Fiona Marshall & Elisabeth Hildebrand
Cattle Before Crops: The Beginnings of Food Production in Africa
Journal of World Prehistory, Vol. 16, No. 2, June 2002. pp. 99-143
Domestication of plants: why so late?
In most regions of the world, plants were domesticated before animals. In Africa, plants were domesticated long after herding was first established. Delays in domestication of African plants are due to a number of factors, many of which relate to scheduled consumption, predictability, and the mobile nature of early pastoral societies. We discuss general reasons for the late domestication of plants; reasons for domestication of particular plant species are beyond the scope of this paper.
The first domestic plants in Africa were southwest Asian crops c. 7000 BP, which were confined to the Nile Valley by their need for winter rain (Wetterstrom, 1993). Indigenous African crops can be grouped into three complexes (Table I): savanna, forest margin, and Ethiopian (Harlan 1982, 1992b). On the basis of the geographic distribution of wild progenitors of African crops, Harlan (1971) argues that in Africa, unlike many other regions of the world, domestication of plants was noncentric: crops did not spread from a single geographic point of origin (Fig. 3). Rather, domestication occurred under a variety of conditions in widely dispersed regions of Africa.
The first indigenous domestic grains appear after early movement of Saharan groups into the west African grasslands c. 4000–3500 BP (Smith 1980, 1984) (Fig. 4). Domestic pearl millet is found as impressions in sherds at Dhars Tichitt and Oualata in Mauritania from c. 3500 BP (Amblard, 1996; Amblard and Pernès, 1989). Charred grains of domestic pearl millet are directly dated to c. 3460 BP at Birimi in northern Ghana (D’Andrea et al., 2001), c. 2840 BP at Ti-n-Akof in northern Burkina Faso, and c. 2930 BP at Gajiganna and c. 2430 BP at Kursakata in the Chad basin of northeast Nigeria (Neumann et al., 1996). Recent studies have resolved the longstanding question of whether pearl millet was domesticated by hunter-gatherers, early mobile herders, or later, more sedentary herders (Clark, 1976; Shaw, 1977). It now appears that domestication took place among semisedentary herders in west African savannas, well after the first appearance of pastoralists (D’Andrea et al., 2001; Holl, 1985; Neumann, 1999). Domestic pearl millet appears abruptly 600 years after early pastoralists enter Burkina Faso and Nigeria (Neumann, 1999).
Fig. 3. Proposed areas of domestication of African plants (after Harlan, 1971). 1. Guinea millet; 2. Fiono and black fonio; 3. African rice; 4. Yam (Dioscorea cayenensis complex); 5. Enset; 6. Tef; 7. Groundnuts (Kerstingiella and Vooandzeia); 8. Sorghum; 9. Bullrush/pearl millet; 10. Finger millet.
Fig. 4. Distribution of sites with early cultivated or domesticated plants.
Sites in the eastern Sahara and Sudanese Nile reveal evidence for early intensive use of sorghum, but for late morphological change. Sorghum dating to c. 7950–8020 BP at Nabta has lipids that differ from the wild form, but it is morphologically wild (Wasylikowa and Dahlberg, 1999; Wasylikowa, 2001; Wendorf et al., 1998). Its abundance indicates intensive use (Close, 2001) or perhaps occasional cultivation of wild sorghum (Wasylikowa et al., 1997). Impressions of wild sorghum appear in sherds at Um Direiwa, Kadero 1, and Zakiab (Stemler, 1990). On the basis of these, and on the high frequency of grindstones (e.g., 30,000 at Um Direiwa), Abdel-Magid (1989) and Haaland (1981, 1992, 1999) argue that wild sorghum was cultivated by c. 5000 BP. Secure dates for domestic sorghum are late: c. 2060 BP at Jenné Jenno (McIntosh, 1995) and 20–250 AD at the historic sites of Meroë, Jebel Tomat, and Qasr Ibrim in the Sudan (Clark and Stemler, 1975; Rowley-Conwy et al., 1999; Stemler and Falk, 1981) (Fig. 4). Morphologically domestic sorghum (race bicolor) from Qasr Ibrim is genetically identical to both wild and domestic modern sorghum in the area of the genome studied, supporting the idea of recent domestication (Rowley-Conwy et al., 1999).
Despite the importance of the Horn of Africa as a center for agricultural origins, little is known about domestication processes there. Recent excavations in the Aksum area (Fig. 4) have found domestic tef in historic periods: c. 500 BC at sites D and K (Boardman, 1999) and during the fifth century AD at Bieta Giyorgis (Bard et al., 1997). Sorghum and noog oilseeds appear during the sixth century AD (Boardman, 1999). Ethnoarchaeological studies of crop processing stages for tef (D'Andrea et al., 1999) and of differences between wild and domestic enset (Hildebrand, 2001) will facilitate future archaeological investigation. The relative antiquity of indigenous versus exotic Near Eastern crops in Ethiopia has yet to be fully explored.
East Africa has long been regarded as the locus for domestication of finger millet (Harlan, 1992a). Little research has been conducted on Later Stone Age sites in Uganda, but domestic finger millet occurs by the early seventh centuryADat Aksum (Boardman, 1999), and c. 1185 BP at Deloraine Farm in Kenya (Ambrose, 1984c) (Fig. 4).
The forests and forest margins of central and western Africa have yielded little archaeological data on the domestication of plants. After domestic sheep and goat appeared c. 3500 BP at Kintampo, use of local legumes and oil palm may have increased (but see Maley, 2001), yet local wild resources such as Canarium and small wild animals remained important (Anquandah, 1993; Flight, 1976; Stahl, 1985, 1993). Domestic cowpeas and Bambara groundnuts are first found in Iron Age contexts in western Africa (Vogelsang et al., 1999).
No indigenous domesticates are known from the southern half of the continent. Rather, pearl and finger millet, sorghum, and domestic pulses appear in conjunction with Early Iron Age mixed farming in Zambia, Zimbabwe, and South Africa (Maggs, 1984). The earliest grain, domestic Pennisetum, dates to c. 270 AD at Silver Leaves (Klapwijk, 1974; Klapwijk and Huffman, 1996) (Fig. 4).
Why So Late?
Many African cereals are amenable to domestication because of their appealing taste, rapid growth, high yields, drought tolerance, good storage potential, and single-locus genetic inheritance of key traits (Diamond, 1997; Harlan, 1992a). The absence of evidence for early domestic plants has been attributed to several causes. Young and Thompson (1999) argue that poor preservation of seeds and cereals in tropical east Africa results from rapid carbon cycling, wetting and drying cycles, and a high level of soil microbial activity, and contributes to a lack of evidence for early crops. Insufficient archaeobotanical research is also an issue (Bar-Yosef, 1998; Wetterstrom, 1998). Poor preservation and lack of research do not completely account for the absence of morphologically domestic plant remains before 4000 BP, however. A growing body of paleoethnobotanical data from northern Africa attests to the use of wild grasses and other plants by hunter-gatherers and early herders (Abdel-Magid, 1989; Magid and Caneva, 1998; Stemler, 1990; Wasylikowa, 1993; Wasylikowa et al., 1993, 1997). Possible reasons for the late onset of morphological change among African crops include harvesting practices, plant biology (Abdel-Magid, 1989; Haaland, 1992), the unpredictable plant productivity of the Sahara and its margins, and the mobility of early herders.
Uprooting plants or cutting seed heads with sickles can select for indehiscence, synchronous ripening, and loss of dormancy. Not all harvesting methods result in domestication, however (Harlan, 1992a). Tuareg and Zaghawa pastoralists today collect Panicum, Cenchrus, and other wild African grains by hand-stripping, by using a swinging basket, or by beating (Harlan, 1989, 1992c; Nicolaisen, 1963). These methods select for brittle rachises, and hence will not result in domestication. Evidence for prehistoric selection processes is mixed. Blades or flakes with gloss, possible indicators of sickle harvesting, are absent at Zakiab, Kadero 1, and Um Direiwa (Haaland, 1992), but appear during the mid-Holocene at Dakhleh (McDonald, 1998a), Adrar Bous III (Roset, 1987), and Farafra (Barich, 1998), and at Laga Oda in Ethiopia during the last 3000 years (Clark and Prince, 1978). More detailed microwear studies are needed to identify the precise cause of the gloss (e.g., grain cutting or hide preparation); if, however, these pieces are indeed found to be sickles, they might be indicative of harvesting practices that could have led to domestication.
Even if harvesting methods favor tough rachises, grains must be replanted and genetically isolated for selection to result in morphological change (Harlan, 1989, 1992c; Hillman and Davies, 1990). Late domestication of sorghum and pearl millet has been attributed to their ability to outcross, which can impede genetic isolation. According to this view, domestication of sorghum became possible only when early Sudanese pastoralists took it outside its natural range (Abdel-Magid, 1989; Haaland, 1992, 1995). Outcrossing rates vary among African cereals, however (National Research Council, 1996): finger millet is almost entirely self-pollinating, sorghum has highly variable rates of outcrossing, and pearl millet outcrosses the most but appears to have been domesticated the earliest. Although genetic isolation is a well-known general requirement of domestication, the actual mechanisms for isolation during prehistoric times are far from clear.
Climate and ecology in the southern Sahara may have fostered interactions between people and plants that would not have led to domestication. As already noted, low and variable rainfall affects productivity of plants more than that of animals, because the timing of rainfall relative to plant growth phases is crucial, and plants have fixed positions, whereas animals can move or disperse. All of these factors make cultivation of grains in marginal environments more risky than plant collecting. Even today, much of the northern Sahel can support pastoralism but not farming. Areas with less than 558 mm p. a. average one crop failure in 10 years, and farming is rarely attempted with less than 348 mm p. a. (Mortimore, 1998, p. 77).
If hunter-gatherers or pastoralists with marginal subsistence due to low and variable rainfall sought more predictable access to food, then emphasizing livestock would have been more logical than undertaking cultivation. Planting would have entailed gambling on when and where rain would fall and which cereals the rainfall distribution would suit. Different grains have different growth cycles: pearl millet matures quickly, whereas sorghum grows slowly but can use residual moisture (Mortimore, 1998, p. 89). By focusing on stock, herders could move animals to pasture and continue exploiting wild grass stands wherever they occurred in any given year. From initial stages of cattle domestication until the fourth millenium BP, pastoral strategies provided more predictable access to food than did intensification of plants. Mobile pastoral strategies precluded steady selection on populations of useful plants.When pastoralists moved into wetter grasslands and became more sedentary, then selection pressures on plants became sufficiently constant to cause morphological change.
Domestication requires a constellation of cultural plant management practices, such as reaping with a sickle and replanting harvested grain that is either self-fertilizing or genetically isolated year after year. If not all of the requisite practices or conditions are in place, then selection is not maintained, and morphological change does not take place. In Africa, suitable sets of factors for domestication came together late, rarely, and in highly varied circumstances. This is largely because the unpredictable environments of the early middle Holocene Sahara, and the mobile pastoral lifestyles they fostered, together created circumstances in which humans would not have exercised continuous, directional selection on cereals. Continuing intensive use of wild plants indigenous to different parts of Africa led to the continent's noncentric pattern of domestication.