Knowledge From the Afrikan World, For the World

PMC Articles

Agricultural diversification in West Africa: an archaeobotanical study of the site of Sadia (Dogon Country, Mali)

PMCID: PMC7937602

PMID: 33758626

Abstract

While narratives of the spread of agriculture are central to interpretation of African history, hard evidence of past crops and cultivation practices are still few. This research aims at filling this gap and better understanding the evolution of agriculture and foodways in West Africa. It reports evidence from systematic flotation samples taken at the settlement mounds of Sadia (Mali), dating from 4 phases (phase 0=before first–third century AD; phase 1=mid eighth–tenth c. AD; phase 2=tenth–eleventh c. AD; phase 3=twelfth–late thirteenth c. AD). Flotation of 2200 l of soil provided plant macro-remains from 146 archaeological samples. As on most West African sites, the most dominant plant is pearl millet ( Pennisetum glaucum ). But from the tenth century AD, sorghum ( Sorghum bicolor ) and African rice ( Oryza glaberrima ) appear in small quantities, and fonio ( Digitaria exilis ) and barnyard millet/hungry rice ( Echinochloa sp.), sometimes considered weeds rather than staple crops, are found in large quantities. Some samples also show remains of tree fruits from savannah parklands, such as baobab ( Adansonia digitata ), marula ( Sclerocarya birrea ), jujube ( Ziziphus sp.), shea butter ( Vittelaria paradoxa ) and African grapes ( Lannea microcarpa ). Fonio and Echinochloa sp. cultivation appears here to be a later addition that helped to diversify agriculture and buffer against failures that might affect the monoculture of pearl millet. This diversification at the end of the 1st millennium AD matches with other evidence found in West Africa.

Full Text

Food and drink form a foremost concern of all societies. Much of human energy is spent fuelling nutritional needs and tastes. Productivity of agriculture may ultimately determine population density, community size, and rate at which growing populations expand through migration (e.g. Fuller et al. 2019). Beyond subsistence needs, food is also crucial for structuring social relationships, reinforcing cultural identity and marking the calendar through seasonal feasts (Hastorf 2016: 142ff; Twiss 2019: 98ff). Feasting event and beer production are usually required by social conventions, and from African ethnographic examples it is estimated that 15–20% of agricultural production goes towards such productions (Dietler 2001: 81–82). Archaeobotanical evidence provide empirical data on past productive systems, from which estimates of the productive capacity and the potential raw materials for daily cuisine and feasts can be reconstructed (Mercuri et al. 2018a; Champion and Fuller 2019).
The aim of this paper is to contribute new empirical evidence that documents some of the evolution of West African food traditions during the past two millennia through the analysis of an important assemblage of archaeobotanical samples coming from the excavations of the stratified and well-dated settlement site of Sadia in Mali (Fig. 1). A key issue is the timing and extent of agricultural diversification in Africa, as crops such as African rice (Oryza glaberrima Steud.), fonio (Digitaria exilis (Kippist) Stapf.), cowpea (Vigna unguiculata) and sorghum (Sorghum bicolor (L.) Moench) helped to expand the agricultural repertoire beyond the Neolithic reliance on pearl millet (Pennisetum glaucum) (Kahlheber and Neumann 2007; Champion and Fuller 2018a; Kay et al. 2019).

Map of the sites cited in the text (based on the database of Champion 2019)
Archaeological work was part of an extensive interdisciplinary study of the Dogon Country conducted within the project ‘Human population and Palaeoenvironment in Africa’ under the leadership of E. Huysecom (Huysecom et al. 2004; Mayor 2011 b; Mayor et al. 2005, 2014; Ozainne 2013; Ozainne et al. 2017). Extensive surveys and small-scale test pits were first conducted along the Guringin valley (Loukou et al. 2013), and then two campaigns of excavations took place at the settlement site of Sadia in 2010 and 2011, the first one focusing on a vertical exploration of four mounds, and the second one on a spatial exposure at the summit of the main mound (Huysecom et al. 2011, 2012, 2015).
In the region, modern rainfall reaches 500–600 mm per annum, but probably varied in the past and was higher during the time of Sadia’s occupation (Mayor et al. 2005; Garnier et al. 2015). The environment is today a largely open Sahelian grassland with scattered trees and shrubs. A long history of pastoralism has encouraged high proportions of spiny taxa like Acacia, while selective cutting, tree management and possibly planting has helped to create a parkland of many useful tree species like Faidherbia albida, Vitellaria paradoxa, Adansonia digitata, Borassus aethiopum, Cordyla pinnata, Tamarindus indica and Sclerocarya birrea (Boffa 1999; Höhn and Neumann 2012; Taïbi 2019). Several of these trees produce edible fruits.
Traditional local agriculture focuses on pearl millet, sorghum and minor millets such as fonio as carbohydrate staples, with various vegetables and pulses, such as cowpea (Vigna unguiculata (L.) Walp.), as companion crops. Occasional wild rice (e.g. Oryza barthii) and various other wild grasses gathering has been documented in the region, namely during periods of starvation or bad success of agriculture (Selleger 2014; personal inquiries). Currently, rice cultivation is mostly concentrated to the wetter setting in the Niger valley, such as the Inland Delta, and in the Sourou valley at the border with Burkina Faso. However, rice is also cultivated in the Dogon country, but represents only a very small contribution to local agriculture.
A total of 27 radiocarbon dates analysed with Bayesian modelling has enabled us to build a chrono-cultural framework of the site of Sadia, extending from about 750 to 1300 AD (Huysecom et al. 2011, 2012, 2015). The mounds’ development is preceded by a pre-tell occupation (phase 0) in a yellow sandy layer dated from a single radiocarbon date (1895±35 BP, ETH-40328) as a terminus ante quem of first to third century AD. The tell occupation itself includes three main phases (Fig. 2):

Synthesised transects showing the extension of each phase in the four investigated mounds of Sadia (vertical scale magnified 7.5 times; From Huysecom et al. 2015:12)
Extensive excavations of 143 m2 conducted on the top of mound I allowed a better knowledge of the architectural characteristics of phase 3 (Figs. 3 and 4). Around 200 structures were found and revealed domestic units composed of circular rooms and spaces of different shapes limited by strait walls. Some of the circular rooms are too small for houses and were probably related to specific activities such as storing or cooking. Most of the buildings were made with mud bricks. Occupation floors, post holes, refuse pits, hearths and two burials of children were also found in deeper layers (Huysecom et al. 2015).

Plan of Sadia mound I extensive excavation. Overview of architectural features of phase 3 (from Huysecom et al. 2015:15)
Animal remains studied by Louis Chaix (Museum of Natural History in Geneva) are mainly based on domesticates. Caprine remains are the largest part of the livestock (51%) with goat (Capra) being more important than sheep (Ovis). Cattle (Bos Taurus) represent around 30% of the animal remains. Donkeys (Equus asinus) and chickens (Gallus gallus) are present in small quantities. Wild animals (<5%) are represented by varieties of gazelles, giant rats, birds and reptiles (crocodiles, frogs and snakes). Fish are well-represented by species found in floodplains and temporary ponds. Diachronic analysis has shown a reduction in domesticates with a decrease of cattle that parallels an increase of caprines through the sequence (Huysecom et al. 2015).
Besides objects of local production like ceramics, iron objects or grindstones, some goods from medium- and long-distance trade were also found, such as copper jewellery and some carnelian and glass beads (Huysecom et al. 2015). Chemical analysis of one glass bead (eleventh–thirteenth c. AD) indicated a composition like the ones of beads discovered in the Iberian Peninsula and Southern Africa, but with an origin of the glass still unknown. These results highlight the trade connections between the Seno plain, the Niger Bend and the major trans-Saharan trade routes (Truffa Giachet 2019; Truffa Giachet et al. 2020).
Regional surveys have shown that many settlements of similar size were established in the Seno plain, suggesting the development of non-centralised agro-pastoral societies during a peaceful period. The turn of the thirteenth to the fourteenth century AD witnessed a large abandonment of the sites in the region, including Sadia, but also in most of West Africa. From abrupt climate changes to war, invasion (i.e. Songhay) and diseases (i.e. the plague), several reasons are proposed for this sub-continental settlement rupture. However, no strong evidence can confirm or refute any of the reasons yet (Huysecom et al. 2015; Champion 2019).
In total, around 2200 l of archaeological soil matrix coming from 146 samples was processed by bucket flotation directly in the field. Flotation is a method widely used to separate materials of different density, and is the most common way of recovering charred plant remains from archaeological contexts, as soil particles sink and charred plant remains float (Nesbitt 1995). Whether mechanical (using a flotation drum or machine) or manual (using buckets), the basic principle is to mix the sediment with water to wet it and break it apart. The buoyant material—including the plant macro-remains—is then poured off and collected in a fine mesh (Fuller 2008; Champion and Fuller 2019). The weight and volume of each flot was then measured, and each sample was sieved into five size fractions: >4 mm, 2–4 mm, 1–2 mm, 0.5–1 mm, <0.5 mm. The ≥0.5 mm sub-samples were sorted for charred plant elements such as seeds, fruits, and parenchyma with the aid of a low-power stereomicroscope (×10–40).
Identification of cereals, chaff, pulses, millets, weeds, nuts and fruits was made by comparing archaeological morphotypes with taxonomic specimens in the modern reference collection housed at the Institute of Archaeology, University College London, and isolate samples at the University of Geneva. Taxa identification follows the nomenclature of The Useful Plants of West Tropical Africa (Burkill 1985–2000). Photographs and descriptions of identification from a variety of published literature were also used (Fuller 2003; Fuller and Harvey 2006; Nixon 2011; Murray 2000, 2007; D’Andrea et al. 2001). For all the samples analysed, whole vegetal items were recorded by count in a list of taxa, from which tables of relative frequency were constructed, and ubiquity calculated. The relative frequency is the percentage of total seed count. It can be calculated by sample, phase, trench or site, as necessary. In addition, every taxon was logged in a ubiquity table. The ubiquity is the number of samples in which a taxon is present as a percentage of the total number of samples.
Here we present the results of our analysis of seeds from all samples, integrating a previous preliminary analysis by B. Eichhorn and S. Kahlheber including grains and wood charcoal, which was published elsewhere (Huysecom et al. 2012, 2015). Of the 146 collected samples analysed here, only five were entirely devoid of identifiable macro-botanical remains. In total, around 10,104 plant remains were recorded within 141 samples for an average density of 4.5 remains per litre of soil. Almost all the remains comprise carbonised seeds, although some silicified African rice husks (mainly lemma) were found, and these are also included in the counts in Table 1, and Figs. 5 and 6, that provide frequency and ubiquity of important species.

Pearl millet is represented by 4907 remains (chaff and grains). Pearl millet is present in 91% of the samples and constitutes 48% of the total frequency. The seeds represent 37% (1819 grains) of the millet assemblage, while chaff (bristles and involucre bases, B in Fig. 7) represent 63% (3088) (Table 1). Due to their morphological characteristics, club-shaped grains and stalked involucres, these are all determined as domesticated pearl millet (after Brunken et al. 1977; Manning et al. 2011).

A total of 968 fonio grains, 10% of the total assemblage, were found in 40% of the samples (Table 1, D and E in Fig. 7). Fonio is a small-seeded millet, around 1.2 × 0.7 × 0.5 mm. The grain form is ovoid to oblong with an ovoid scutellum, less than half the length of the grain, and a roundish hilum on the ventral surface (Stapf 1915; Portères 1976).
African rice was found in 43 samples and represented 2% of the total assemblage (Table 1). This includes 12 grains and 167 spikelet bases (C in Fig. 7) and hundreds of husks (lemma/palea fragments).
Rice spikelet bases provide a good way to differentiate wild and domesticated rice. One of the most important key effects of plant domestication is a shift from natural seed dispersal to human seed dispersal. Mature wild grains fall naturally from the plant creating a smooth, round and regular abscission scar with a small distinct vascular port on the spikelet base. On the opposite, domesticated non-shattered grain needs human threshing to fall apart. The result of this labour is to create a less symmetrical scar with a dimpled appearance (Fuller et al. 2009; Fuller and Weisskopf 2011; Ishii et al. 2010). It is observed that within the 167 spikelet bases of the Sadia dataset, 60% are domesticated and only 6.5% are wild. A further 12% are classified as immature and, due to their bad preservation, the remaining 22% could not be determined. Excluding immature and indeterminate forms, the domesticated proportion can thus be estimated as ~90%. These data indicate that rice at Sadia was from already domesticated populations, with a small presence (~10%) of wild-type or weedy rice (O. barthii).
Only 90 (less than 1% of the total assemblage) sorghum remains, mainly grains (57) but also chaff (33), were found in 18% of the samples (Table 1). The chaff is composed of husks (lemma/palea), spikelet bases (A in Fig. 7), with the torn rachilla characteristic of domesticated sorghum. Such remains probably come from the dehusking waste of race bicolor (on races, see Fuller and Stevens 2018).
In Sadia, cowpea remains are few, only 10 seeds (representing less than 1% of the total assemblage) were found. The seeds are flat with a sub-rectangular shape. The hilum is ovate and generally placed asymmetrically on the hilum edge (Fuller 2003; Fuller and Harvey 2006).
In total, 34 Roselle caryopses, representing less than 1% of the assemblage, were found. The identification was done following the criteria published by Kahlheber (2004).
In total, 479 remains of fruits, 5% of the total assemblage, were found in 52% of the samples (Table 1). Charred remains include baobab (Adansonia digitata L.), marula (Sclerocarya birrea A. Rich. syn. Poupartia birrea (A. Rich.) A. Chev.), jujube (Ziziphus sp.), shea butter (Vittelaria paradoxa Gaertn. f.), African grapes (Lannea acida A. Rich. Syn. L. microcarpa Engl. & Krause) and tamarind (Tamarindus indicus L.).
A number of seeds of known weed taxa (299 seeds for 3% of the total assemblage in 37% of the samples) were recovered (Table 1). They include carpet weeds (Trianthema sp., family Aizoaceae) and button weeds (Borreira sp./Spermococce sp., family Rubiaceae), characterised by intermediate size, large-seeded weedy grasses (Paspalum and Echinochloa), and other monocots (Commelina spp.), characterised by quite large and heavy seeds.
The 146 samples studied have provided plant remains from four archaeological phases at Sadia. The diachronic distribution of archaeobotanical remains is summarised in Table 1, and Figs. 5 and 6.
Spatial analysis may offer interesting insights into intra-site patterns that could relate to activity areas or differences in food use across the site (Fig. 8). This kind of analysis, however, is made difficult by the fact that preserved plant remains require charring for preservation. Thus, food processing, crop-processing and other activities, if waste is simply left in situ, will not be readily visible due to decomposition, unless the building has burned down. Instead, it must be recognised that the waste of food plant use activities must have been gathered up and disposed in fires after which some of this material survived charring and got redeposited nearby (Fuller et al. 2014). Nevertheless, broad patterns contrasting across building contexts or areas of a site may still be possible (e.g. Hastorf 1991).

The total assemblage distribution shows that remains are mainly found around the centre of the excavated area, while they appear in low amounts in the western part of the surface and show a more patchy pattern in the eastern part of the site. The separate analysis of the five hearth samples (small circles in Fig. 8) indicates similar proportions of the main crop remains as the other samples, with two exceptions. The first exception is Echinochloa sp. remains, which are not present inside the hearths but are often found in nearby structures. Echinochloa is also found in higher amounts in structure 7, which was interpreted as a silo pit (Huysecom et al. 2012). The second exception is fonio, which is mainly present in two specific places. One is structure 4, a pit that exclusively contained charred fonio grains that is 380 grains of the 968 grains from the total assemblage, representing 40% of the total fonio remains. The other is structure 26, a pit filled with a succession of dense ashy layers and carbonised wood and grain layers (see Fig. 4 for the numbered structures). This pit may contain refuse of the two hearths located close by. This group of structures including two hearths, one refuse pit and burnt clay bricks, full of domestic remains like animal bones and diversified charred plant remains may be interpreted as a kitchen area.
Based on modern ethnographic and agronomic observations as well as recent parallels (e.g. Portères 1976; Burnham 1980; Fuller et al. 2019; Champion 2019), the archaeobotanical remains indicate that Sadia’s inhabitants were practising primarily dry agriculture, possibly coupled with some wet rice agriculture in the later phase of the occupation. Sadia’s dry agriculture was probably composed of pearl millet and sorghum fields intercropped with cowpea. The field’s edges were probably, as it is often the case today, demarcated by sorrel (Hibiscus) cultivation. Fonio was likely grown in third or fourth years before fallowing or in areas of shallower soils. Fields were probably located in sandier soils at some distance from the Guringin river or from the eventual ponds, and the only water source was rainfall. The second type of agriculture found at Sadia was possibly related to the river itself. Indeed, the finds of African rice and Echinochloa sp. remains suggest that wet agriculture was also in practised around Sadia. Today, West African wet agricultural systems are dominated by rice with some décrue sorghum cultivation. Also, one of the most common weeds, especially in the Inland Niger Delta but also in the valley of the Sourou, only 60 km away, found growing in association with rice, is Bourgou grass, Echinochloa stagnina, which is gathered with the rice crop. Bourgou grass is also collected as fodder for livestock. However, Echinochloa sp. could be considered a cultivated crop as it may be an intended product from prepared fields, even if it lacks full domestication syndrome. It is fairly common that modern agricultural researchers consider small millet grasses as weeds rather than cultivars. This means that they have been under-researched in terms of intentional cultivation or domestication adaptations. It is worth noting that Echinochloa sp. has been domesticated several times in Asia (De Wet et al. 1983), and widely documented as gathered food grains (Harlan 1989; Yang et al. 2015). It has recently been shown to undergo selection and possible domestication also in the Middle Holocene Sahara (Mercuri et al. 2018b).
Pearl millet is one of the most common staple cereals used in Africa. Today, the seeds are generally transformed into porridge or beer (Jolly 2004; Haaland 2007; Ricquier 2014). So far, the first archaeological evidence of pearl millet consists of a couple of charred grains and seed impressions in ceramics that are dated to from ca. 2500 BC to 2000 BC and were found in the lower Tilemsi valley (Manning et al. 2011; Manning and Fuller 2014; Neumann 2018). Earlier evidence in the form of ceramic temper from the desert of Northern Mali suggest a sequence of pre-domestication cultivation starting from 4000 to 5000 BC (Fuller et al. in press). This supports domestication in or around the Tilemsi valley and subsequent spread southwards into the savannas and through the Sahel. Finds from Ounjougou and Windé Koroji south of the Niger river, and Birimi in Ghana, attest to the rapid spread of pearl millet cultivation by ca. 1700 BC (D’Andrea et al. 2001; Ozainne et al. 2009; Manning et al. 2011). Nevertheless, the possibility of a second independent domestication of pearl millet in Mauretania associated with the Tichitt-Oualata Neolithic cultures remains plausible (MacDonald et al. 2009; Manning and Fuller 2014). Regardless of the number of origins, pearl millet was established since the Neolithic as the primary staple grain in West Africa, as it continued for many areas into modern times. Sadia fits into this long-term tradition.
Concerning sorghum, genetic analyses on modern sorghum and recent archaeobotanical evidence support domestication of this cereal in the Eastern Sahel region around central/eastern Sudan (Fuller and Hildebrand 2013; Fuller and Stevens 2018). Data from sorghum chaff and spikelet impressions in ceramics indicate that morphologically domesticated forms were close to 70% the population before 3000 cal BC around the upper Atbara river east of Khartoum (Winchell et al. 2017, 2018; Barron et al. 2020), with domestication fully established at Jebel Moya, ca. 2500 BC (Brass et al. 2019), although some communities continued to use a mixture of wild and domesticated sorghum, such as at Kasala (eastern Sudan) at ca. 1850 cal BC (Beldados et al. 2018). Already between 2000 and 1700 cal BC, sorghum appears to have been introduced to India (Fuller and Stevens 2018). The first sorghum evidence in West Africa is inferred to date by around 650 cal BC from Alibori sites in North Benin (Champion and Fuller 2018a, 2018b), and sorghum is reported from sites near Lake Chad in the later First Millennium BC, such as Mege and Kursakata (Klee et al. 2004; Bigga and Kahlheber 2011; Fuller and Stevens 2018). In the Inland Niger Delta in Mali, the earliest evidence of sorghum appears around 250 cal BC in Jenne-Jeno and starts to spread in the area around 500 cal AD. It is only around the tenth century AD that sorghum is found outside the Niger river delta and banks, such as at the sites of Sadia, Tongo Maaré Diabal and Oursi. By that period, sorghum starts to be more widespread in West Africa, but always in smaller proportion than pearl millet or rice.
A very important characteristic of cowpea, from an agronomic perspective, is that it is very fast growing and relatively easy to cultivate. It is a versatile crop that can be intercropped with pearl millet and sorghum (D’Andrea et al. 2007; Fuller 2003), and its seeds and leaves can be collected several times before the plant dies. As with other Fabaceae (legumes), it fixes atmospheric nitrogen and therefore helps to boost soil nitrates. Thus, its cultivation in rotations can to a degree substitute for a fallow or lengthen the cycle between fallows.
It was during the initial pearl millet dispersion period that cowpea (Vigna unguiculata) appears to have been domesticated as an addition to agriculture, evident from finds in Ghana dated around 1700–1500 cal BC and from the Nok Culture around 800 cal BC (D’Andrea et al. 2007; Franke 2016).
In the current state of research, the earliest evidence for Oryza glaberrima comes from the Inland Niger Delta at the site of Dia. AMS dates on rice grains coming from the earliest occupation layers of this settlement mound place them in the 800–400 cal BC bracket (Murray 2007). Further chronological resolution is difficult, as this falls in the Iron Age calibration plateau, and thus a date of fifth century BC is just as probable as eighth century BC. This rice is regarded as already domesticated, on the basis of rice grain measurements, as spikelet bases were not recovered (Murray 2007). Given these data, it is not clear whether an earlier domestication process in the Inland Niger Delta should be hypothesised (as suggested by Portères 1976), or whether rice had spread to this area from another domestication zone, such as the tributaries of the Niger River to the southwest, as suggested by modern genetic studies (Meyer et al. 2016). Interestingly, the first divergence within African rice is inferred to be between western Guinea (coast and hill) races and those of the Niger basin (Meyer et al. 2016). But currently, with the exception of the site of Juffure in Gambia, dated to 1650–1900 cal AD (Gijanto and Walshaw 2014) and from Sadia (this study), all the sites with domesticated African rice are coming from within the Niger River Basin itself, including the Inland Niger Delta and river banks, like the Dendi in North Benin (Bedaux et al. 1978; Champion and Fuller 2018a, 2018b; Murray 2007).
More generally, rice is well attested during this period on several archaeological sites located within the nearby Inland Niger Delta (Dia, Jenné-Jeno, Thièl, Tato à Sanouna, Toguéré Galia, Toguere Doupwil). Therefore, its presence at Sadia could result from trade and exchange, along with other items that circulated between both parts of the Bandiagara escarpment at this time (Huysecom et al. 2015). More easily, rice could also come from the Sourou valley located around 50 km south-east of Sadia on the other side of the flat sandy Seno plain, at the current border with Burkina Faso. The Sourou Valley is a large but understudied region of rice cultivation (Zwarts et al. 2019; Bin 2009; Yaro 1998). The region forms a 560-km2 humid area where every type of rice is currently cultivated (irrigated, rainfed and mainly deep-water rice). Inhabitants from the Sadia region are still buying and consuming the Sourou rice (Huysecom and Mayor personal observations). Although we do not know the ancestry of rice cultivation in this wet region, it is thus possible that the rice found in Sadia originated from the Sourou valley, an amazingly rich rice granary region where archaeological mound sites similar to Sadia are present but were never studied (Huysecom and Mayor observation).
Finally, several variants of wild rice are found today at the foot of the cliff and on the Bandiagara Plateau, as well as in some depressions of the Seno Plain. Collection from the wild and local cultivation of rice is practiced today in the region (Selleger 2014; Huysecom and Mayor personal observations). The rice found in Sadia could therefore be the product of local cultivation, especially if climatic conditions were a bit wetter than today, as rainfed rice typically requires a minimum of 700–800 mm of rainfall (Portères 1959; Jacquot and Courtois 1987).
Weedy taxa are likely to enter the site through being harvested with crops, and then removed during various stages of crop-processing, alongside chaff (Hillman 1984; Jones 1987; Thompson 1996; Fuller and Weber 2005; Harvey and Fuller 2005). Small, light weed seeds tend to be removed early in processing with the initial threshing and winnowing, while larger and heavier seeds are removed in subsequent winnowing (or sieving) after dehusking, and those seeds closest in weight/size to crop grains must be removed by final hand-picking of processed grain.
With their intermediate size and weight, carpet weeds and button weeds can be expected to have contaminated rice spikelets after threshing and winnowing. They are both widely reported groups of weeds of dry tropical agriculture, such as in millet cultivation or in rainfed rice, both in Asia and in Africa (Fuller and Boivin 2009). Commelina spp. produce quite large and heavy seeds, so they are likely to be removed in final processing after dehusking, including final hand-picking of clean grain. They are also well-known weeds of rice cultivation of both upland rainfed and lowland irrigated systems (Moody 1989; Weisskopf et al. 2014; Portères 1976; Bezançon 1995). These three types of weed taxa would be expected to have contaminated rice spikelets (i.e. in the husk), a form in which crops were likely to be traded as the husks would protect the grains from various forms of spoilage during exchange. The high presence of these species is an argument for the local cultivation of rice in Sadia.
Large-seeded weedy grasses Paspalum and Echinochloa could be contaminants removed in the latter processing stages, but it is also possible that they were present as gathered foods. It is currently not possible to determine whether these grasses represent weeds of harvested rice, gathered (or even managed/cultivated) foods or a combination of both. These large-seeded weedy grasses cannot be used as indicators of rice cultivation. However, it is worth noting that Oryza and Echinochloa show different spatial distributions in Sadia (Fig. 8), which supports the idea of Echinochloa as an important gathered or cultivated resource in its own right.
The most common “Fonio” cereal is Digitaria exilis (commonly named true fonio or white fonio) domesticated from Digitaria longiflora (Hilu 1997; Adoukonou-Sagbadja et al. 2007). Fonio is widely cultivated across western Africa, but generally is considered a minor crop that is presently in decline. The arduousness of fonio processing due to the small size of the seeds seems to be related to its progressive disappearance (Portères 1976; Cruz et al. 2011). Moreover, the seeds require laborious dehusking, unlike pearl millet which is free-threshing and thus easily processed to clean grain for cooking or flour making (Fuller and Weber 2005). On the other hand, brewing can be achieved with some or all grain remaining in their husks, and fonio beers are common.
Economic importance of fonio lies in its ability to grow very quickly in poor, especially shallow, soils, in contrast to the deeper-rooted tall crops of sorghum and pearl millet. Nevertheless, yields are quite low. A study in Nigeria reports fonio yields between 22 and 322 kg/ha (Aliero and Morakinyo 2001), whereas pearl millet yields can be expected to be between 300 and 900 kg/ha (Fuller et al. 2019). Consequently, the interest of fonio is its ability to give reliable production in a very short time, even in low yields, on poorly cleared or shallow soils, which are marginal for pearl millet or sorghum and too dry for rice. It may be grown, for example, after a couple seasons of pearl millet or sorghum, before a fallow cycle (see, e.g. Portères 1976; Burnham 1980; Cruz et al. 2011).
Currently, fonio cultivation is declining. Boureima Tessougué, one of our informants from Dimbal (a village currently located 6 km from Sadia), explained in 2018 (pers. comm.) that cultivators are sowing less and less fonio for different reasons: the difficulty of cultivation and processing, and the lack of main d’oeuvre due to seasonal migration of young people to urban centres. They maintain the practice more for tradition, to keep seeds for the following year, than for real nutritional needs. But the importance of fonio is also symbolic in the Dogon Country, having a high significance in cosmological tales and in various rituals linked to animist cults (Dieterlen 1955; Griaule 1948).
Currently, timing of fonio domestication is unclear. However, Portères suggested that it was probably domesticated in the Inland Niger Delta (Portères 1959; Portères 1976). Its geographically widespread but patchy distribution could be taken to imply that it was once continuously distributed, and might therefore be of great antiquity (Blench 2006). However, finds have been strikingly absent from most archaeobotanical studies of either Neolithic or Iron Age sites, leading to the alternative hypothesis that it was a late, secondary domestication associated with diversification of agriculture, sometimes in a context of urbanisation (Champion and Fuller 2019). So far, the only strong regional evidence of fonio is coming from the site of Kirikongo in Burkina Faso (100–1500 cal AD), where fonio grains compose 8% of the total assemblage. However, only preliminary results are available from this site (Gallagher et al. 2016). In the current state of knowledge, the adoption of fonio in Sadia seems to occur shortly after its domestication.
The high frequency of Echinochloa in Sadia raises the possibility that this was a major food resource, although domesticated form is not known in West Africa. However, it is an important gathered resource for some current populations (Portères 1976; Harlan 1989) and namely in the Dogon Country (Selleger 2014). Echinochloa colona and Brachiaria ramosa are still largely collected on a large scale in the Niger Valley during rice harvesting (Boré 1983). Moreover, it has been found as a major grain in some archaeobotanical studies, as in Essouk in North Mali (Nixon et al. 2011), and in the Inland Niger Delta at Jenné-jeno (McIntosh 1995). Indeed, this is often a utilised grain throughout the broad West African and Sahelian regions (Harlan 1989; Mercuri et al. 2018b), as well as Asia (Yang et al. 2015).
At Sadia, due to the quantity (up to 9% of the assemblage for phase 3 and 7% for phase 2) and size of the seeds, it seems probable that they represent a gathered or even cultivated cereal. The three main Echinochloa species (E. stagnina, E. pyramidalis and E. colona) are common weeds. E. stagnina is the bourgou grass growing in deep water (up to 4 m of water) and is usually gathered at the same time as rice harvesting and as the Paspalum species that are difficult to separate from rice, like in the valley of Sourou. It is considered the most common weed for floating rice types. However, bourgou is more often gathered as fodder for animal stock than as a grain for human consumption. In the Inland Niger Delta, this plant is claimed to be the most useful of all wild plants, providing, in addition to food and drink, fodder, thatch, caulking for boats, vegetable salt after calcining which is used to make soap and indigo dye. Also, it is the most common plant (domesticated or wild) that populates the Niger River floodplain (Burkill 1985; Chevalier 1932; Dalziel 1937; Portères 1952). E. colona, jungle rice, is also a wild weedy grass commonly found in irrigation ditches or near waterways but never in deep water. More common in the Senegal valley than in the Niger valley, it is often misidentified as E. stagnina. It is used more for the grain than for the fodder. E. colona is so appreciated as a minor cereal that it is the object of a kind of proto-cultivation as in some areas like in Sierra Leone, where it is cultivated (Burkill 1985). Asian populations of this species are the presumed ancestors of domesticated Indian Sawa millet (De Wet et al. 1983). E. pyramidalis is a Sahelian grass widely gathered on river edges. It is a weed of sorghum and irrigated rice cultivation. All the species are usually prepared in a similar manner to rice and are sometimes used to produce beer (Burkill 1985; Portères 1959; Boré 1983; Dalziel 1937).
Hibiscus flowers are often used as a tea like beverage called ‘Carcadet’ or ‘Bissap’, but the seeds and leaves are also used to prepare sauces (Burkill 1997). It is certainly native to the African savannahs, but little is known of its origins as a cultivar, although Fuller and Hildebrand (2013) suggested it might have originated in the Sudan region. Currently, in West Africa, pearl millet fields and garden edges are often demarcated by Hibiscus sabdariffa cultivation (Gissima or sorrel). Presence of Hibiscus sabdariffa remains at Sadia provides some early evidence for its use.
Interestingly, archaeobotanical results from Sadia indicate that during phase 0 (before the first to third century AD) and phase 1 (mid eighth to tenth c. AD), the only cultivated crop was pearl millet, which composed around half of the botanical assemblage. However, phase 0 is clearly not connected to the three other phases, as it is characterised by ceramics decorated with composite cord-wrapped roulettes, a décor typical of the late regional Neolithic period, which ends around 500 cal BC (Ozainne 2013; Huysecom et al. 2015). Thus, phase 0 may rather be related to the end of the initial wave of agricultural diffusion from the Tilemsi Valley, mainly characterised by the cultivation of pearl millet and the presence of chaff temper and cord-wrapped rouletted décors in the pottery production, along with domesticated livestock, mainly cattle remains.
Phase 1 may coincide with a later diffusion of agricultural population originating from the Tichitt area in Southern Mauritania, through the Inland Niger Delta, bringing mound site development and social complexity (Mayor et al. 2014; Champion and Fuller 2018a; Champion 2019). This wave is accompanied by the introduction of iron production technology (Robion-Brunner 2008) and more complex agricultural systems, i.e. mainly augmentation of production and crop diversification (Champion 2019; Champion and Fuller 2018a, 2018b). At Sadia, the ceramic assemblage also shows a marked transition from phase 1 to phase 2, with an increase of braided and folded strip roulettes décors and an abrupt decrease of composite roulettes (Huysecom et al. 2015).
Phase 2 witnesses the rise of agricultural complexity brought by the addition of four new domesticated crops: rice, fonio, sorghum and cowpea. The advent of rice, which could be related to trade as well as local production, possibly inspired by agricultural practices in the Inland Niger Delta, may echo the diversification of materials used for jewellery in the cultural sequence, suggesting the development of connections with regional urban trade centres along the Niger and Bani Rivers, such as Gao and Jenné-Jeno (Huysecom et al. 2015; Mayor et al. 2014; Truffa Giachet et al. 2020).
The emergence of these new staple crops (sorghum, fonio, African rice, cowpea) and pulses suggests that the expansion of the site is based on more diversified agricultural practices. Thus, phase 2 is interpreted as a period in which demand for increased grain production was met by some expansion of agriculture into marginal soils with low labour investment through the cultivation of “catch crops” of fonio, together with gathering of wild grain from open savannah grassland and fallows (i.e. Echinochloa). Environmental conditions are thought to remain stable, and generally wetter than later or modern periods, under conditions of the Medieval Climatic Anomaly (Nash et al. 2016; Mayor et al. 2005). Locally, climatic stability from phases 1 to 3 is supported by previously published wood charcoal analyses from Sadia (Huysecom et al. 2015). In other words, increasing population indicated by site size increase may have put pressure on established agricultural production and encouraged extensification and diversification. This expansion would have been focused on predictable availability of crops rather than increasing yields through intensification, a pattern that mirrors the model for initial African domestications posited by Marshall and Hildebrand (2002).
Phase 3, which immediately precedes the site’s abandonment, coincides with the rise of fonio cultivation, corresponding to a general west African pattern visible on most sites being abandoned around 1300–1400 cal AD (Champion 2019). This suggests a possible reduction in woodland cover and increasing cultivation of more marginal soils with fonio. Assuming that Echinochloa was gathered from wild grassland stands, its decrease could also result from an increase in fonio cultivation. Keeping in mind that the yields of fonio are perhaps only half or less than equivalent areas of pearl millet cultivation, caloric staple production through fonio would have required more land.
Wood charcoal analysis done by B. Eichhorn indicated a similar pattern of agricultural expansion through time. Indeed, the presence of Prosopis africana, Terminalia sp. and Vitellaria paradoxa (Shea butternut tree) wood charcoals corroborates that phase 0 is characterised by a savannah with open areas of pearl millet fields, in which the presence of fallow is inferred from the presence of Terminalia sp. that readily grow in old fallow. Prosopis sp. are characteristic of a savannah with limited agricultural influence. Wood charcoals from phase 1 seem to indicate an extension of the cultivation area with the rise of Shea butternut tree and decrease of Prosopis africana. During phases 2 and 3, the tree and shrub macro-remains dwindle, coinciding with a diminution of the useful tree species (Vitex sp., Vitellaria sp.) present in the wood charcoals. Moreover, the rise of old and young fallow species would indicate that the total land area under cultivation increased at this time, with more field area devoted to agriculture, some left uncultivated as long fallow, allowing for increased regeneration (Detarium microcarpa, Pterocarpus erinaceus, Terminalia etc.), and other areas with short fallow periods (Guiera senegalesis) (Huysecom et al. 2012, 2015).
Not far from Sadia, on the Bandiagara plateau, the Ounjougou area provides a great number of Holocene archaeological sites excavated during the same project (Ozainne et al. 2014, 2017; Fig. 9). None of these sites were systematically sampled and analysed for macro-archaeobotanical purposes. Nevertheless, direct evidence of domesticated pearl millet cultivation dated to the 2nd millennium BC was found at the site of Varves–Ouest. Two pearl millet grains were directly dated to 1981–1494 cal BC (95.4% calibration) or 1692–1533 cal BC (Bayesian HPD) (Erl-9196; 3416±109 BP), and to 1628–976 cal BC (95.4% calibration) or 1285–1055 cal BC (Bayesian HPD) (Erl-9197; 3078±131 BP) (Ozainne et al. 2009). Moreover, bilobate phytoliths (characteristic of pearl millet phytoliths) from the Yamé River deposit dated to 2600–2300 BC may constitute indirect evidence of pearl millet cultivation, several hundred years before the appearance of its charred grains in the archaeological record (Garnier et al. 2013). Both direct and indirect indicators of pearl millet agriculture coincide with a radical change in ceramic tradition associated with the presence of numerous grinding stones and the appearance of settlement structures (Huysecom et al. 2004; Ozainne et al. 2009; Eichhorn and Neumann 2014). Such material culture and the presence of domesticated pearl millet coincide with contemporaneous similar finds at Windé Koroji around 2200–1900 cal BC (Macdonald et al. 2017). In the Ounjougou region, settlement mound sites occur from around 800 BC. From around 400 cal BC to 300 cal AD, a noticeable occupational decrease is noted on the Bandiagara plateau and the Seno Plain, the only evidence of human activities being recorded in a few caves of the Bandiagara cliff (Mayor et al. 2005, 2014; Huysecom et al. 2015; Fig. 9). The final Holocene (300–1500 cal AD) is marked by the growth of mound sites into larger mounds on the Seno plain, such as Sadia and the appearance of iron production sites on the plateau (Ozainne et al. 2017; Robion-Brunner 2008). The apogee of Sadia, around 900–1250 AD, ‘corresponds to the establishment of networks of villages, self-reliant in food resources and everyday consumption goods, alongside the first urban settlements along the Niger (e.g. Jenné-Jeno)’ (Ozainne et al. 2017: 223).

Secondly, the utilisation of new soil (more marginal soil) with the adoption of new cultivars (cowpea, fonio, Echinochloa) could also result from growing insecurity (ecological and social) that pushed people to explore agricultural risk buffering. Indeed, Sadia’s growth and abandonment coincide with a more general trend observed in Sahelian West Africa where most of the sites collapsed between 1200 and 1500 AD (Mayor et al. 2005; Mayor 2011). From abrupt climate changes, to war (i.e. the Songhay invasion) and diseases (i.e. the plague the ravaged Europe in the 1300s), several reasons are proposed for this sub-continental settlement rupture (Breunig 2013; Macdonald 2013; McIntosh 1993, 2000). However, no strong evidence can confirm or refute any of these potential causes. Nevertheless, all these factors could have contributed to insecurity/instability that pushed towards both more specialised and more diversified agriculture. Although there is a lack of high precision climatic proxies near Sadia, the wider western African region appears to have experienced wetter conditions during the Medieval Climatic Anomaly (900–1250 AD) after which aridity set in (Nash et al. 2016). This could plausibly contribute to Sadia’s decline. In contrast, whereas a role for the plague pandemic in west Africa for abandonments of many sites from ca. 1350–1500 AD has been recently explored (Chouin 2018; Gallagher and Dueppen 2018), this is later than the abandonment of Sadia (ca. 1300 AD). Nevertheless, some inherent declines in soil fertility as result of agricultural expansion and accumulating soil exhaustion over the course of phases 2 and 3 could also have contributed to the site’s decline.
The absence of agricultural diversification at Tongo Maaré Diabal is also visible through isotope analyses. Recent carbon and nitrogen isotope analysis on human teeth from people buried in the Bandiagara escarpment indicate diets with a major input of C4 foods (e.g. pearl millet) in the region of Tongo Maaré Diabal, and more diversified diets mixing C3 (e.g. rice and cowpea) and C4 foods in the southern part of the Bandiagara escarpment, close to Sadia. There, a temporal shift from a C4 food system to a mix system (C3 and C4) is also observed (Dlamini et al. 2019) and is convergent with our archaeobotanical results (Fig. 6, Table 1).
In the same way, nitrogen isotopes on pearl millet grains from Tongo Maaré Diabal show a lack of increase in pearl millet δ15N values during the 650-year occupation of the site, implying that manure/household waste was added only sporadically to pearl millet (Styring et al. 2019). In other words, manuring was not a normal practice for continued annual processing, suggesting instead fallowing cycles. This continuity in soil maintenance practice reflects the general lack of change in the botanical and charcoal assemblages.
Thirdly, to the east, in the Oursi region in Burkina Faso, archaeobotanical evidence is available since the Late Stone Age. Late Stone Age (Neolithic) Oursi inhabitants were probably seasonal agro-pastoralists, who were coming into this zone during the dry season and hunting wild animals and cultivating pearl millet. From 400 cal BC, permanent settlement mounds started to appear and grow. Economic subsistence was based on pearl millet cultivation. Cowpea was added to this agricultural system around 1 cal BC/AD, and sorghum around 500 cal AD, but in very low quantity. Throughout the sequence, fruit and wood charcoal remains of Adansonia digitata, Sclerocarya birrea, Balanites aegyptiaca and Vitellaria paradoxa suggest the presence of an agro-parkland system. Charred wood results illustrate that cultivation and settlement activities intensified considerably during the course of the occupation. For example, the system of fields and fallows, typical for this cultural landscape, replaced the natural Acacia savannah on the dune, as evidenced by the decrease of Acacia and the increase of Combretaceae charcoals (Höhn and Neumann 2012).
Finally, to the south, the site of Kirikongo also provides archaeobotanical data. This settlement site is composed of mounds representing household compounds. This site complex was founded around 100 cal AD and abandoned like many other sites around 1400–1500 cal AD, possibly including the impact of plague (Gallagher and Dueppen 2018). Inhabitants were performing a large array of economic tasks, from herding cattle, producing and forging iron, potting and cultivating. The most ubiquitous and numerically frequent plant taxon at the site is Shea Butter nut (Vitellaria paradoxa). The large quantity of carbonised shea fragments is attributed to the use of parching ovens, one of the steps in Shea butter production (Gallagher et al. 2016). Agricultural activities were based on pearl millet and fonio cultivation. Inhabitants were also collecting wild grasses and arboricultural products (baobab and shea butter). Archaeobotanical samples from the different settlement mounds is not fully published but initial results attest to the presence of sorghum from phase red II (1100–1260 cal AD) (Gallagher et al. 2016).

Sections

"[{\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR45\", \"CR57\", \"CR112\", \"CR31\", \"CR86\", \"CR22\"], \"section\": \"Introduction\", \"text\": \"Food and drink form a foremost concern of all societies. Much of human energy is spent fuelling nutritional needs and tastes. Productivity of agriculture may ultimately determine population density, community size, and rate at which growing populations expand through migration (e.g. Fuller et al. 2019). Beyond subsistence needs, food is also crucial for structuring social relationships, reinforcing cultural identity and marking the calendar through seasonal feasts (Hastorf 2016: 142ff; Twiss 2019: 98ff). Feasting event and beer production are usually required by social conventions, and from African ethnographic examples it is estimated that 15\\u201320% of agricultural production goes towards such productions (Dietler 2001: 81\\u201382). Archaeobotanical evidence provide empirical data on past productive systems, from which estimates of the productive capacity and the potential raw materials for daily cuisine and feasts can be reconstructed (Mercuri et al. 2018a; Champion and Fuller 2019).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"Fig1\", \"CR71\", \"CR20\", \"CR69\"], \"section\": \"Introduction\", \"text\": \"The aim of this paper is to contribute new empirical evidence that documents some of the evolution of West African food traditions during the past two millennia through the analysis of an important assemblage of archaeobotanical samples coming from the excavations of the stratified and well-dated settlement site of Sadia in Mali (Fig. 1). A key issue is the timing and extent of agricultural diversification in Africa, as crops such as African rice (Oryza glaberrima Steud.), fonio (Digitaria exilis (Kippist) Stapf.), cowpea (Vigna unguiculata) and sorghum (Sorghum bicolor (L.) Moench) helped to expand the agricultural repertoire beyond the Neolithic reliance on pearl millet (Pennisetum glaucum) (Kahlheber and Neumann 2007; Champion and Fuller 2018a; Kay et al. 2019).\\n\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR19\"], \"section\": \"\", \"text\": \"Map of the sites cited in the text (based on the database of Champion 2019)\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR61\", \"CR80\", \"CR81\", \"CR82\", \"CR96\", \"CR99\", \"CR73\", \"CR62\", \"CR63\", \"CR64\"], \"section\": \"Localisation and current environment\", \"text\": \"Archaeological work was part of an extensive interdisciplinary study of the Dogon Country conducted within the project \\u2018Human population and Palaeoenvironment in Africa\\u2019 under the leadership of E. Huysecom (Huysecom et al. 2004; Mayor 2011 b; Mayor et al. 2005, 2014; Ozainne 2013; Ozainne et al. 2017). Extensive surveys and small-scale test pits were first conducted along the Guringin valley (Loukou et al. 2013), and then two campaigns of excavations took place at the settlement site of Sadia in 2010 and 2011, the first one focusing on a vertical exploration of four mounds, and the second one on a spatial exposure at the summit of the main mound (Huysecom et al. 2011, 2012, 2015).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR81\", \"CR50\", \"CR10\", \"CR60\", \"CR108\"], \"section\": \"Localisation and current environment\", \"text\": \"In the region, modern rainfall reaches 500\\u2013600 mm per annum, but probably varied in the past and was higher during the time of Sadia\\u2019s occupation (Mayor et al. 2005; Garnier et al. 2015). The environment is today a largely open Sahelian grassland with scattered trees and shrubs. A long history of pastoralism has encouraged high proportions of spiny taxa like Acacia, while selective cutting, tree management and possibly planting has helped to create a parkland of many useful tree species like Faidherbia albida, Vitellaria paradoxa, Adansonia digitata, Borassus aethiopum, Cordyla pinnata, Tamarindus indica and Sclerocarya birrea (Boffa 1999; H\\u00f6hn and Neumann 2012; Ta\\u00efbi 2019). Several of these trees produce edible fruits.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR105\"], \"section\": \"Localisation and current environment\", \"text\": \"Traditional local agriculture focuses on pearl millet, sorghum and minor millets such as fonio as carbohydrate staples, with various vegetables and pulses, such as cowpea (Vigna unguiculata (L.) Walp.), as companion crops. Occasional wild rice (e.g. Oryza barthii) and various other wild grasses gathering has been documented in the region, namely during periods of starvation or bad success of agriculture (Selleger 2014; personal inquiries). Currently, rice cultivation is mostly concentrated to the wetter setting in the Niger valley, such as the Inland Delta, and in the Sourou valley at the border with Burkina Faso. However, rice is also cultivated in the Dogon country, but represents only a very small contribution to local agriculture.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR62\", \"CR63\", \"CR64\", \"Fig2\"], \"section\": \"Chrono-stratigraphy\", \"text\": \"A total of 27 radiocarbon dates analysed with Bayesian modelling has enabled us to build a chrono-cultural framework of the site of Sadia, extending from about 750 to 1300 AD (Huysecom et al. 2011, 2012, 2015). The mounds\\u2019 development is preceded by a pre-tell occupation (phase 0) in a yellow sandy layer dated from a single radiocarbon date (1895\\u00b135 BP, ETH-40328) as a terminus ante quem of first to third century AD. The tell occupation itself includes three main phases (Fig. 2):\\n\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR64\"], \"section\": \"\", \"text\": \"Synthesised transects showing the extension of each phase in the four investigated mounds of Sadia (vertical scale magnified 7.5 times; From Huysecom et al. 2015:12)\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"Fig3\", \"Fig4\", \"CR64\"], \"section\": \"Important features\", \"text\": \"Extensive excavations of 143 m2 conducted on the top of mound I allowed a better knowledge of the architectural characteristics of phase 3 (Figs. 3 and 4). Around 200 structures were found and revealed domestic units composed of circular rooms and spaces of different shapes limited by strait walls. Some of the circular rooms are too small for houses and were probably related to specific activities such as storing or cooking. Most of the buildings were made with mud bricks. Occupation floors, post holes, refuse pits, hearths and two burials of children were also found in deeper layers (Huysecom et al. 2015).\\n\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR64\"], \"section\": \"\", \"text\": \"Plan of Sadia mound I extensive excavation. Overview of architectural features of phase 3 (from Huysecom et al. 2015:15)\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR64\"], \"section\": \"Important features\", \"text\": \"Animal remains studied by Louis Chaix (Museum of Natural History in Geneva) are mainly based on domesticates. Caprine remains are the largest part of the livestock (51%) with goat (Capra) being more important than sheep (Ovis). Cattle (Bos Taurus) represent around 30% of the animal remains. Donkeys (Equus asinus) and chickens (Gallus gallus) are present in small quantities. Wild animals (<5%) are represented by varieties of gazelles, giant rats, birds and reptiles (crocodiles, frogs and snakes). Fish are well-represented by species found in floodplains and temporary ponds. Diachronic analysis has shown a reduction in domesticates with a decrease of cattle that parallels an increase of caprines through the sequence (Huysecom et al. 2015).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR64\", \"CR110\", \"CR111\"], \"section\": \"Important features\", \"text\": \"Besides objects of local production like ceramics, iron objects or grindstones, some goods from medium-\\u2009and long-distance trade were also found, such as copper jewellery and some carnelian and glass beads (Huysecom et al. 2015). Chemical analysis of one glass bead (eleventh\\u2013thirteenth c. AD) indicated a composition like the ones of beads discovered in the Iberian Peninsula and Southern Africa, but with an origin of the glass still unknown. These results highlight the trade connections between the Seno plain, the Niger Bend and the major trans-Saharan trade routes (Truffa Giachet 2019; Truffa Giachet et al. 2020).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR64\", \"CR19\"], \"section\": \"Important features\", \"text\": \"Regional surveys have shown that many settlements of similar size were established in the Seno plain, suggesting the development of non-centralised agro-pastoral societies during a peaceful period. The turn of the thirteenth to the fourteenth century AD witnessed a large abandonment of the sites in the region, including Sadia, but also in most of West Africa. From abrupt climate changes to war, invasion (i.e. Songhay) and diseases (i.e. the plague), several reasons are proposed for this sub-continental settlement rupture. However, no strong evidence can confirm or refute any of the reasons yet (Huysecom et al. 2015; Champion 2019).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR93\", \"CR36\", \"CR22\"], \"section\": \"Material and methods\", \"text\": \"In total, around 2200 l of archaeological soil matrix coming from 146 samples was processed by bucket flotation directly in the field. Flotation is a method widely used to separate materials of different density, and is the most common way of recovering charred plant remains from archaeological contexts, as soil particles sink and charred plant remains float (Nesbitt 1995). Whether mechanical (using a flotation drum or machine) or manual (using buckets), the basic principle is to mix the sediment with water to wet it and break it apart. The buoyant material\\u2014including the plant macro-remains\\u2014is then poured off and collected in a fine mesh (Fuller 2008; Champion and Fuller 2019). The weight and volume of each flot was then measured, and each sample was sieved into five size fractions: >4 mm, 2\\u20134 mm, 1\\u20132 mm, 0.5\\u20131 mm, <0.5 mm. The \\u22650.5\\u00a0mm sub-samples were sorted for charred plant elements such as seeds, fruits, and parenchyma with the aid of a low-power stereomicroscope (\\u00d710\\u201340).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR15\", \"CR17\", \"CR35\", \"CR37\", \"CR95\", \"CR90\", \"CR91\", \"CR26\"], \"section\": \"Material and methods\", \"text\": \"Identification of cereals, chaff, pulses, millets, weeds, nuts and fruits was made by comparing archaeological morphotypes with taxonomic specimens in the modern reference collection housed at the Institute of Archaeology, University College London, and isolate samples at the University of Geneva. Taxa identification follows the nomenclature of The Useful Plants of West Tropical Africa (Burkill 1985\\u20132000). Photographs and descriptions of identification from a variety of published literature were also used (Fuller 2003; Fuller and Harvey 2006; Nixon 2011; Murray 2000, 2007; D\\u2019Andrea et al. 2001). For all the samples analysed, whole vegetal items were recorded by count in a list of taxa, from which tables of relative frequency were constructed, and ubiquity calculated. The relative frequency is the percentage of total seed count. It can be calculated by sample, phase, trench or site, as necessary. In addition, every taxon was logged in a ubiquity table. The ubiquity is the number of samples in which a taxon is present as a percentage of the total number of samples.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR63\", \"CR64\", \"Tab1\", \"Fig5\", \"Fig6\"], \"section\": \"Results\", \"text\": \"Here we present the results of our analysis of seeds from all samples, integrating a previous preliminary analysis by B. Eichhorn and S. Kahlheber including grains and wood charcoal, which was published elsewhere (Huysecom et al. 2012, 2015). Of the 146 collected samples analysed here, only five were entirely devoid of identifiable macro-botanical remains. In total, around 10,104 plant remains were recorded within 141 samples for an average density of 4.5 remains per litre of soil. Almost all the remains comprise carbonised seeds, although some silicified African rice husks (mainly lemma) were found, and these are also included in the counts in Table 1, and Figs. 5 and 6, that provide frequency and ubiquity of important species.\\n\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"Fig7\", \"Tab1\", \"CR14\", \"CR78\"], \"section\": \"Pearl millet, Pennisetum glaucum (L.) R. Br., Poaceae\", \"text\": \"Pearl millet is represented by 4907 remains (chaff and grains). Pearl millet is present in 91% of the samples and constitutes 48% of the total frequency. The seeds represent 37% (1819 grains) of the millet assemblage, while chaff (bristles and involucre bases, B in Fig. 7) represent 63% (3088) (Table 1). Due to their morphological characteristics, club-shaped grains and stalked involucres, these are all determined as domesticated pearl millet (after Brunken et al. 1977; Manning et al. 2011).\\n\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"Tab1\", \"Fig7\", \"CR106\", \"CR102\"], \"section\": \"Fonio, Digitaria exilis (Kippist.) Staph., Poaceae.\", \"text\": \"A total of 968 fonio grains, 10% of the total assemblage, were found in 40% of the samples (Table 1, D and E in Fig. 7). Fonio is a small-seeded millet, around 1.2 \\u00d7 0.7 \\u00d7 0.5 mm. The grain form is ovoid to oblong with an ovoid scutellum, less than half the length of the grain, and a roundish hilum on the ventral surface (Stapf 1915; Port\\u00e8res 1976).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"Tab1\", \"Fig7\"], \"section\": \"African Rice, Oryza glaberrima Steud., Poaceae\", \"text\": \"African rice was found in 43 samples and represented 2% of the total assemblage (Table 1). This includes 12 grains and 167 spikelet bases (C in Fig. 7) and hundreds of husks (lemma/palea fragments).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR43\", \"CR42\", \"CR65\"], \"section\": \"African Rice, Oryza glaberrima Steud., Poaceae\", \"text\": \"Rice spikelet bases provide a good way to differentiate wild and domesticated rice. One of the most important key effects of plant domestication is a shift from natural seed dispersal to human seed dispersal. Mature wild grains fall naturally from the plant creating a smooth, round and regular abscission scar with a small distinct vascular port on the spikelet base. On the opposite, domesticated non-shattered grain needs human threshing to fall apart. The result of this labour is to create a less symmetrical scar with a dimpled appearance (Fuller et al. 2009; Fuller and Weisskopf 2011; Ishii et al. 2010). It is observed that within the 167 spikelet bases of the Sadia dataset, 60% are domesticated and only 6.5% are wild. A further 12% are classified as immature and, due to their bad preservation, the remaining 22% could not be determined. Excluding immature and indeterminate forms, the domesticated proportion can thus be estimated as ~90%. These data indicate that rice at Sadia was from already domesticated populations, with a small presence (~10%) of wild-type or weedy rice (O. barthii).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"Tab1\", \"Fig7\", \"CR40\"], \"section\": \"Sorghum, Sorghum bicolor subsp. bicolor (L.) Moench., Poaceae.\", \"text\": \"Only 90 (less than 1% of the total assemblage) sorghum remains, mainly grains (57) but also chaff (33), were found in 18% of the samples (Table 1). The chaff is composed of husks (lemma/palea), spikelet bases (A in Fig. 7), with the torn rachilla characteristic of domesticated sorghum. Such remains probably come from the dehusking waste of race bicolor (on races, see Fuller and Stevens 2018).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR35\", \"CR37\"], \"section\": \"Cowpea, Vigna unguiculata (L.) Walp., Fabaceae\", \"text\": \"In Sadia, cowpea remains are few, only 10 seeds (representing less than 1% of the total assemblage) were found. The seeds are flat with a sub-rectangular shape. The hilum is ovate and generally placed asymmetrically on the hilum edge (Fuller 2003; Fuller and Harvey 2006).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR70\"], \"section\": \"Roselle, Hibiscus sabdariffa Lin., Malvaceae\", \"text\": \"In total, 34 Roselle caryopses, representing less than 1% of the assemblage, were found. The identification was done following the criteria published by Kahlheber (2004).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"Tab1\"], \"section\": \"Fruits, tree and bushes\", \"text\": \"In total, 479 remains of fruits, 5% of the total assemblage, were found in 52% of the samples (Table 1). Charred remains include baobab (Adansonia digitata L.), marula (Sclerocarya birrea A. Rich. syn. Poupartia birrea (A. Rich.) A. Chev.), jujube (Ziziphus sp.), shea butter (Vittelaria paradoxa Gaertn. f.), African grapes (Lannea acida A. Rich. Syn. L. microcarpa Engl. & Krause) and tamarind (Tamarindus indicus L.).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"Tab1\"], \"section\": \"Weeds and/or wild gathered plants\", \"text\": \"A number of seeds of known weed taxa (299 seeds for 3% of the total assemblage in 37% of the samples) were recovered (Table 1). They include carpet weeds (Trianthema sp., family Aizoaceae) and button weeds (Borreira sp./Spermococce sp., family Rubiaceae), characterised by intermediate size, large-seeded weedy grasses (Paspalum and Echinochloa), and other monocots (Commelina spp.), characterised by quite large and heavy seeds.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"Tab1\", \"Fig5\", \"Fig6\"], \"section\": \"Diachronic analysis\", \"text\": \"The 146 samples studied have provided plant remains from four archaeological phases at Sadia. The diachronic distribution of archaeobotanical remains is summarised in Table 1, and Figs. 5 and 6.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"Fig8\", \"CR44\", \"CR56\"], \"section\": \"Spatial distribution for phase 3\", \"text\": \"Spatial analysis may offer interesting insights into intra-site patterns that could relate to activity areas or differences in food use across the site (Fig. 8). This kind of analysis, however, is made difficult by the fact that preserved plant remains require charring for preservation. Thus, food processing, crop-processing and other activities, if waste is simply left in situ, will not be readily visible due to decomposition, unless the building has burned down. Instead, it must be recognised that the waste of food plant use activities must have been gathered up and disposed in fires after which some of this material survived charring and got redeposited nearby (Fuller et al. 2014). Nevertheless, broad patterns contrasting across building contexts or areas of a site may still be possible (e.g. Hastorf 1991).\\n\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"Fig8\", \"CR63\", \"Fig4\"], \"section\": \"Spatial distribution for phase 3\", \"text\": \"The total assemblage distribution shows that remains are mainly found around the centre of the excavated area, while they appear in low amounts in the western part of the surface and show a more patchy pattern in the eastern part of the site. The separate analysis of the five hearth samples (small circles in Fig. 8) indicates similar proportions of the main crop remains as the other samples, with two exceptions. The first exception is Echinochloa sp. remains, which are not present inside the hearths but are often found in nearby structures. Echinochloa is also found in higher amounts in structure 7, which was interpreted as a silo pit (Huysecom et al. 2012). The second exception is fonio, which is mainly present in two specific places. One is structure 4, a pit that exclusively contained charred fonio grains that is 380 grains of the 968 grains from the total assemblage, representing 40% of the total fonio remains. The other is structure 26, a pit filled with a succession of dense ashy layers and carbonised wood and grain layers\\u00a0(see\\u00a0Fig. 4 for the numbered structures). This pit may contain refuse of the two hearths located close by. This group of structures including two hearths, one refuse pit and burnt clay bricks, full of domestic remains like animal bones and diversified charred plant remains may be interpreted as a kitchen area.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR102\", \"CR18\", \"CR45\", \"CR19\", \"CR29\", \"CR54\", \"CR116\", \"CR87\"], \"section\": \"Agricultural practices and food preparation\", \"text\": \"Based on modern ethnographic and agronomic observations as well as recent parallels (e.g. Port\\u00e8res 1976; Burnham 1980; Fuller et al. 2019; Champion 2019), the archaeobotanical remains indicate that Sadia\\u2019s inhabitants were practising primarily dry agriculture, possibly coupled with some wet rice agriculture in the later phase of the occupation. Sadia\\u2019s dry agriculture was probably composed of pearl millet and sorghum fields intercropped with cowpea. The field\\u2019s edges were probably, as it is often the case today, demarcated by sorrel (Hibiscus) cultivation. Fonio was likely grown in third or fourth years before fallowing or in areas of shallower soils. Fields were probably located in sandier soils at some distance from the Guringin river or from the eventual ponds, and the only water source was rainfall. The second type of agriculture found at Sadia was possibly related to the river itself. Indeed, the finds of African rice and Echinochloa sp. remains suggest that wet agriculture was also in practised around Sadia. Today, West African wet agricultural systems are dominated by rice with some d\\u00e9crue sorghum cultivation. Also, one of the most common weeds, especially in the Inland Niger Delta but also in the valley of the Sourou, only 60\\u00a0km away, found growing in association with rice, is Bourgou grass, Echinochloa stagnina, which is gathered with the rice crop. Bourgou grass is also collected as fodder for livestock. However, Echinochloa sp. could be considered a cultivated crop as it may be an intended product from prepared fields, even if it lacks full domestication syndrome. It is fairly common that modern agricultural researchers consider small millet grasses as weeds rather than cultivars. This means that they have been under-researched in terms of intentional cultivation or domestication adaptations. It is worth noting that Echinochloa sp. has been domesticated several times in Asia (De Wet et al. 1983), and widely documented as gathered food grains (Harlan 1989; Yang et al. 2015). It has recently been shown to undergo selection and possible domestication also in the Middle Holocene Sahara (Mercuri et al. 2018b).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR67\", \"CR53\", \"CR103\", \"CR78\", \"CR77\", \"CR94\", \"CR46\", \"CR26\", \"CR97\", \"CR78\", \"CR75\", \"CR77\"], \"section\": \"Pearl millet and sorghum\", \"text\": \"Pearl millet is one of the most common staple cereals used in Africa. Today, the seeds are generally transformed into porridge or beer (Jolly 2004; Haaland 2007; Ricquier 2014). So far, the first archaeological evidence of pearl millet consists of a couple of charred grains and seed impressions in ceramics that are dated to from ca. 2500 BC to 2000 BC and were found in the lower Tilemsi valley (Manning et al. 2011; Manning and Fuller 2014; Neumann 2018). Earlier evidence in the form of ceramic temper from the desert of Northern Mali suggest a sequence of pre-domestication cultivation starting from 4000 to 5000 BC (Fuller et al. in press). This supports domestication in or around the Tilemsi valley and subsequent spread southwards into the savannas and through the Sahel. Finds from Ounjougou and Wind\\u00e9 Koroji south of the Niger river, and Birimi in Ghana, attest to the rapid spread of pearl millet cultivation by ca. 1700 BC (D\\u2019Andrea et al. 2001; Ozainne et al. 2009; Manning et al. 2011). Nevertheless, the possibility of a second independent domestication of pearl millet in Mauretania associated with the Tichitt-Oualata Neolithic cultures remains plausible (MacDonald et al. 2009; Manning and Fuller 2014). Regardless of the number of origins, pearl millet was established since the Neolithic as the primary staple grain in West Africa, as it continued for many areas into modern times. Sadia fits into this long-term tradition.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR39\", \"CR40\", \"CR114\", \"CR115\", \"CR3\", \"CR12\", \"CR5\", \"CR40\", \"CR20\", \"CR21\", \"CR72\", \"CR7\", \"CR40\"], \"section\": \"Pearl millet and sorghum\", \"text\": \"Concerning sorghum, genetic analyses on modern sorghum and recent archaeobotanical evidence support domestication of this cereal in the Eastern Sahel region around central/eastern Sudan (Fuller and Hildebrand 2013; Fuller and Stevens 2018). Data from sorghum chaff and spikelet impressions in ceramics indicate that morphologically domesticated forms were close to 70% the population before 3000 cal BC around the upper Atbara river east of Khartoum (Winchell et al. 2017, 2018; Barron et al. 2020), with domestication fully established at Jebel Moya, ca. 2500 BC (Brass et al. 2019), although some communities continued to use a mixture of wild and domesticated sorghum, such as at Kasala (eastern Sudan) at ca. 1850 cal BC (Beldados et al. 2018). Already between 2000 and 1700 cal BC, sorghum appears to have been introduced to India (Fuller and Stevens 2018). The first sorghum evidence in West Africa is inferred to date by around 650 cal BC from Alibori sites in North Benin (Champion and Fuller 2018a, 2018b), and sorghum is reported from sites near Lake Chad in the later First Millennium BC, such as Mege and Kursakata (Klee et al. 2004; Bigga and Kahlheber 2011; Fuller and Stevens 2018). In the Inland Niger Delta in Mali, the earliest evidence of sorghum appears around 250 cal BC in Jenne-Jeno and starts to spread in the area around 500 cal AD. It is only around the tenth century AD that sorghum is found outside the Niger river delta and banks, such as at the sites of Sadia, Tongo Maar\\u00e9 Diabal and Oursi. By that period, sorghum starts to be more widespread in West Africa, but always in smaller proportion than pearl millet or rice.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR27\", \"CR35\"], \"section\": \"Cowpea\", \"text\": \"A very important characteristic of cowpea, from an agronomic perspective, is that it is very fast growing and relatively easy to cultivate. It is a versatile crop that can be intercropped with pearl millet and sorghum (D\\u2019Andrea et al. 2007; Fuller 2003), and its seeds and leaves can be collected several times before the plant dies. As with other Fabaceae (legumes), it fixes atmospheric nitrogen and therefore helps to boost soil nitrates. Thus, its cultivation in rotations can to a degree substitute for a fallow or lengthen the cycle between fallows.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR28\", \"CR34\"], \"section\": \"Cowpea\", \"text\": \"It was during the initial pearl millet dispersion period that cowpea (Vigna unguiculata) appears to have been domesticated as an addition to agriculture, evident from finds in Ghana dated around 1700\\u20131500 cal BC and from the Nok Culture around 800 cal BC (D\\u2019Andrea et al. 2007; Franke 2016).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR91\", \"CR91\", \"CR102\", \"CR88\", \"CR88\", \"CR51\", \"CR4\", \"CR20\", \"CR21\", \"CR91\"], \"section\": \"African rice\", \"text\": \"In the current state of research, the earliest evidence for Oryza glaberrima comes from the Inland Niger Delta at the site of Dia. AMS dates on rice grains coming from the earliest occupation layers of this settlement mound place them in the 800\\u2013400 cal BC bracket (Murray 2007). Further chronological resolution is difficult, as this falls in the Iron Age calibration plateau, and thus a date of fifth century BC is just as probable as eighth century BC. This rice is regarded as already domesticated, on the basis of rice grain measurements, as spikelet bases were not recovered (Murray 2007). Given these data, it is not clear whether an earlier domestication process in the Inland Niger Delta should be hypothesised (as suggested by Port\\u00e8res 1976), or whether rice had spread to this area from another domestication zone, such as the tributaries of the Niger River to the southwest, as suggested by modern genetic studies (Meyer et al. 2016). Interestingly, the first divergence within African rice is inferred to be between western Guinea (coast and hill) races and those of the Niger basin (Meyer et al. 2016). But currently, with the exception of the site of Juffure in Gambia, dated to 1650\\u20131900 cal AD (Gijanto and Walshaw 2014) and from Sadia (this study), all the sites with domesticated African rice are coming from within the Niger River Basin itself, including the Inland Niger Delta and river banks, like the Dendi in North Benin (Bedaux et al. 1978; Champion and Fuller 2018a, 2018b; Murray 2007).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR64\", \"CR118\", \"CR8\", \"CR117\"], \"section\": \"African rice\", \"text\": \"More generally, rice is well attested during this period on several archaeological sites located within the nearby Inland Niger Delta (Dia, Jenn\\u00e9-Jeno, Thi\\u00e8l, Tato \\u00e0 Sanouna, Togu\\u00e9r\\u00e9 Galia, Toguere Doupwil). Therefore, its presence at Sadia could result from trade and exchange, along with other items that circulated between both parts of the Bandiagara escarpment at this time (Huysecom et al. 2015). More easily, rice could also come from the Sourou valley located around 50 km south-east of Sadia on the other side of the flat sandy Seno plain, at the current border with Burkina Faso. The Sourou Valley is a large but understudied region of rice cultivation (Zwarts et al. 2019; Bin 2009; Yaro 1998). The region forms a 560-km2 humid area where every type of rice is currently cultivated (irrigated, rainfed and mainly deep-water rice). Inhabitants from the Sadia region are still buying and consuming the Sourou rice (Huysecom and Mayor personal observations). Although we do not know the ancestry of rice cultivation in this wet region, it is thus possible that the rice found in Sadia originated from the Sourou valley, an amazingly rich rice granary region where archaeological mound sites similar to Sadia are present but were never studied (Huysecom and Mayor observation).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR105\", \"CR101\", \"CR66\"], \"section\": \"African rice\", \"text\": \"Finally, several variants of wild rice are found today at the foot of the cliff and on the Bandiagara Plateau, as well as in some depressions of the Seno Plain. Collection from the wild and local cultivation of rice is practiced today in the region (Selleger 2014; Huysecom and Mayor personal observations). The rice found in Sadia could therefore be the product of local cultivation, especially if climatic conditions were a bit wetter than today, as rainfed rice typically requires a minimum of 700\\u2013800 mm of rainfall (Port\\u00e8res 1959; Jacquot and Courtois 1987).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR58\", \"CR68\", \"CR109\", \"CR41\", \"CR55\"], \"section\": \"Weeds and/or gathered seeds\", \"text\": \"Weedy taxa are likely to enter the site through being harvested with crops, and then removed during various stages of crop-processing, alongside chaff (Hillman 1984; Jones 1987; Thompson 1996; Fuller and Weber 2005; Harvey and Fuller 2005). Small, light weed seeds tend to be removed early in processing with the initial threshing and winnowing, while larger and heavier seeds are removed in subsequent winnowing (or sieving) after dehusking, and those seeds closest in weight/size to crop grains must be removed by final hand-picking of processed grain.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR38\", \"CR89\", \"CR113\", \"CR102\", \"CR6\"], \"section\": \"Weeds and/or gathered seeds\", \"text\": \"With their intermediate size and weight, carpet weeds and button weeds can be expected to have contaminated rice spikelets after threshing and winnowing. They are both widely reported groups of weeds of dry tropical agriculture, such as in millet cultivation or in rainfed rice, both in Asia and in Africa (Fuller and Boivin 2009). Commelina spp. produce quite large and heavy seeds, so they are likely to be removed in final processing after dehusking, including final hand-picking of clean grain. They are also well-known weeds of rice cultivation of both upland rainfed and lowland irrigated systems (Moody 1989; Weisskopf et al. 2014; Port\\u00e8res 1976; Bezan\\u00e7on 1995). These three types of weed taxa would be expected to have contaminated rice spikelets (i.e. in the husk), a form in which crops were likely to be traded as the husks would protect the grains from various forms of spoilage during exchange. The high presence of these species is an argument for the local cultivation of rice in Sadia.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"Fig8\"], \"section\": \"Weeds and/or gathered seeds\", \"text\": \"Large-seeded weedy grasses Paspalum and Echinochloa could be contaminants removed in the latter processing stages, but it is also possible that they were present as gathered foods. It is currently not possible to determine whether these grasses represent weeds of harvested rice, gathered (or even managed/cultivated) foods or a combination of both. These large-seeded weedy grasses cannot be used as indicators of rice cultivation. However, it is worth noting that Oryza and Echinochloa show different spatial distributions in Sadia (Fig. 8), which supports the idea of Echinochloa as an important gathered or cultivated resource in its own right.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR59\", \"CR1\", \"CR102\", \"CR25\", \"CR41\"], \"section\": \"Fonio\", \"text\": \"The most common \\u201cFonio\\u201d cereal is Digitaria exilis (commonly named true fonio or white fonio) domesticated from Digitaria longiflora (Hilu 1997; Adoukonou-Sagbadja et al. 2007). Fonio is widely cultivated across western Africa, but generally is considered a minor crop that is presently in decline. The arduousness of fonio processing due to the small size of the seeds seems to be related to its progressive disappearance (Port\\u00e8res 1976; Cruz et al. 2011). Moreover, the seeds require laborious dehusking, unlike pearl millet which is free-threshing and thus easily processed to clean grain for cooking or flour making (Fuller and Weber 2005). On the other hand, brewing can be achieved with some or all grain remaining in their husks, and fonio beers are common.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR2\", \"CR45\", \"CR18\", \"CR25\"], \"section\": \"Fonio\", \"text\": \"Economic importance of fonio lies in its ability to grow very quickly in poor, especially shallow, soils, in contrast to the deeper-rooted tall crops of sorghum and pearl millet. Nevertheless, yields are quite low. A study in Nigeria reports fonio yields between 22 and 322 kg/ha (Aliero and Morakinyo 2001), whereas pearl millet yields can be expected to be between 300 and 900 kg/ha (Fuller et al. 2019). Consequently, the interest of fonio is its ability to give reliable production in a very short time, even in low yields, on poorly cleared or shallow soils, which are marginal for pearl millet or sorghum and too dry for rice. It may be grown, for example, after a couple seasons of pearl millet or sorghum, before a fallow cycle (see, e.g. Port\\u00e8res 1976; Burnham 1980; Cruz et al. 2011).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR30\", \"CR52\"], \"section\": \"Fonio\", \"text\": \"Currently, fonio cultivation is declining. Boureima Tessougu\\u00e9, one of our informants from Dimbal (a village currently located 6 km from Sadia), explained in 2018 (pers. comm.) that cultivators are sowing less and less fonio for different reasons: the difficulty of cultivation and processing, and the lack of main d\\u2019oeuvre due to seasonal migration of young people to urban centres. They maintain the practice more for tradition, to keep seeds for the following year, than for real nutritional needs. But the importance of fonio is also symbolic in the Dogon Country, having a high significance in cosmological tales and in various rituals linked to animist cults (Dieterlen 1955; Griaule 1948).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR101\", \"CR102\", \"CR9\", \"CR22\", \"CR48\"], \"section\": \"Fonio\", \"text\": \"Currently, timing of fonio domestication is unclear. However, Port\\u00e8res suggested that it was probably domesticated in the Inland Niger Delta (Port\\u00e8res 1959; Port\\u00e8res 1976). Its geographically widespread but patchy distribution could be taken to imply that it was once continuously distributed, and might therefore be of great antiquity (Blench 2006). However, finds have been strikingly absent from most archaeobotanical studies of either Neolithic or Iron Age sites, leading to the alternative hypothesis that it was a late, secondary domestication associated with diversification of agriculture, sometimes in a context of urbanisation (Champion and Fuller 2019). So far, the only strong regional evidence of fonio is coming from the site of Kirikongo in Burkina Faso (100\\u20131500 cal AD), where fonio grains compose 8% of the total assemblage. However, only preliminary results are available from this site (Gallagher et al. 2016). In the current state of knowledge, the adoption of fonio in Sadia seems to occur shortly after its domestication.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR102\", \"CR54\", \"CR105\", \"CR11\", \"CR95\", \"CR84\", \"CR54\", \"CR87\", \"CR116\"], \"section\": \"Echinochloa\", \"text\": \"The high frequency of Echinochloa in Sadia raises the possibility that this was a major food resource, although domesticated form is not known in West Africa. However, it is an important gathered resource for some current populations (Port\\u00e8res 1976; Harlan 1989) and namely in the Dogon Country (Selleger 2014). Echinochloa colona and Brachiaria ramosa are still largely collected on a large scale in the Niger Valley during rice harvesting (Bor\\u00e9 1983). Moreover, it has been found as a major grain in some archaeobotanical studies, as in Essouk in North Mali (Nixon et al. 2011), and in the Inland Niger Delta at Jenn\\u00e9-jeno (McIntosh 1995). Indeed, this is often a utilised grain throughout the broad West African and Sahelian regions (Harlan 1989; Mercuri et al. 2018b), as well as Asia (Yang et al. 2015).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR15\", \"CR23\", \"CR28\", \"CR100\", \"CR15\", \"CR29\", \"CR15\", \"CR101\", \"CR11\", \"CR28\"], \"section\": \"Echinochloa\", \"text\": \"At Sadia, due to the quantity (up to 9% of the assemblage for phase 3 and 7% for phase 2) and size of the seeds, it seems probable that they represent a gathered or even cultivated cereal. The three main Echinochloa species (E. stagnina, E. pyramidalis and E. colona) are common weeds. E. stagnina is the bourgou grass growing in deep water (up to 4 m of water) and is usually gathered at the same time as rice harvesting and as the Paspalum species that are difficult to separate from rice, like in the valley of Sourou. It is considered the most common weed for floating rice types. However, bourgou is more often gathered as fodder for animal stock than as a grain for human consumption. In the Inland Niger Delta, this plant is claimed to be the most useful of all wild plants, providing, in addition to food and drink, fodder, thatch, caulking for boats, vegetable salt after calcining which is used to make soap and indigo dye. Also, it is the most common plant (domesticated or wild) that populates the Niger River floodplain (Burkill 1985; Chevalier 1932; Dalziel 1937; Port\\u00e8res 1952). E. colona, jungle rice, is also a wild weedy grass commonly found in irrigation ditches or near waterways but never in deep water. More common in the Senegal valley than in the Niger valley, it is often misidentified as E. stagnina. It is used more for the grain than for the fodder. E. colona is so appreciated as a minor cereal that it is the object of a kind of proto-cultivation as in some areas like in Sierra Leone, where it is cultivated (Burkill 1985). Asian populations of this species are the presumed ancestors of domesticated Indian Sawa millet (De Wet et al. 1983). E. pyramidalis is a Sahelian grass widely gathered on river edges. It is a weed of sorghum and irrigated rice cultivation. All the species are usually prepared in a similar manner to rice and are sometimes used to produce beer (Burkill 1985; Port\\u00e8res 1959; Bor\\u00e9 1983; Dalziel 1937).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR16\", \"CR39\"], \"section\": \"Hibiscus\", \"text\": \"Hibiscus flowers are often used as a tea like beverage called \\u2018Carcadet\\u2019 or \\u2018Bissap\\u2019, but the seeds and leaves are also used to prepare sauces (Burkill 1997). It is certainly native to the African savannahs, but little is known of its origins as a cultivar, although Fuller and Hildebrand (2013) suggested it might have originated in the Sudan region. Currently, in West Africa, pearl millet fields and garden edges are often demarcated by Hibiscus sabdariffa cultivation (Gissima or sorrel). Presence of Hibiscus sabdariffa remains at Sadia provides some early evidence for its use.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR96\", \"CR64\"], \"section\": \"Sadia and the agricultural diversification of the Dogon Country\", \"text\": \"Interestingly, archaeobotanical results from Sadia indicate that during phase 0 (before the first to third century AD) and phase 1 (mid eighth to tenth c. AD), the only cultivated crop was pearl millet, which composed around half of the botanical assemblage. However, phase 0 is clearly not connected to the three other phases, as it is characterised by ceramics decorated with composite cord-wrapped roulettes, a d\\u00e9cor typical of the late regional Neolithic period, which ends around 500 cal BC (Ozainne 2013; Huysecom et al. 2015). Thus, phase 0 may rather be related to the end of the initial wave of agricultural diffusion from the Tilemsi Valley, mainly characterised by the cultivation of pearl millet and the presence of chaff temper and cord-wrapped rouletted d\\u00e9cors in the pottery production, along with domesticated livestock, mainly cattle remains.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR82\", \"CR20\", \"CR19\", \"CR104\", \"CR19\", \"CR20\", \"CR21\", \"CR64\"], \"section\": \"Sadia and the agricultural diversification of the Dogon Country\", \"text\": \"Phase 1 may coincide with a later diffusion of agricultural population originating from the Tichitt area in Southern Mauritania, through the Inland Niger Delta, bringing mound site development and social complexity (Mayor et al. 2014; Champion and Fuller 2018a; Champion 2019). This wave is accompanied by the introduction of iron production technology (Robion-Brunner 2008) and more complex agricultural systems, i.e. mainly augmentation of production and crop diversification (Champion 2019; Champion and Fuller 2018a, 2018b). At Sadia, the ceramic assemblage also shows a marked transition from phase 1 to phase 2, with an increase of braided and folded strip roulettes d\\u00e9cors and an abrupt decrease of composite roulettes (Huysecom et al. 2015).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR64\", \"CR82\", \"CR111\"], \"section\": \"Sadia and the agricultural diversification of the Dogon Country\", \"text\": \"Phase 2 witnesses the rise of agricultural complexity brought by the addition of four new domesticated crops: rice, fonio, sorghum and cowpea. The advent of rice, which could be related to trade as well as local production, possibly inspired by agricultural practices in the Inland Niger Delta, may echo the diversification of materials used for jewellery in the cultural sequence, suggesting the development of connections with regional urban trade centres along the Niger and Bani Rivers, such as Gao and Jenn\\u00e9-Jeno (Huysecom et al. 2015; Mayor et al. 2014;\\u00a0Truffa Giachet et al.\\u00a02020).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR92\", \"CR81\", \"CR64\", \"CR79\"], \"section\": \"Sadia and the agricultural diversification of the Dogon Country\", \"text\": \"The emergence of these new staple crops (sorghum, fonio, African rice, cowpea) and pulses suggests that the expansion of the site is based on more diversified agricultural practices. Thus, phase 2 is interpreted as a period in which demand for increased grain production was met by some expansion of agriculture into marginal soils with low labour investment through the cultivation of \\u201ccatch crops\\u201d of fonio, together with gathering of wild grain from open savannah grassland and fallows (i.e. Echinochloa). Environmental conditions are thought to remain stable, and generally wetter than later or modern periods, under conditions of the Medieval Climatic Anomaly (Nash et al. 2016; Mayor et al. 2005). Locally, climatic stability from phases 1 to 3 is supported by previously published wood charcoal analyses from Sadia (Huysecom et al. 2015). In other words, increasing population indicated by site size increase may have put pressure on established agricultural production and encouraged extensification and diversification. This expansion would have been focused on predictable availability of crops rather than increasing yields through intensification, a pattern that mirrors the model for initial African domestications posited by Marshall and Hildebrand (2002).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR19\"], \"section\": \"Sadia and the agricultural diversification of the Dogon Country\", \"text\": \"Phase 3, which immediately precedes the site\\u2019s abandonment, coincides with the rise of fonio cultivation, corresponding to a general west African pattern visible on most sites being abandoned around 1300\\u20131400 cal AD (Champion 2019). This suggests a possible reduction in woodland cover and increasing cultivation of more marginal soils with fonio. Assuming that Echinochloa was gathered from wild grassland stands, its decrease could also result from an increase in fonio cultivation. Keeping in mind that the yields of fonio are perhaps only half or less than equivalent areas of pearl millet cultivation, caloric staple production through fonio would have required more land.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR63\", \"CR64\"], \"section\": \"Sadia and the agricultural diversification of the Dogon Country\", \"text\": \"Wood charcoal analysis done by B. Eichhorn indicated a similar pattern of agricultural expansion through time. Indeed, the presence of Prosopis africana, Terminalia sp. and Vitellaria paradoxa (Shea butternut tree) wood charcoals corroborates that phase 0 is characterised by a savannah with open areas of pearl millet fields, in which the presence of fallow is inferred from the presence of Terminalia sp. that readily grow in old fallow. Prosopis sp. are characteristic of a savannah with limited agricultural influence. Wood charcoals from phase 1 seem to indicate an extension of the cultivation area with the rise of Shea butternut tree and decrease of Prosopis africana. During phases 2 and 3, the tree and shrub macro-remains dwindle, coinciding with a diminution of the useful tree species (Vitex sp., Vitellaria sp.) present in the wood charcoals. Moreover, the rise of old and young fallow species would indicate that the total land area under cultivation increased at this time, with more field area devoted to agriculture, some left uncultivated as long fallow, allowing for increased regeneration (Detarium microcarpa, Pterocarpus erinaceus, Terminalia etc.), and other areas with short fallow periods (Guiera senegalesis) (Huysecom et al. 2012, 2015).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR98\", \"CR99\", \"Fig9\", \"CR97\", \"CR49\", \"CR61\", \"CR97\", \"CR33\", \"CR76\", \"CR81\", \"CR82\", \"CR64\", \"Fig9\", \"CR99\", \"CR104\", \"CR99\"], \"section\": \"Sadia and the agricultural diversification of the Dogon Country\", \"text\": \"Not far from Sadia, on the Bandiagara plateau, the Ounjougou area provides a great number of Holocene archaeological sites excavated during the same project (Ozainne et al.\\u00a02014, 2017; Fig. 9). None of these sites were systematically sampled and analysed for macro-archaeobotanical purposes. Nevertheless, direct evidence of domesticated pearl millet cultivation dated to the 2nd millennium BC was found at the site of Varves\\u2013Ouest. Two pearl millet grains were directly dated to 1981\\u20131494 cal BC (95.4% calibration) or 1692\\u20131533 cal BC (Bayesian HPD) (Erl-9196; 3416\\u00b1109 BP), and to 1628\\u2013976 cal BC (95.4% calibration) or 1285\\u20131055 cal BC (Bayesian HPD) (Erl-9197; 3078\\u00b1131 BP) (Ozainne et al. 2009). Moreover, bilobate phytoliths (characteristic of pearl millet phytoliths) from the Yam\\u00e9 River deposit dated to 2600\\u20132300 BC may constitute indirect evidence of pearl millet cultivation, several hundred years before the appearance of its charred grains in the archaeological record (Garnier et al. 2013). Both direct and indirect indicators of pearl millet agriculture coincide with a radical change in ceramic tradition associated with the presence of numerous grinding stones and the appearance of settlement structures (Huysecom et al. 2004; Ozainne et al. 2009; Eichhorn and Neumann 2014). Such material culture and the presence of domesticated pearl millet coincide with contemporaneous similar finds at Wind\\u00e9 Koroji around 2200\\u20131900 cal BC (Macdonald et al. 2017). In the Ounjougou region, settlement mound sites occur from around 800 BC. From around 400 cal BC to 300 cal AD, a noticeable occupational decrease is noted on the Bandiagara plateau and the Seno Plain, the only evidence of human activities being recorded in a few caves of the Bandiagara cliff (Mayor et al. 2005, 2014; Huysecom et al. 2015; Fig. 9). The final Holocene (300\\u20131500 cal AD) is marked by the growth of mound sites into larger mounds on the Seno plain, such as Sadia and the appearance of iron production sites on the plateau (Ozainne et al. 2017; Robion-Brunner 2008). The apogee of Sadia, around 900\\u20131250 AD, \\u2018corresponds to the establishment of networks of villages, self-reliant in food resources and everyday consumption goods, alongside the first urban settlements along the Niger (e.g. Jenn\\u00e9-Jeno)\\u2019 (Ozainne et al. 2017: 223).\\n\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR81\", \"CR80\", \"CR13\", \"CR74\", \"CR83\", \"CR85\", \"CR92\", \"CR24\", \"CR47\"], \"section\": \"Sadia and the agricultural diversification of the Dogon Country\", \"text\": \"Secondly, the utilisation of new soil (more marginal soil) with the adoption of new cultivars (cowpea, fonio, Echinochloa) could also result from growing insecurity (ecological and social) that pushed people to explore agricultural risk buffering. Indeed, Sadia\\u2019s growth and abandonment coincide with a more general trend observed in Sahelian West Africa where most of the sites collapsed between 1200 and 1500 AD (Mayor et al. 2005;\\u00a0Mayor 2011). From abrupt climate changes, to war (i.e. the Songhay invasion) and diseases (i.e. the plague the ravaged Europe in the 1300s), several reasons are proposed for this sub-continental settlement rupture (Breunig 2013; Macdonald 2013; McIntosh 1993, 2000). However, no strong evidence can confirm or refute any of these potential causes. Nevertheless, all these factors could have contributed to insecurity/instability that pushed towards both more specialised and more diversified agriculture. Although there is a lack of high precision climatic proxies near Sadia, the wider western African region appears to have experienced wetter conditions during the Medieval Climatic Anomaly (900\\u20131250 AD) after which aridity set in (Nash et al. 2016). This could plausibly contribute to Sadia\\u2019s decline. In contrast, whereas a role for the plague pandemic in west Africa for abandonments of many sites from ca. 1350\\u20131500 AD has been recently explored (Chouin 2018; Gallagher and Dueppen 2018), this is later than the abandonment of Sadia (ca. 1300 AD). Nevertheless, some inherent declines in soil fertility as result of agricultural expansion and accumulating soil exhaustion over the course of phases 2 and 3 could also have contributed to the site\\u2019s decline.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR32\", \"Fig6\", \"Tab1\"], \"section\": \"Sadia in its regional context\", \"text\": \"The absence of agricultural diversification at Tongo Maar\\u00e9 Diabal is also visible through isotope analyses. Recent carbon and nitrogen isotope analysis on human teeth from people buried in the Bandiagara escarpment indicate diets with a major input of C4 foods (e.g. pearl millet) in the region of Tongo Maar\\u00e9 Diabal, and more diversified diets mixing C3 (e.g. rice and cowpea) and C4 foods in the southern part of the Bandiagara escarpment, close to Sadia. There, a temporal shift from a C4 food system to a mix system (C3 and C4) is also observed (Dlamini et al. 2019) and is convergent with our archaeobotanical results (Fig. 6, Table 1).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR107\"], \"section\": \"Sadia in its regional context\", \"text\": \"In the same way, nitrogen isotopes on pearl millet grains from Tongo Maar\\u00e9 Diabal show a lack of increase in pearl millet \\u03b415N values during the 650-year occupation of the site, implying that manure/household waste was added only sporadically to pearl millet (Styring et al. 2019). In other words, manuring was not a normal practice for continued annual processing, suggesting instead fallowing cycles. This continuity in soil maintenance practice reflects the general lack of change in the botanical and charcoal assemblages.\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR60\"], \"section\": \"Sadia in its regional context\", \"text\": \"Thirdly, to the east, in the Oursi region in Burkina Faso, archaeobotanical evidence is available since the Late Stone Age. Late Stone Age (Neolithic) Oursi inhabitants were probably seasonal agro-pastoralists, who were coming into this zone during the dry season and hunting wild animals and cultivating pearl millet. From 400 cal BC, permanent settlement mounds started to appear and grow. Economic subsistence was based on pearl millet cultivation. Cowpea was added to this agricultural system around 1 cal BC/AD, and sorghum around 500 cal AD, but in very low quantity. Throughout the sequence, fruit and wood charcoal remains of Adansonia digitata, Sclerocarya birrea, Balanites aegyptiaca and Vitellaria paradoxa suggest the presence of an agro-parkland system. Charred wood results illustrate that cultivation and settlement activities intensified considerably during the course of the occupation. For example, the system of fields and fallows, typical for this cultural landscape, replaced the natural Acacia savannah on the dune, as evidenced by the decrease of Acacia and the increase of Combretaceae charcoals (H\\u00f6hn and Neumann 2012).\"}, {\"pmc\": \"PMC7937602\", \"pmid\": \"33758626\", \"reference_ids\": [\"CR47\", \"CR48\", \"CR48\"], \"section\": \"Sadia in its regional context\", \"text\": \"Finally, to the south, the site of Kirikongo also provides archaeobotanical data. This settlement site is composed of mounds representing household compounds. This site complex was founded around 100 cal AD and abandoned like many other sites around 1400\\u20131500 cal AD, possibly including the impact of plague (Gallagher and Dueppen 2018). Inhabitants were performing a large array of economic tasks, from herding cattle, producing and forging iron, potting and cultivating. The most ubiquitous and numerically frequent plant taxon at the site is Shea Butter nut (Vitellaria paradoxa). The large quantity of carbonised shea fragments is attributed to the use of parching ovens, one of the steps in Shea butter production (Gallagher et al. 2016). Agricultural activities were based on pearl millet and fonio cultivation. Inhabitants were also collecting wild grasses and arboricultural products (baobab and shea butter). Archaeobotanical samples from the different settlement mounds is not fully published but initial results attest to the presence of sorghum from phase red II (1100\\u20131260 cal AD) (Gallagher et al. 2016).\"}]"

Metadata

"{}"