Early humans recipe for processing plants foods at Blombos Cave, South Africa (85-82 ka)


 We present the results of archaeobotanical research conducted into the plant diet of early modern humans who intermittently occupied Blombos Cave on the southern Cape coast of South Africa during the Middle Stone Age (MSA). Botanical samples were taken from two combustion events in the MSA sequence dated to 85 and 82 kya (kya = thousands of years ago). Analysis of these samples shows charred starchy plants had undergone processing sequences prior to cooking. The plant cells show evidence of having been pounded whilst still fresh and some of these fragments indicate pounded starchy plant food was mixed with pounded seeds prior to cooking. We infer that these samples represent the earliest known example of food processing and mixing in a ‘recipe’ as part of early human foodways. We consider these findings and their role in human evolution in the context of human behavioural modernity.


Introduction
Here we present evidence of starchy plant remains recovered from combustion events dated to 85 kya and 82 kya in Blombos Cave, South Africa. We infer from indications of pre-processing, in this case pounding, then mixing of different plants followed by re treatment, the earliest attested case of a 'recipe'.
Blombos Cave (BBC) is situated on the southern Cape coast of South Africa, within the Greater Cape Floristic Kingdom, around 300 km east of Cape Town (34°24 × 51´´S, 21°13 × 19´´E). The vegetation surrounding Blombos is a subtropical thicket-fynbos mosaic (strandveld) on calcareous coastal dunes 1 .
Fynbos is rich in underground storage organs making edible starchy plants then, as now, available yearround. Blombos Cave has a well-dated and strati ed sequence that spans the Middle Stone Age 2 deposits (127.5 ± 16.9-67.8 ± 4.2 kya) (Fig. 1). The MSA layers are topped by a layer of sterile aeolian sand dune (~ 70 kya) and a Later Stone Age sequence dated to 2 kya 3,4 .
The MSA at Blombos is identi ed in four phases (phases = layers identi ed with a human occupation characterised by particular lithic styles and hunting and foraging strategies). The upper two phases, BBC M1 and BBC M2 upper are associated with the Still Bay techno-complex dated to 77 − 73 kya [5][6][7] . Other cultural markers from these phases include bone awls and points, engravings of parallel and joining lines on bone 8,9 , beads made from Nassarius kraussianus shells 10,11 , at least eight pieces of ochre engraved with geometric patterns 12,13 and most recently an abstract drawing on a ground silcrete ake, dated to 73 kya 3 . The M2 lower phase dated to 85 − 82 kya, is represented by a low-intensity occupational horizon lacking Still Bay artefacts. The nal M3 phase (127 − 94 kya) contains abundant evidence of ochre use, including engraved ochre and a liquid pigmented compound stored in abalone shells from a layer dated to 100 kya 4,7 . Context This paper focusses on two of the ten botanical samples taken from across the Blombos Cave southern section ( Fig. 1), that come from distinct anthropogenic combustion features, identi ed in section, from two layers of the M2 lower phase. Sample 1 from layer CGABh1, quadrant G7b, which is dated to 82 kya (total station ref: 2688) and Sample 2 from layer CGAC, quadrant G7b, which is dated to 85 kya (total station ref: 2728) (see Fig. 1). It is likely that these combustion events are remnant hearths that may have been either trampled and/or cleaned.
To establish that the route into the archaeological record was anthropogenic and that the manner of deposition was consistent with processing and cooking, we considered whether burrowing animals, other organic deposits (possibly aeolian) or other anthropogenic activity could have been responsible for these assemblages The most likely taxa that would burrow in this cave would be Procavia capensis (Rock hyrax), which occur in the vicinity of Blombos today and Bathyergus suillus (Cape dune mole rat). Both species have a diet that is dominated by grasses and sedges, although B. suillus also eat bulbs and tubers. Neither species store food underground. Only B. suillus nests underground [14][15][16] . Both species defecate and urinate in outside latrines, which can sometimes be metres thick and which are visible to the naked eye in the archaeological record 17,18 . Any assemblage attributable to either of these burrowing animals would contain a signi cant number of phytoliths and no phytoliths were present in this assemblage. We have, therefore, dismissed this possibility.
There are organic deposits with humi ed remains and these have the appearance of combustion features in section but contain no/very few charred plant remains of any kind. It is possible that this vegetation may have blown into the cave, as has signi cant amounts of sand. These deposits may account for three unsuccessful samples out of the ten (see Table S The other route into the archaeological record may be attributed to plants, such as sedge grasses, used by hunter-gatherers as bedding. Occasionally, the rhizomes and tubers remained attached to these grasses when gathered and used as bedding as at Sibudu Cave 19,20 . The bedding was then burned as part of a periodic cleansing of the cave leaving charred rhizomes and tubers. This scenario leaves a distinctive archaeological signature of long burned layers in the stratigraphy, or if not burned, then long white layers created by phytoliths from decomposed sedge grass stems, both visible at Sibudu Cave 20 . At Sibudu the parenchyma from these rhizomes and tubers has perfectly preserved cell structure, indicating they had dried before burning. This is not the case at Blombos and no phytoliths were found in the SEM process. The density of charred starchy plant fragments recovered from these two contexts (see Results - Table   S.3 in the S.I.) were signi cantly higher, forming around 40% of all charred plant fragments in the sample, compared to the zero presence and low density of parenchyma fragments in the control sample and non-combustion event samples. This result attests to the likelihood of these samples coming from an anthropogenic combustion event and that their presence was not accidentally tracked or trodden into the cave, where the density is likely to have been more random and less spatially concentrated.
That these combustion events may have been used for cooking food is supported by the presence of other anthropogenic evidence, including vertebrate remains and three lithic fragments, the latter also coated in burned sediment and ash.
Finally, the evidence of the fragments themselves. The 'folded' appearance of the root and tuber cell structure is the unique result of the starchy tissue being pounded when fresh. This pounded tissue was then mixed with pounded seeds and cooked. Only humans are known to cook and use this multi-step processing prior to cooking 21 .

Analysis Of Starchy Plant Tissue -Parenchyma
Starch is the means by which all green plants store energy. Plants store starch in the form of granules in tissue called parenchyma, which can be found in leaves, stems and seed endosperm. For many plants the energy reserve to feed new plant growth comes from starch stored in underground storage organs, such as underground stems (tubers), rhizomes and roots 22,23 . The parenchyma found in underground storage organs has a number of distinct morphological and identifying characteristics: the cells have a distinct organisation structure; the cells are different from the elongated bres, tracheids and vessels in wood fragments; rhizomes and tubers can have secretory cavities (that can contain essential oils such as the avour in ginger or secondary metabolites as a defence mechanism) and lateral or terminal buds and detachment scars on the endermis; and distinct organisation of the vascular structure 24,25 . These features allow the identi cation of cooked starchy plants 26 .
The moist conditions at Blombos Cave are not conducive to the preservation of plant remains by desiccation. The parenchyma tissue from these burned samples is preserved as carbon. Charred parenchyma has also recently been found at Klasies River dating to 120 kya 26 and at Border Cave dating to 170 kya 27 and form testament to a deep history of human cooking and eating starchy plant foods.

Results
The full archaeobotanical methods for analysis and modern reference collection creation can be found in the supplementary information.
The parenchyma fragments recovered from both samples as a percentage of all charred fragments is high, 41% and 43% respectively (see S.I.). The high density of parenchyma suggests that these fragments came from anthropogenic combustion event, albeit not intact. The evidence of processed and mixed starchy plant foods within these charred contexts is explained by human cooking of starchy plants.
The charred tuber parenchyma identi ed from sample 1 is rounded and indicates it has been subjected to mechanical friction, supporting the hypothesis that the context had been subject to either trampling or cleaning 25 (Fig. 2A). Fragment 2B is an example of rhizome pith parenchyma that had been cooked from fresh where the cells have been deformed by escaping steam (Fig. 2B) 25 . Eleven fragments from both samples displayed this cell structure. The evidence that they were cooked while fresh and moist would indicate that these rhizomes and tubers were neither used for bedding or fuel but were plant foods brought into the cave by humans.
Poorly preserved and mechanically stressed or broken fragments are common throughout the Blombos samples. The poor preservation is often a result of phosphatisation (Fig. 2C) and mechanical breakage caused by movement with the sediment and trampling (Fig. 2D).
The broken parenchyma cells (Fig. 2D)   Comparative cell wall structures were found in the samples from Blombos Cave. Parenchymatous tissue from the tubers in the experiment in Fig. 3 were rst pounded when the tuber was raw and moist and then cooked 28 . The result indicates that when fresh roots and tubers are pounded, the cell walls of the parenchyma do not break but fold inwards. This process could only occur whilst fresh. Once cooked, the cell walls become brittle and when broken, such as by trampling, the parenchyma breaks, exhibiting straight edges across cells at random.
In the experiment it was noted that heating alone did not soften sea club rush tubers, although heating caused the starch granules to swell and gelatinise, a process that makes starch more susceptible to alpha-amylase breakdown during digestion. In fact, boiling for any length of time did not appear to soften the tissue, until pounded.
The micrographs in Figs. 4A and 4B exhibit evidence of the parenchyma having been processed by pounding. The cell walls appear 'folded' rather than sharp-edged and broken or decomposed by phosphatisation. Sample 1 has 27 fragments and Sample 2 has 32 fragments that exhibit signs of pounding processing prior to cooking. Figures 4C and 4D provide evidence of ruptured seeds combined with the pounded parenchyma. Six burned fragments in total exhibit pounded parenchyma with seed inclusions. The seed testa morphology appears similar in each case with testa cells measuring 10-18 µm. Seeds rarely rupture unless they are ground or pounded 29 and are often found whole and capable of identi cation to species from archaeological latrines and sewers, e.g. the Carco V sewer at a Roman site in the Bay of Naples 30 . The evidence from Blombos cave suggests mixing and cooking of plant foods that include pounded tubers and pounded seeds for food and avour.

Discussion
Although only a few thousand years apart, these hearths were formed during different climate cycles, 85 kya corresponding with a cold stadial and 82 kya with a warmer inter-stadial but each representing a persistence in behaviour (Fig. 5). When combined with the ndings of cooked roots and tubers from intact hearths from Klasies River, South, 26 , the cooking and consumption of starchy plants appears to have persisted during signi cant oscillations in climate (Fig. 5). These variations in climate would have meant cold, wet, arid and/or extreme heat.
So whilst lithic technologies such as Howiesons Poort and Still Bay may be climate-mediated responses to ecological changes 32-34 , the foraging for, cooking and processing of starchy plants appears to have persisted.
Plant processing methods, whether grinding, pounding, grating, soaking/leaching, fermenting or boiling/roasting transforms the bioavailability for humans of the carbohydrates found in starchy plants [35][36][37][38][39][40] . These processes essentially start the digestion process externally. They can also, in some species, neutralise some of the irritants and secondary compounds in plants, rendering them safer and more digestible. Different processes are required for different plants, as the physical and chemical composition of plants mean they respond in different ways to processing 36 . Humans have a wide range of processing methods that they can apply either individually or in combination. The evidence presented here represents a complex multi-step processing sequence not previously observed in the archaeological record.
The discovery of mixed plant foods, with pulverised seeds forming an occasional ingredient in a pounded tuber-like matrix suggests that there may be taste/ avour preferences being expressed, beyond just the meeting of subsistence needs. The seeds look to be similar, although their ruptured form inhibits identi cation to species, but their regular appearance throughout the assemblage, incorporated in parenchymous matrices, suggests a taste preference.
Before the use of re, exploitation of starchy plant foods would have been limited 21,36 . Hardy et al. (2015) argue that humans are biologically adapted to a starch diet, with the brain, red blood cells and reproductive organs dependent upon glucose. A carbohydrate diet would have been essential to provide the energy needed for the morphological changes that evolve in the genus Homo, such as larger brains, longer limbs and larger babies 37 . This argument is consistent with evidence for the duplication in our salivary amylase genes in Homo sapiens 41 . It also indicates an early adaptation to a diet that included a larger proportion of starch 42 . The ability of humans to adapt their generalist diet to ecological niches has allowed them to colonise most biomes across the planet 43 , and yet the ecological intelligence required to accomplish this may have been underestimated 44 . The botanical and ecological knowledge to identify safe plant species, in a seasonal botanical landscape, not only as a food itself but in its ability to attract animals 45 , the technical knowledge to know which and how many processing steps will make them edible and the culinary skill to mix plant foods for avour is possibly another measure of what it is to be human.

Conclusion
We offer evidence of the changed morphology of charred plant remains as the earliest multi-step plant processing and cooking, along with the mixing of plants from levels dated to 85kya and 81kya. This evidence comes from two combustion events from the M2 lower phase of the MSA levels in Blombos Cave, South Africa; each level representing very different climatic contexts. Whilst hunting technologies and strategies have changed in response to varying ecologies, these results would suggest that the hunter-gatherers at Blombos Cave retained the ability to nd starchy plants that they were able to make digestible by processing. This persistent behaviour of processing, combining and cooking foraged starchy plants in the form of a 'recipe', we suggest, is a signi cant addition to the suite of modern human behavioural attributes.

Competing Interests
None of the authors has any competing interests Author Contributions CL: Conception; design; data acquisition; analysis; interpretation of data