁹⁾, Schoeninger & Deniro⁽Reference Schoeninger and Deniro²⁴⁰⁾, Mbabazi et al. ⁽Reference Mbabazi, Makanga and Orach-Meza²⁴¹⁾, Schoeninger et al. ⁽Reference Schoeninger, Deniro and Tauber^242, Reference Schoeninger, Moore and Sept²⁴⁶⁾, Sponheimer & Lee-Thorp⁽Reference Sponheimer and Lee-Thorp^243, Reference Sponheimer and Lee-Thorp^248, Reference Sponheimer and Lee-Thorp²⁴⁹⁾ and van der Merwe et al. ⁽Reference van der Merwe, Masao and Bamford²⁴⁵⁾.

Consistent with the data above, Schoeninger et al. ⁽Reference Schoeninger, Deniro and Tauber²⁴²⁾ showed that European agriculturalists consuming C₃ grasses had much lower δ¹³C values in bone collagen ( − 21 to − 19 ‰) compared with Mesoamerican agriculturalist consuming C₄ maize ( − 7 to − 5 ‰), while North American and European fisher–gatherers had intermediate values ( − 15 to − 11 ‰). However, bone collagen proved less reliable to study early hominin diets. For that purpose the ¹³C:¹²C ratio is preferably measured in tooth enamel. To compare collagen δ¹³C values with enamel δ¹³C values, an additional (8 ‰) correction has to be made. Similar to the clear distinctions between bone collagen of modern grazers, browsers and their carnivores, tooth enamel data from fossilised fauna from South Africa showed similar differences for Plio-Pleistocene C₃ feeders ( − 11·5 ‰) and C₄ feeders ( − 0·5 ‰), with Australopithecus, Paranthropus and Homo taking intermediate positions ( − 10 to − 4 ‰)⁽Reference Sponheimer, Lee-Thorp and de Ruiter^231, Reference Peters and Vogel^234, Reference Lee-Thorp, Thackeray and van der Merwe^237, Reference Sponheimer and Lee-Thorp²⁴³⁾, which compared well with the values for contemporaneous felids ( − 10 to − 0·5 ‰)⁽Reference Sponheimer, Lee-Thorp and de Ruiter^231, Reference Lee-Thorp, Thackeray and van der Merwe^237, Reference Lee-Thorp, Sponheimer and Luyt²⁴⁴⁾ (Fig. 5). These results clearly demonstrated that a significant proportion of the diets of the early hominins from Swartkrans, Makapansgat and Sterkfontein derived from C₄ resources. Using these data it was calculated that South African Paranthropus derived 14–47 % of its diet from C₄ sources, compared with 5–64 % in Australopithecus and 20–35 % in Homo ⁽Reference van der Merwe, Masao and Bamford²⁴⁵⁾.

A second study in Olduvai showed that the Tanzanian Paranthropus boisei derived 77–81 % and Homo 23–49 % from its diet from C₄ resources⁽Reference van der Merwe, Masao and Bamford²⁴⁵⁾. The low nutritional value of grasses, and microwear studies (see above) render it unlikely that humans were directly eating grass⁽Reference Sponheimer, Lee-Thorp and de Ruiter^231, Reference Sponheimer and Lee-Thorp²⁴³⁾. Analogously, the sizeable carnivory of C₄-consuming mammals (such as cane rats, hyraxes or juvenile bovis) was argued to be practically impossible and thus unable to leave a strong C₄ signature⁽Reference Sponheimer and Lee-Thorp²⁴³⁾. However, at about 1·8 Mya, there were extensive wetlands in the Olduvai area, where a river from the Ngorongoro mountains entered the area, while at 1·5 Mya the Peninj river produced wetlands near Lake Natron⁽Reference Stewart, Cunnane and Stewart¹⁵⁶⁾. Some researchers investigated the edible plants in a present-day wetland (Okavango Delta) and found that the rhizomes and culms of three species of C₄ sedges were edible, the most common one of which is Cyperus papyrus ⁽Reference van der Merwe, Masao and Bamford²⁴⁵⁾. However, it seems unlikely that the C₄ signature in all early hominins derived from the consumption of papyrus. It was recently suggested that P. robustus and especially P. boisei had a diet of primarily C₄ resources, most probably grasses or sedges, from savanna or wetland environments, respectively⁽Reference Cerling, Mbua and Kirera¹⁹⁰⁾. Theoretically, a good source of C₄ foods would be a seasonal freshwater wetland with floodplains and perennial marshlands, with an abundance of easy accessible aquatic foods, large aggregations of nesting birds and calving ungulates⁽Reference Peters and Vogel²³⁴⁾. Consumption of termites could have contributed to the high C₄ signature observed in hominin fossils (Fig. 5), but it seems unlikely that termites could explain values as high as 50 % of the diet from C₄⁽Reference Sponheimer and Lee-Thorp²⁴³⁾. Finally, an enamel C₄ signature of − 10 to − 4 ‰ in hominins, which translates into a soft tissue signature of − 23 to − 17 ‰ and a collagen signature of − 18 to − 12 ‰ (see above), might also derive from the consumption of small freshwater aquatic animals or fish, since they compare well with the δ¹³C values of − 24 to − 13 ‰ for freshwater fish⁽Reference Mbabazi, Makanga and Orach-Meza²⁴¹⁾ and − 18 to − 9 ‰ in collagen of crustaceans and anthropods⁽Reference Schoeninger and Deniro²⁴⁰⁾, respectively. Moreover, δ¹³C values for hominins are similar to those reported for marine mammal hunters, freshwater fish, crustaceans, fisher–gatherers, marine and freshwater carnivores and marine fish (Fig. 5).

In agreement with the variability selection hypothesis of Potts⁽Reference Potts¹⁷⁰⁾, which states that large disparities in environmental conditions were responsible for important episodes of adaptive evolution, the wide range in δ¹³C values in particularly Australopithecus suggests that early hominins utilised a wide range of dietary sources, including C₄ resources. This contrasts with chimpanzees, which, even in the most arid and open areas of their range, are known to consume negligible amounts of C₄ resources, despite their local abundancy. Consequently, chimpanzees show very little variability in their δ¹³C carbon signature⁽Reference Schoeninger, Moore and Sept^246, Reference Sponheimer, Loudon and Codron²⁴⁷⁾. This underscores that even if contemporaneous chimpanzees and early hominins inhabited similar habitats, hominins had broadened their dietary range sufficiently to survive in habitats uninhabitable by chimpanzees. The latter assumption provides an interesting perspective on the recent data, which suggest that C₄ foods were absent in the diet of Ardipithecus ramidus at 4·4 Mya. Consequently, it has been proposed that the origins of the introduction of C₄ foods into the hominin diet lie in the period between 3 and 4 Mya⁽Reference Lee-Thorp, Sponheimer and Passey²²⁹⁾.

Limited evidence from the ¹⁵N:¹⁴N ratio

Another stable-isotope ratio that has received considerable attention is the N isotope (¹⁵N:¹⁴N) ratio. A number of food web studies have shown that each step in the food chain is accompanied by 3–4 ‰ enrichment in δ¹⁵N⁽Reference Ambrose and Deniro^228, Reference Kelly²³⁹⁾ and that δ¹⁵N can therefore be useful as a trophic level indicator. Additionally, animals feeding in marine ecosystems have higher values compared with animals feeding on terrestrial resources⁽Reference Schoeninger, Deniro and Tauber²⁴²⁾. For example, North American and European fisher–gatherers and North American marine mammal hunters and salmon fishers had much higher δ¹⁵N values (+13 to +20 ‰) compared with agriculturalists (+6 to +12 ‰). Analyses of phyto- and zooplankton suggest that freshwater organisms have δ¹⁵N values intermediate to terrestrial and marine organisms⁽Reference Schoeninger, Deniro and Tauber²⁴²⁾. δ¹⁵N values are routinely measured in bone collagen, but it has been shown that good-quality collagen (preserving the original δ¹⁵N value) can, and only under favourable conditions, survive up to a maximum of 200 000 years⁽Reference Lee-Thorp and Sponheimer²²⁵⁾. This limits δ¹⁵N isotopic studies to Late Pleistocene hominins (see below), but with improved technology, future studies using collagen extracted from tooth enamel may expand their application to early hominins⁽Reference Sponheimer, Lee-Thorp, de Ruiter and Ungar²³⁸⁾.

Limited evidence from the ¹⁸O:¹⁶O ratio

A final isotope that might provide information about an animal's diet and thermophysiological adaptations is the oxygen isotope ratio (¹⁸O:¹⁶O). More energy is needed to vaporise H₂¹⁸O than H₂¹⁶O. When ocean water evaporates and during evapotranspiration, i.e. the sum of evaporation and plant transpiration from the earth's land surface to the atmosphere, more of the lighter isotope evaporates as H₂¹⁶O. The ensuing ¹⁸O enrichment of transpiring leaves results in ¹⁸O enrichment in typical browsers such as kudu and giraffe who rely less on free drinking water and derive most of their water from the consumption of the ¹⁸O-enriched plant water. As the ¹⁶O-enriched water vapour in clouds moves inland, some of it condenses as rain, during which more of the heavier isotope (as H₂¹⁸O) rains out, making the δ¹⁸O of coastal rain only slightly less enriched than the original vaporated ocean water, while the δ¹⁸O of the remaining water vapour that eventually comes down is highly negative (i.e. more ¹⁸O depleted). Consequently, river water from rain and melting ice is more δ¹⁸O negative than seawater. Roots derive their water from meteoric or underground water that is thus relatively depleted from ¹⁸O and so become animals that are consuming these roots⁽Reference Sponheimer and Lee-Thorp^248, Reference Sponheimer and Lee-Thorp²⁴⁹⁾. Browsers of leaves undergoing evapotranspiration and consumers of roots may thus be expected to have high and low δ¹⁸O values, respectively.

Australopiths showed lower δ¹⁸O values compared with Paranthropus, but the meaning of this difference remains uncertain. However, one might argue that Australopithecus preferred less arid conditions compared with Paranthropus or was more dependent on seasonal drinking water⁽Reference Sponheimer, Lee-Thorp and de Ruiter²³¹⁾. Low δ¹⁸O was additionally found in primates and suids, which might be linked to frugivory, although this is not supported by the higher ¹⁸O values found in Ardipithecus ramidus compared with Australopithicines ⁽Reference White, Asfaw and Beyene^68, Reference Lee-Thorp, Sponheimer and Passey²²⁹⁾. Taken together, the use of δ¹⁸O for exploration of ancient human diets is still in its infancy, but might, especially in combination with other isotope ratios, become more appreciated in the future.

Isotopic data for more recent hominins

It would be of high interest to explore the hominin diet during the last spurt of encephalisation between 1·9 Mya to 100 Kya, when brain size tripled in size to volumes between 1200 and 1490 cm³ for Homo erectus, H. heidelbergensis, H. neanderthalensis and modern H. sapiens ⁽Reference Cunnane, Cunnane and Stewart¹⁰⁰⁾. Isotopic data for this period are, however, absent. Due to the limited preservation of collagen beyond 200 000 years, and the near absence of C₄ plants in Europe, these answers will have to come from further studies with tooth enamel in Africa and Asia. So far, there is no isotope evidence for the diet of Homo between 1·5 Mya up to 50 Kya (MP Richards, personal communication).

Dietary information from more recent humans comes from data on δ¹³C, supplemented with data on δ¹⁵N and the ¹³C:¹⁵N ratio. The ¹⁵N-isotope values of bone collagen⁽Reference Schoeninger, Deniro and Tauber²⁴²⁾ for differentiation between aquatic and agricultural diets were additionally verified by the study of the sulfur isotope ratios (³⁴S:³²S), since high intakes of marine organisms also result in higher δ³⁴S values⁽Reference Hu, Shang and Tong²⁵⁰⁾. Combined isotope studies reveal high intakes of animal protein, with substantial proportions derived from freshwater fish by Upper and Middle Palaeolithic (40–12 Kya) humans in Eurasia, indicating that in some populations about 30 % of dietary protein came from marine sources⁽Reference Hu, Shang and Tong^250–Reference Richards and Trinkaus²⁵⁶⁾. In contrast, isotopic evidence indicates that Neanderthals were top-level carnivores that obtained most of their dietary protein from large terrestrial herbivores, although even Neanderthals certainly exploited shellfish such as clams, oysters, mussels and fish on occasion⁽Reference Hu, Shang and Tong^250–Reference Richards, Pettitt and Stiner²⁵²⁾. At the onset of the Neolithic period (5200 years ago), there was a rapid and complete change from aquatic- to terrestrial-derived proteins among both coastal and inland Britons compared with Mesolithic (9000–5200 years ago) British humans⁽Reference Richards, Schulting and Hedges²⁵⁴⁾, which coincides precisely with the local onset of the Agricultural Revolution in Europe.

Conclusions from isotope studies

The isotope systems that have been studied thus far in hominin bone and teeth provide evidence that early hominins were opportunistic feeders⁽Reference van der Merwe, Thackeray and Lee-Thorp²⁵⁷⁾. The spread of C₄ foods in East Africa, and subsequently in the hominin food chain between 3 and 4 Mya, is in agreement with a niche of early hominins that locates close to the water. This conception is in agreement with the palaeo-environmental evidence. However, many questions still remain unanswered. With regard to the possible niche in the water–land interface, it seems interesting to include aquatic as well as terrestrial piscivorous animals into future studies. The data of combined studies of early hominins and the more recent hominins suggest a gradual increase in dietary animal protein, a part of which may derive from aquatic resources. In the more recent human ancestors, a substantial part of the dietary protein was irrefutably derived from marine resources, and this habit was only abandoned in some cases after the introduction of agriculture at the onset of the Neolithic⁽Reference Richards, Schulting and Hedges²⁵⁴⁾.

Living in the water–land ecosystem

Because sea levels have risen up to 150 m in the past 17 000 years, a substantial part of the evidence for the exploitation of aquatic resources is hidden below sea level, if not permanently destroyed by the water⁽Reference Erlandson^267, Reference Marean, Bicho, Haws and Davis²⁶⁸⁾. However, in Kenya, a site in East Turkana provides solid evidence that at about 1·95 Mya hominins enjoyed carcasses of both terrestrial and aquatic animals including turtles, crocodiles and fish, which were associated with Oldowan artifacts⁽Reference Braun, Harris and Levin²⁶¹⁾. More ambiguous evidence for the exploitation of freshwater fish, crocodiles, turtles, amphibians and molluscs by Homo habilis in the Olduvai Gorge in Tanzania goes back as far as 1·8–1·1 Mya⁽Reference Erlandson^267, Reference Stewart²⁶⁹⁾. Subsequent tentative evidence from the Olduvai Gorge dates the use of similar aquatic resources by Homo erectus to 1·1–0·8 Mya⁽Reference Erlandson^267, Reference Stewart²⁶⁹⁾. Also the out-of-Africa diaspora probably took place largely via the coastlines⁽Reference Stringer⁸¹⁾, even after the crossing of the Bering Strait into North America⁽Reference Wang, Lewis and Jakobsson²⁷⁰⁾ (Fig. 2). In Koa Pah Nam, Thailand, 700 000-year-old piles of freshwater oyster shells were associated with Homo erectus ⁽Reference Pope^271, Reference Fagan²⁷²⁾. In Holon, Israel, freshwater turtles, shells and hippopotamus bones were associated with Homo erectus and dated to 500–400 Kya⁽Reference Bar-Yosef²⁷³⁾. Homo erectus fossils associated with seal remains in Mas del Caves (Lunel-Viel, France) were dated to 400 Kya⁽Reference Cleyet-Merle, Madelaine and Fischer²⁷⁴⁾.

The archeological evidence for aquatic resource use increases with the appearance of archaic Homo sapiens ⁽Reference Erlandson²⁶⁷⁾. Although dominated by land mammal bones, 400 000–200 000-year-old remains from penguin and cormorants in Duinefontein, South Africa were associated with early Homo ⁽Reference Klein, Avery and Cruz-Uribe²⁷⁵⁾. Shellfish and possibly fish remains, dated 300–230 Kya, were associated with the French coastal campsite at Terra Amata⁽Reference de Lumley^276, Reference Villa²⁷⁷⁾, while marine shellfish and associated early human remains, dated 186–127 Kya, were found in Lazaret, France⁽Reference Cleyet-Merle, Madelaine and Fischer²⁷⁴⁾. Marean et al. ⁽Reference Marean, Bar-Matthews and Bernatchez²⁷⁸⁾ found evidence for the inclusion of marine resources, at 164 Kya, in the diet of anatomically modern humans from the Pinnacle Point Caves (South Africa). At the Eritrea Red Sea coast, Middle Stone Age artifacts on a fossil reef support the view that early humans exploited near-shore marine food resources by at least 125 Kya⁽Reference Walter, Buffler and Bruggemann²⁷⁹⁾. In several North African sites, dated to 40–150 Kya⁽Reference Erlandson²⁶⁷⁾, human remains were associated with shell middens and aquatic resources such as aquatic snails, monk seals, mussels and crabs. Several European sites, dated to 30–125 Kya, are comparable with archeological sites that reveal evidence ranging from thick layers of mussels and large heaps of marine shells in Gibraltar⁽Reference Erlandson²⁶⁷⁾ to diverse marine shells in Italy⁽Reference Stiner²⁸⁰⁾, and to a casual description of the presence of marine shells of unknown density in Gruta da Figueira in Portugal⁽Reference Erlandson²⁶⁷⁾. Further evidence for the use of shellfish, sea mammals and flightless birds comes from: Klasies River Mouth (South Africa) dated between 130 and 55 Kya⁽Reference Erlandson^267, Reference von den Driesch^281, Reference Erlandson, Cunnane and Stewart²⁸²⁾; from Boegoeberg, where 130 000–40 000-year-old shell middens and cormorant bones were associated with Homo sapiens ⁽Reference Erlandson^267, Reference Klein, Cruz-Uribe and Halkett²⁸³⁾; from Herolds Bay Cave, where 120 000–80 000-year-old shell middens, shellfish, mussels and otter remains were associated with human hearths⁽Reference Erlandson²⁶⁷⁾; from Die Kelders (75–55 Kya), where abundant remains of sea mammals, birds and shellfish were found in cave deposits⁽Reference Erlandson^267, Reference Marean, Goldberg and Avery²⁸⁴⁾; and from Hoedjies Punt (70–60 Kya), Sea Harvest (70–60 Kya) and Blombos Cave (60–50 Kya) for the use of shellfish, sea mammals and fish⁽Reference Erlandson^267, Reference Henshilwood and Sealy²⁸⁵⁾. From this period onwards, human settlements are strongly associated with the exploitation of aquatic resources⁽Reference Erlandson^267, Reference Marean, Bicho, Haws and Davis^268, Reference Erlandson, Cunnane and Stewart^282, Reference Klein, Avery and Cruz-Uribe^286–Reference Henshilwood, d'Errico and Vanhaeren²⁸⁸⁾. Evidence for more sophisticated fishing by use of barbed bone harpoon points dates back to 90–75 Kya in Katanda, Semlike River, Zaire⁽Reference Harris, Williamson, Morris and Boaz^289, Reference Meylan and Boaz²⁹⁰⁾ and to 70 Kya in South Africa⁽Reference Henshilwood, Sealy and Yates²⁹¹⁾. Finally, indications for seafaring are dated to 42–15 Kya⁽Reference Wickler and Spriggs^292–Reference O'Connor, Ono and Clarkson²⁹⁴⁾. Possibly, seafaring dates as far back as 800 Kya, as indicated by the finding of Homo erectus stone tools at the Indonesian island of Flores, which is located on the other side of a deep sea strait⁽Reference Brown, Sutikna and Morwood^79, Reference Morwood, O'Sullivan and Aziz^295–Reference Brumm, Jensen and van den Bergh²⁹⁹⁾. In general, many archeological sites are found along channels, lake- and seashores⁽Reference Broadhurst, Cunnane and Crawford^99, Reference Broadhurst, Wang and Crawford^266, Reference Erlandson^267, Reference Erlandson, Cunnane and Stewart²⁸²⁾ and reveal aquatic fauna, such as catfish, crocodile and hippopotamus⁽Reference Sept³⁰⁰⁾, but its proves difficult to relate their possible utilisation to our early ancestors.

Several events within the time span of the past about 2 million years have been attributed to the increase in brain size and intelligence. The introduction of meat in the hominin diet, which resulted in a higher dietary quality, has been discussed above. Claims for controlled fire in the Olduvai Gorge (Tanzania) and Koobi Fora (Kenya) go as far back as 1·5 Mya⁽Reference Wrangham, Jones and Laden^215, Reference Gibbons^301, Reference Wobber, Hare and Wrangham³⁰²⁾. Evidence for cooking is as old as 250 Kya⁽Reference Gibbons³⁰¹⁾, but possibly dates back to 800 Kya⁽Reference Goren-Inbar, Alperson and Kislev³⁰³⁾, when indications of controlled fire were found to be present. However, recently it has been concluded that solid evidence for systematic use of fire is only found from 400 to 300 Kya onwards⁽Reference Roebroeks and Villa³⁰⁴⁾. Evidence that cooking provided increased dietary quality was recently provided by Wrangham et al. ⁽Reference Wrangham, Jones and Laden²¹⁵⁾. According to archeological evidence, this could only have played an important role since the appearance of Homo sapiens ⁽Reference Wrangham, Jones and Laden^215, Reference Gibbons³⁰¹⁾ and Neanderthals⁽Reference Henry, Brooks and Piperno³⁰⁵⁾. Also, the inclusion of aquatic resources as an attributor to human brain evolution has been suggested⁽Reference Crawford, Cunnane and Stewart^2, Reference Broadhurst, Cunnane and Crawford^99–Reference Cunnane^101, Reference Crawford, Bloom and Broadhurst^130, Reference Broadhurst, Wang and Crawford^266, Reference Crawford, Bloom and Cunnane^306–Reference Parkington, Roggenpoel, Halkett and Conard³⁰⁸⁾, but remains a matter of debate⁽Reference Carlson and Kingston^134, Reference Joordens, Kuipers and Muskiet^135, Reference Marean, Bicho, Haws and Davis^268, Reference Cordain, Eaton and Sebastian³⁰⁹⁾.

From hunting–gathering to agriculture

The hunter–gatherer lifestyle continued worldwide for several millions of years and ended quite abruptly with the introduction of agriculture. The first indications for the abandonment of the hunter–gatherer lifestyle towards settlement come from a 23 000 year-old fisher–hunter–gatherer's camp at the shore of the Sea of Galilee⁽Reference Nadel, Weiss and Simchoni^310, Reference Weiss, Wetterstrom and Nadel³¹¹⁾. The associated return from diets containing substantial amounts of protein (from hunting and gathering) back to substantial amounts of carbohydrates is supported by indications for the ground collecting of wild cereals⁽Reference Kislev, Weiss, Hartmann and Hartmann³¹²⁾. This was slowly followed by the large-scale utilisation of cereals starting with the onset of the Agricultural Revolution some 10 Kya.

As indicated above (see ‘Biogeochemistry’), there is much controversy about the diet of the earliest humans and until now it is often stated that fishing was only introduced until more recently. From an anthropological perspective this might be true, since certain types (for example, deepwater) of fishing require advanced techniques⁽Reference O'Connor, Ono and Clarkson²⁹⁴⁾. However, from a nutritional point of view, ‘fishing’ might include anything from collecting sessile shellfish to the seasonal hand capture or clubbing of migrating or spawning fish in very shallow water. Since fresh drinking water is the single most important aquatic resource for humans, hominins probably observed predators and scavengers feeding on aquatic animals. This makes it unlikely that they would not have participated in opportunistic harvesting of the shallow-water flora and fauna, such as molluscs, crabs, sea urchins, barnacles, shrimp, fish, fish roe or spawn, amphibians, reptiles, small mammals, birds or weeds⁽Reference Joordens, Wesselingh and de Vos^175, Reference Erlandson²⁶⁷⁾. There are many indications suggesting that the evolution of early Homo and its development to Homo sapiens did not take place in the ‘classical’ hot, arid and waterless savanna, but occurred in African ecosystems that were notably located in places where the land meets the water (with the land ecosystem possibly consisting of – depending on rainfall – wooded grasslands). Compared with terrestrial hunting and/or scavenging in the savanna, food from this land–water ecosystem is relatively easy to obtain and is rich in the aforementioned combination of haem-Fe, iodine, Se, vitamins A and D, and long-chain n-3-fatty acids⁽Reference Cunnane, Cunnane and Stewart^100, Reference Cunnane^101, Reference Broadhurst, Wang and Crawford²⁶⁶⁾.

In conclusion, there is ample archeological evidence for a shift from the consumption of plant towards animal foods. Second, although there is an extensive archeological record for aquatic fossils (representing possible food) in the vicinity of human remains, their co-occurence is usually attributed to the preferential conservation of human remains in the vicinity of water. The present review provides support for the notion that the exploitation of these aquatic resources by hominins in coastal areas should be the default assumption, unless proven otherwise⁽Reference Joordens¹⁵⁴⁾. For a long time period in hominin evolution, hominins derived large amounts of energy from (terrestrial and aquatic) animal fat and protein. This habit became reversed only by the onset of the Neolithic Revolution in the Middle East starting about 10 Kya.

The hunter–gatherer diet

The Homo genus has been on earth for at least 2·4 million years⁽Reference Kimbel, Walter and Johanson³¹³⁾ and for over 99 % of this period has lived as hunter–gatherers⁽Reference Lee, Lee and DeVore³¹⁴⁾. Surprisingly, very little information is available on the macro- and micronutrient compositions of their diet in this extended and important period of human evolution⁽Reference Eaton, Eaton and Sinclair^34, Reference Cordain, Miller and Eaton¹⁴⁸⁾. Since the onset of agriculture, about 10 Kya, agriculturalists and nomadic pastoralists have been expanding at the expense of hunter–gatherers⁽Reference Lee, Lee and DeVore³¹⁴⁾, with agricultural densities increasing by a factor of 10–1000 compared with the highest hunter densities. For this reason, present-day hunter–gatherers are often found in marginal environments, unattractive for crop cultivation or animal husbandry.

In order to study the original hunter–gatherer way of life, it is appropriate to aim at the few hunter–gatherer communities living in the richer environments that bear closer resemblance to those in which the evolution of the genus Homo probably took place. Most studies on hunter–gatherers and their diets are, however, performed by anthropologists⁽Reference Murdock³¹⁵⁾, whose primary interests are different from those of nutritionists. Anthropologists would, for example, conclude that ‘fishing was so unimportant as to be a type of food collection’⁽Reference Stewart, Lee and DeVore³¹⁶⁾, or consider collecting both small land fauna and shellfish⁽Reference Lee, Lee and DeVore³¹⁴⁾ as part of ‘gathering’, whereas from a nutritional point of view considerable differences exist in energy density, macro- and micronutrient composition between plants, terrestrial and aquatic animal foods.

Brain-selective nutrients

Nutrients and other environmental factors are increasingly recognised to influence epigenetic marks⁽Reference Godfrey, Lillycrop and Burdge^319–Reference Godfrey, Sheppard and Gluckman³²³⁾, either directly or indirectly via many bodily sensors. Food from the diverse East African aquatic ecosystems is rich in haem-Fe, iodine, Se, vitamins A and D, and n-3 fatty acids from both vegetable origin and fish⁽Reference Cunnane¹⁰¹⁾. All of these nutrients seem to act at the crossroad of metabolism and inflammation⁽Reference Muskiet, Montmayeur and Coutre²⁴⁾. For example, PPAR⁽Reference Bensinger and Tontonoz^324, Reference Castrillo and Tontonoz³²⁵⁾ are lipid-driven nuclear receptors with key cellular functions in metabolism and inflammation⁽Reference Feige, Gelman and Michalik²⁶⁾. TR⁽Reference Song, Yao and Ying³²⁶⁾, VDR⁽Reference Bouillon, Bischoff-Ferrari and Willett³²⁷⁾, RXR and RAR⁽Reference McGrane³²⁸⁾ are other examples of nuclear transcription factors that serve functions as ligand-driven sensors. The iodine- and Se-dependent hormone triiodothyronine (T₃)⁽Reference Venturi, Donati and Venturi^329–Reference Gilbert, McLanahan and Hedge³³¹⁾ is a ligand of TR⁽Reference Song, Yao and Ying³²⁶⁾, many fatty acids and their derivatives are ligands of PPAR⁽Reference Desvergne and Wahli³³²⁾, the vitamin D-derived 1,25-dihydroxyvitamin D hormone is a ligand of the VDR⁽Reference Norman and Bouillon³³³⁾, 9-cis-retinoic acid and the fish oil fatty acid DHA are ligands of RXR⁽Reference McGrane³²⁸⁾, while RAR interacts with vitamin A (retinol) and many of its derivatives such as all-trans-retinoic acid, retinal and retinyl acetate⁽Reference McGrane³²⁸⁾. The ligated nuclear transcription factors usually do not support transcription by themselves, but need to homodimerise or heterodimerise notably with RXR to facilitate gene transcription. Examples of the latter are TR/RXR, PPAR/RXR, VDR/RXR and RAR/RXR. It has become clear that their modes of action illustrate the need of balance between, for example, iodine, Se, fish oil fatty acids and vitamins A and D, a balance that is notably found in the land–water ecosystem.

Deficiencies of the above ‘brain-selective nutrients’ are among the most widely encountered in the current world population⁽Reference Cunnane^101, Reference Holick and Chen³³⁴⁾. While iodide is added to table salt in many countries, margarines and milk have become popular food products for fortification with vitamins A and D. After discussing some general health differences between traditionally living individuals and those living in Westernised countries, we focus on the importance of LCP and notably those of the n-3-series, as examples of the above-mentioned nutrients that are especially abundant in the land–water ecosystem.

Hunter–gatherer v. ‘Western’ physiology

There are many differences in health indicators between traditionally living individuals and those living in Western societies. For instance, primary and secondary intervention trials with statins indicate lowest CHD risk at an LDL-cholesterol of 500–700 mg/l (1·3–1·8 mmol/l), which is consistent with levels encountered in primates in the wild and hunter–gatherer populations with few deaths from CVD⁽Reference Mann, Roels and Price^335–Reference O'Keefe, Cordain and Harris³³⁹⁾. Another example of the healthy lifestyle of present-day hunter–gatherers comes from the observed ‘insulinopenia’ or ‘impaired insulin secretion’ following an oral glucose tolerance test (Fig. 6) in Central African Pygmies and Kalahri Bushmen⁽Reference Joffe, Jackson and Thomas^340, Reference Merimee, Rimoin and Cavalli-Sforza³⁴¹⁾, respectively. As opposed to the ‘impairments’ noted by these authors, it may also be argued that these researchers were actually witnessing an insulin sensitivity that has become sporadic in Western countries as a consequence of the decrease in physical activity and fitness, increase in fat mass and as a result of the quantity and quality of the foods consumed⁽Reference O'Keefe and Cordain^36, Reference Cordain, Eaton and Sebastian^133, Reference Ramsden, Faurot and Carrera-Bastos³⁴²⁾. The current consensus is that ‘fat is bad’ and especially saturated fats have become associated with CVD⁽Reference Keys^343–Reference Kuipers, de Graaf and Luxwolda³⁴⁵⁾. However, traditional Maasai consumed diets high in protein and fat (milk and meat) and low in carbohydrates⁽Reference Ho, Biss and Mikkelson^346, Reference Biss, Ho and Mikkelson³⁴⁷⁾. They had high intakes of saturated fat and cholesterol, showed extensive atherosclerosis with lipid infiltration and fibrous changes, but had very few complicated lesions, and were virtually devoid of CVD⁽Reference Mann, Spoerry and Gray³⁴⁸⁾. The average total and LDL-cholesterol in these societies was low and did not increase with age⁽Reference Mann, Spoerry and Gray³⁴⁸⁾. Finally, the physical fitness of individuals in such traditional societies, such as the Maasai, is often remarkable⁽Reference Mann, Shaffer and Rich³³⁷⁾.

Fig. 6 ‘Abnormal’ insulin response but normal glucose response after oral glucose tolerance test in African Bushmen and Pygmies, compared with Western controls. (a) Plasma glucose response after an oral glucose load of 50 g (Bushmen (♦) and white controls (○)) or 100 g (Pygmy (■), Pygmy with 2 weeks' daily supplementation of 150 g carbohydrates before testing (▲), Bantu (X) and American controls ()). Of note is that Bushmen and Pygmies have significantly lower body weights as compared with Bantu and white and American controls (average weight Bushmen/Pygmy males 46 kg, females 38 kg; controls 65 kg), while each group received the same unadjusted loading dose of 50 g or 100 g glucose. (b) The so-called ‘abnormal’ insulin response or ‘impaired’ insulin secretion as observed by the authors in both Bushmen and Pygmies⁽Reference Joffe, Jackson and Thomas^340, Reference Merimee, Rimoin and Cavalli-Sforza³⁴¹⁾.

In Kalahari Desert Bushmen and Central African Pygmies, observers could not find any case of high blood pressure and blood pressure did not increase with age⁽Reference Mann, Roels and Price^335, Reference Kaminer and Lutz³⁴⁹⁾. Dental surveys of Kalahari Bushmen⁽Reference Clement, Fosdick and Plotkin³⁵⁰⁾ and other hunter–gatherers⁽Reference Larsen⁴⁹⁾ showed a remarkable absence of caries. The absence was explained by the repetitive annual abstinence of fermentable sugars in their diet, with a consequent inability to build a cariogenic oral Lactobacillus flora⁽Reference Clement, Fosdick and Plotkin³⁵⁰⁾. Inhabitants of Kitava (Trobriand Islands, Papua New Guinea) have high intakes (70 en%) of carbohydrates from yams, high intakes of SFA from coconuts, and a high fish intake⁽Reference Lindeberg and Lundh^44, Reference Lindeberg, Nilsson-Ehle and Terent^45, Reference Lindeberg, Berntorp and Nilsson-Ehle³⁵¹⁾. Although both high intakes of carbohydrates and saturated fat have been related to the metabolic syndrome and CVD, these traditional Kitavians do not show symptoms of either the metabolic syndrome and are virtually free from the Western diseases that ensue from it.

Evidence-based medicine as applied to long-chain PUFA in CVD and depression

Despite some compelling examples of the healthy lifestyles of traditional populations, current dietary recommendations derive preferably from randomised clinical trials with single nutrients and preferably hard endpoints⁽Reference Sackett, Rosenberg and Gray³⁵²⁾. This approach clearly oversimplifies the effects of dietary nutrients⁽Reference Blumberg, Heaney and Huncharek³⁵³⁾, since neither macronutrients, nor micronutrients, are consumed in isolation and their effects may be the result of a complex web of interactions between all the nutrients present in the biological systems that we consume, such as a banana or a fish.

The current recommendations from many nutritional boards for a daily intake of 450 mg EPA + DHA in adults derive from epidemiological data that demonstrated a negative association of fish consumption with CHD⁽Reference Dyerberg, Bang and Hjorne^354–Reference Mozaffarian and Rimm³⁵⁷⁾ that has subsequently become supported by landmark trials with ALA⁽Reference de Lorgeril, Salen and Martin³⁵⁸⁾ and fish oil^(359–Reference Yokoyama, Origasa and Matsuzaki³⁶¹⁾ in CVD. However, not all trials in CVD have been positive⁽Reference Kromhout, Giltay and Geleijnse³⁶²⁾. In addition, a negative association was observed for fish consumption and depression⁽Reference Hibbeln^363–Reference Hibbeln³⁶⁵⁾ and for homicide mortality⁽Reference Hibbeln³⁶⁶⁾. The causality of these relationships was supported by some, but not all, trials with fish oil in depression⁽Reference Lin and Su^367–Reference Freeman, Hibbeln and Wisner³⁷¹⁾, while a recent meta-analysis demonstrated the beneficial effect of EPA supplements with ≥  60 % EPA of total EPA + DHA in a dose range of 200–2200 mg/d of EPA in excess of DHA⁽Reference Sublette, Ellis and Geant³⁷²⁾.

The influence of polymorphisms in the genome is increasingly recognised, but seldom interpreted in an evolutionary context. As argued above, most polymorphisms were already amongst us when Homo sapiens emerged, some 200 Kya, while that also holds true for most, if not all, currently identified ‘disease susceptibility genes’ that are usually abundant but confer low risk⁽³⁷³⁾. A loss-of-function mutation in a specific biosynthetic pathway might be an evolutionary advantage if the specific endproduct has been a consistent part of the diet, such as is probably applicable to all vitamins, for example, vitamin C⁽Reference Challem^374, Reference Nishikimi, Fukuyama and Minoshima³⁷⁵⁾. Applied to our LCP status, it is nowadays well established that all humans synthesise DHA with difficulty⁽Reference Burdge and Wootton^376, Reference Burdge³⁷⁷⁾. Analogously, the recently discovered polymorphisms of fatty acid desaturases 1 (FADS1; also named Δ-5 desaturase) and FADS2 (Δ-6 desaturase) with lower activities in their conversion of the parent essential fatty acid to LCP suggest that from at least the time of their appearance, the dietary intakes of AA, EPA and DHA have been of sufficient magnitude to balance the LCP n-3:LCP n-6 ratio⁽Reference Kuipers, Luxwolda and Janneke Dijck-Brouwer^378, Reference Luxwolda, Kuipers and Smit³⁷⁹⁾ to maintain good health.

Long-chain PUFA benefits in pregnancy and early life

Another indication for the importance of LCP comes from the higher LCP contents in the fetal circulation compared with the maternal circulation, a process named biomagnification⁽Reference Crawford, Hassam and Williams^380–Reference Crawford, Hassam and Rivers³⁸²⁾, which occurs at the expense of the maternal LCP status⁽Reference Hornstra^383, Reference Hornstra³⁸⁴⁾. The decreasing maternal n-3 LCP status during pregnancy in Western countries is associated with postpartum depression⁽Reference Hibbeln^363, Reference Hibbeln³⁶⁴⁾, although intervention studies with LCP in postpartum depression have been negative so far⁽Reference Freeman^370, Reference Freeman, Hibbeln and Wisner^371, Reference Makrides, Gibson and McPhee^385, Reference Doornbos, van Goor and Dijck-Brouwer³⁸⁶⁾. However, a positive effect was seen for n-3 LCP supplementation on depression during pregnancy⁽Reference Su, Huang and Chiu³⁸⁷⁾ and it has been advocated to start supplementation earlier in pregnancy and with higher dosages⁽Reference Wojcicki and Heyman³⁸⁸⁾.

Maternal LCP intakes have also been related to infant health. AA and DHA in premature and low-birth-weight infants correlated positively with anthropometrics, AA to increased birth weight⁽Reference Koletzko and Braun³⁸⁹⁾ and DHA to prolonged gestation⁽Reference Szajewska, Horvath and Koletzko^390–Reference Olsen, Osterdal and Salvig³⁹²⁾. Studies with supplementation of DHA during pregnancy yielded, for example, evidence for: (i) the maturation of the brain, visual system and retina of the newborn at 2·5 and 4 months, but not at 6 months⁽Reference Malcolm, McCulloch and Montgomery^393–Reference Smithers, Gibson and McPhee³⁹⁷⁾; (ii) increased problem solving at 9 months but no difference in memory⁽Reference Judge, Harel and Lammi-Keefe³⁹⁸⁾; and (iii) superior eye–hand coordination at 2·5 years⁽Reference Dunstan, Simmer and Dixon³⁹⁹⁾ and higher intelligence quotient at 4 years⁽Reference Helland, Smith and Saarem⁴⁰⁰⁾ but not at 7 years of age⁽Reference Helland, Smith and Blomen⁴⁰¹⁾. In contrast to the inconclusive human studies, animal studies and combined human and animal studies showed abnormal behaviour together with disturbed cognition at lower brain DHA levels⁽Reference McCann and Ames⁴⁰²⁾. The importance of dietary AA during pregnancy seems less pronounced, but a positive association between umbilical AA and neonatal neurological development⁽Reference Koletzko and Braun³⁸⁹⁾ and a lower venous AA for those with slightly abnormal neurological development⁽Reference Dijck-Brouwer, Hadders-Algra and Bouwstra⁴⁰³⁾ has been shown. A reduced DHA status in the brain is associated with a mildly increased AA status⁽Reference Moriguchi, Loewke and Garrison⁴⁰⁴⁾, which is in its turn associated with low-grade inflammation⁽Reference Rao, Ertley and DeMar⁴⁰⁵⁾.

Infant health starts with maternal health; thus dietary recommendations issued for pregnant women indirectly also apply to their infants. The recommendation for adults to consume 450 mg DHA + EPA per d translates into a DHA composition in breast milk of about 0·79 %⁽Reference van Goor, Smit and Schaafsma⁴⁰⁶⁾. However, current recommendations for the composition of infant formulae derive mainly from the range of human milk fatty acid compositions as observed in Western countries, which in their turn derive from women with recorded intakes below the 450 mg recommended daily intake of EPA + DHA⁽Reference Brenna, Varamini and Jensen^407, Reference Koletzko, Lien and Agostoni⁴⁰⁸⁾.

The same paradox holds for other fatty acids in breast milk. For instance, there are few recommendations for the medium-chain SFA (MCSFA) content of human milk. High MCSFA contents in some traditional societies derive from their high intakes of 12 : 0 and 14 : 0 from coconuts⁽Reference Kuipers, Smit and van der Meulen⁴⁰⁹⁾. Conversely, the high MCSFA contents in Western populations are primarily influenced by maternal carbohydrate intakes⁽Reference Hachey, Silber and Wong⁴¹⁰⁾, since the mammary gland has the unique ability to convert glucose into MCSFA (6 : 0–14 : 0), mainly lauric (12 : 0) and myristic (14 : 0) acids. However, women with regular consumption of coconuts have a much higher 12 : 0:14 : 0 ratio compared with women with high carbohydrate intakes. Both MCSFA are readily absorbed in the gastrointestinal tract, while antiviral as well as antibacterial properties have been attributed to some MCSFA, but mainly to 12 : 0⁽Reference Kabara^411, Reference Bergsson, Steingrimsson and Thormar⁴¹²⁾.

The PUFA content of Western milk has increased over the last decades⁽Reference Widdowson, Dauncey and Gairdner^413, Reference Ailhaud, Massiera and Weill⁴¹⁴⁾. While the human milk LA content in the USA increased by at least 250 %, its DHA content decreased by almost 50 %⁽Reference Ailhaud, Massiera and Weill⁴¹⁴⁾. The (from an evolutionary point of view) abnormally high LA intake is, despite a lack of evidence⁽Reference Ramsden, Hibbeln and Majchrzak⁴¹⁵⁾, advocated for cardiovascular health⁽Reference Harris, Mozaffarian and Rimm⁴¹⁶⁾. The resulting high LA status is likely to interfere with both the incorporation of AA and DHA into phospholipids and also inhibits their synthesis from their parent essential fatty acid⁽Reference Gibson, Muhlhausler and Makrides⁴¹⁷⁾. Major differences are noted in the comparison of the human milk fatty acid compositions of Western mothers compared with some traditional African women⁽Reference Kuipers, Smit and van der Meulen^409, Reference Koletzko, Thiel and Abiodun⁴¹⁸⁾, with unknown consequences for infant health or the occurrence of disease at adult age (i.e. the ‘Barker hypothesis’)⁽Reference Barker^16, Reference Godfrey, Robinson and Barker⁴¹⁹⁾. It has been proposed that the high concentrations of EPA, DHA as well as AA in human milk, such as described for many fish-consuming societies⁽Reference Kuipers, Smit and van der Meulen^409, Reference Innis and Kuhnlein^420, Reference Ruan, Liu and Man⁴²¹⁾, might be a more appropriate reflection of the Palaeolithic breast milk composition and may therefore constitute a better reference for infant formulae than do Western human milks⁽Reference Muskiet, Kuipers and Smit⁴²²⁾.

The influence of environment

It is estimated that 70 % of all cases of stroke and colon cancer, 80 % of all CVD and 90 % of all cases of type 2 diabetes mellitus have been caused by lifestyle and could have been prevented by paying more attention to modifiable behaviour factors, including specific aspects of diet, overweight, inactivity and smoking⁽Reference Willett⁴²³⁾. The mismatch between the human diet and the Palaeolithic genome might therefore be responsible for many typically Western diseases. In addition to the evidence from many other disciplines, evidence from (patho)physiology and epidemiology adds to the notion that a great deal of information on healthy diets might derive from the study of the diets of the early human ancestors. The metabolic syndrome, characterised by impaired insulin sensitivity, is at the centre of many diseases of civilisation. High intakes of refined carbohydrates as well as low intakes of LCP have been implicated in the development of insulin resistance. As such, low carbohydrate intakes⁽Reference Yudkin^424, Reference Yudkin⁴²⁵⁾ and high LCP⁽Reference Eaton and Konner^8, Reference Eaton, Eaton and Sinclair^34, Reference Eaton^426, Reference Cordain, Watkins and Mann⁴²⁷⁾ intakes by the early human ancestor might explain in part the low incidence of diseases of civilisation in current hunter–gatherer societies. The available evidence from pathophysiology and epidemiology supports the hypothesis that the land–water ecosystem contributed important and indispensable nutrients to evolving hominins.