Archaeobotanical Evidence for Pearl Millet (Pennisefumglaucum) in Sub-Saharan West Africa

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by RERO DOC Digital Library

Archaeobotanical evidence for pearl millet (Pennisefumglaucum) in sub-Saharan West Africa

A.C. D'ANDREA,M. KLEE& J. CASEY*

The remains of pearl millet Pennisetum glaucum) dating to 34602200 and 2960k370 BP have been recovered at the archaeological site of Birimi, northern Ghana, associated with the Kintampo cultural complex. This finding represents the earliest known occurrence of pearl millet in sub-Saharan Africa. Results indicate that Kintampo peoples developed effective subsistence adaptations to savannas as well as tropical forest habitats.

Key-words: West Africa, Kintampo, archaeobotany, pearl millet The origin of agriculture in sub-Saharan West of domestic ovicaprids (cf. Capra) and hunt- Africa has been associated with the Kintampo, ing of species attracted to clearings or settle- a ceramic Late Stone Age (LSA) cultural com- ment areas, such as large rodents (Thryonomys plex occupying the region of present-day Ghana swinderinnus). Plant remains include tropical during the 4th millennium BP. Archaeological forest margin species such as oil palm (Elaeis evidence indicates that Kintampo peoples pos- guineensis),incense tree (Canarium schwein- sessed a material culture similar to that of other furthii),hackberry (Celtis)and probable legumes early farmers, including chipped and ground (cf. Fabaceae). Related palynological studies projectile points, ground stone axes, grinding indicate that oil palm was cultivated or man- stones and ceramics. Sedentary village occu- aged in the region by at least 3500 BP (Sowunmi pation is indicated by remains of daub archi- 1981: 136; Talbot et al. 1984: 185).New macro- tecture, occasionally with stone foundations; botanical data bearing on this question have however rock-shelters also were inhabited been recovered during recent excavations at the (Davies 1962; Flight 1968; Dombrowski 1980; Birimi site in northern Ghana (Casey et al. 1997). Stahl 1985; Casey 1993). Although often de- Analysis of Birimi archaeobotanical samples scribed as the earliest settled agriculturalists reported in this paper have confirmed the as- of West Africa, the nature of Kintampo subsist- sociation of domesticated pearl millet ence has been the object of speculation for sev- (Pennisetum glaucum) with the Kintampo com- eral decades (Davies 1960; Flight 1976; plex. Posnansky 1984; Stahll993;Anquandah 1993). The Birimi site is situated in an area of dry Archaeologists have proposed that Kintampo woodland savanna atop the Gambaga Escarp- peoples cultivated various cereals and yams ment in northern Ghana (FIGURE1). A rich as- (Dioscorea rotundata, D. cayenensis),as well semblage of Kintampo artefacts has been as crop mixtures (Davies 1962: 291; Flight 1976: recovered on the surface and throughout the 218-19; Anquandah 1993: 260); however, di- complex series of pits and lenses of daub ar- rect evidence relating to subsistence has been chitectural remains preserved below. The site sorely lacking (Andah 1993: 250-53). Most extends for approximately one kilometre on both available data are derived from one rock-shel- sides of a gully, which was formed by an an- ter site, K6, located in central Ghana near the cient stream. Eroding sections of the gully edges tropical forest and forest-savanna mosaic bound- have exposed a portion of a Kintampo village ary (FIGURE1) (Stahl 1985: 138-42). This lo- including the remains of at least 34 daub ar- cality has produced evidence for the exploitation chitectural units. Excavated sections include

* D'Andrea, Department of Archaeology, Simon Fraser University, Burnaby, British Columbia, Canada, V5A 1S6. Klee, Universitat Basel, Botanisches Institut, Schonbeinstrasse fi, CH-4056 Basel, Switzerland. Casey, Department of Anthropology, IJniversity of South Carolina, Columbia SC 29208, USA.* Received 10 October 2000, accepted 14 February 2001, revised 13 March 2001. ANTIQTJITY75 (2001): 341-8

Downloaded from https:/www.cambridge.org/core. University of Basel Library, on 30 May 2017 at 18:15:30, subject to the Cambridge Core terms of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/S0003598X00060993 342 A.C. D'ANDREA, M. KLEE & J, CASEY

FIGURE1. Map showing location of sites mentioned in text. Floristic regions after White (1983).

Z cm

X 1- O - 10

- 20

- 30

-40 - 50 - 60 - 70 -80

0 charcoal 0 daub 0 sandstone I lateritr Q Quai M Mudrtone p Pottery

FIGUKE2. Gully profile, pit feature, Birimi.

a profile of the gully, which revealed a well- based on wood charcoal are summarized in defined pit feature (FIGUREa), and area exca- TABLE1 (Casey et al. 1997). vations farther upslope. Radiocarbon Birimi was sampled extensively for charred determinations for the gully profile pit feature macrobotanical remains. A total of 394 litres

depth lab material uncalibrated dendro- calibration curve TABLE1. AMS I4C (cm) no. radiocarbon calibrated intercept (cal BC) dates from gully year BP* age range (BC)* profile pit feature. 32-42 B-099308 charcoal 3310f110 1735-1453 1596-1528 42 B-099306 charcoal 3240+90 1618-1413 1512 45-50 TO-8173 grain 2960+370 1620-795 1130-1 2 50 50-60 TO-8172 grain 3460+200 1980-1520 1740 62 B-099307 charcoal 3550f40 1942-1777 1883 * 68% probability

of sediment taken from the area excavation and species count % total gully profile are included in this analysis, all count of which were processed using a simple sys- Pennisetum glaucum 1366 tem of manual flotation. Summary results of cf. Pennisetum glaucum 300 analyses are presented in TABLE2. Samples are domesticates 1666 61.3 dominated by grains of pearl millet (FIGURES3 Dactyloctenium aegyptium 1 & 4) and probable (cf.)pearl millet, with small Digitaria 2 quantities of wild grasses, other wild plants and Paniceae 72 indeterminate seeds. Two pearl millet grain Poaceae 71 samples from the gully profile pit were sub- wild grasses 146 5.4 mitted for radiocarbon dating by accelerator mass spectrometry (AMS) (TABLE1). The grains Fabaceae 9 produced low quantities of carbon, resulting Mimosaceae 1 in relatively large errors; however, both dates Solanaceae 225 confirm a Kintampo association. These grain other wild plants 235 8.6 ages are consistent with charcoal dates obtained unknown 3 from the same pit (TABLE1) and with an AMS unidentifiable 669 determination of 3490k50 BP (1878-1744 cal indeterminate 672 24.7 BC) obtained from charcoal associated with millet site total 271 9 100.0 grains in the area excavation (Casey et al. 1997). I Identifications of the Birimi pearl millet T~~~~ 2. ~i~i~imacrobotanical remains, specimens (I? glaucum) are based primarily on caryopsis (grain) shape. This feature has been suggested as a more effective criterion than gross grain size for the identification of archaeological pearl millet (Neumann eta]. 1996: 443). In addition, caryopsis shape was found to be the most consistent feature in distinguishing variation in a systematic study of floral characters in over 200 accessions of modern pearl millet landraces (Brunken 1977). Although grains of the domesticate are highly variable in size, they are generally obovate (club-shaped), often broadly so, and terete (circularin transverse section).Grains of the wild progenitor (I? violaceuml) are ellip-

1 l? glnucum is part of a complex of three forms which hybridize freely: the domesticate, wild progenitor, and intermediate forms or shihras. The nomenclature used fol- lows Clayton & Renvoize (1982: 672-3).

FIGURE3 (right).Pennisetum glaucum caryopsis, nrea excavafion, 53S, 114E, 130-140 cm, Birimi. Dorsal view. x60.

FIGURE4. Pennisetum glaucum caryopses, dorsal, lateral, and cross-section views: a area excavation 53s 114E, 130-140 cm; b gully profile, 25- 30 cm; c gullyprofile, 50- 1 mm 60 cm.

tic to lanceolate, and dorsally compressed. In- modern reference specimens of related genera termediate forms known as shibras (P. (including domesticated and wild species), such sieberanum) produce grains that range from as Brachiaria, Digitaria, Echinochloa, Panicum, obovate to elliptic, and are terete or moderately Paspalum, Setaria and Sorghum as well as compressed; in effect, they are intermediate Cenchrus were examined for comparison, with between domesticated and wild types (Brunken special emphasis on species used in ancient 1977: 165, 169, 170). Shibras are spontaneous and recent times by peoples of the Sahara and mimetic weeds which exist as stable genotypes, adjacent regions (Harlan 1989; Wasylikowa 1997; the result of introgression between domesticated Barakat & Fahmy 1999). However, none dis- and wild forms. They developed as a result of played the characteristic club-shape evident in weeding practices, and were probably present both Birimi specimens and modern landraces at the early stages of pearl millet domestica- of domesticated pearl millet. tion (Brunken et al. 1977: 167-8). Consequently, The outstanding feature of the Birimi pearl the occurrence of shibra or shibra-like grains millet is the small caryopsis size compared to in archaeological deposits implies the existence modern populations, a trait also noted at the of domesticated forms. All Birimi specimens site of Kursakata, Nigeria (FIGURE1) (Neumann identified as €? glaucum, regardless of actual eta].1996: 443). To illustrate this unusual char- size, conform to the obovate and terete domesti- acteristic, metrical data for 3 76 undistorted cated shape which is clear in both dorsal and specimens identified to €? glaucum were taken lateral views (FIGURE4). In addition to Pennisetum, for grain thickness vs breadth (TB). It was not

2.6 2.4 2.2 2 1.8 1.6 1.4 1.2 1 .8 .6 .4 I 1 .5 1 1.5 2 2.5 3 3.5 Breadth (mm)

FIGURE5. Thickness vs. breadth dimensions for measurable charred P. glaucum grains from Birimi. Publislied ranges for uncharred modern pearl millet, shibras, and wild forms are superimposed @runken 1977: 165,169, 17Uj.

possible to measure length in many specimens, is consistent with the domesticated form. The because proximal ends (embryo)were missing. occurrence of several large grain fragments (not FIGURE5 and TABLE3 illustrate Birimi grain measurable) indicates that the percentage of total TB dimensions in relation to published data grains falling into the shibra-domesticate range for modern pearl millet, the wild progenitor, is somewhat higher. Even so, the use of metri- and shibras, leaving aside the effects of char- cal criteria alone is clearly problematic in the ring in archaeological grains.2 As evident in identification of early domesticated forms of FIGURE5, the majority of measured Birimi do- pearl millet in the archaeological record. Sev- mesticated grains, 60% (226 specimens), have eral factors mitigate against the use of gross grain no apparent modern analogue (breadth less than size in this manner: the astonishing increase 1.0 mm), 23% (87) fall into the shibra-domes- in pearl millet inflorescence size as the result ticate range (thickness21.0 mm), and 17% (63) of evolution under domestication (Brunken et actually fall into the size range of modern wild al. 1977:167); the extreme variability in grain grains (thickness 51.0 mm), although their shape size documented in modern land races (Brunken et al. 1977: 168);and the somewhat unpredict- 2 The effects of charring on cereal grain dimensions are not well understood, but experiments indicate that TBratios able effects of charring (Nesbitt 1997:186). tend to remain unchanged because both thickness and The Birimi grain samples fall into the same breadth increase slightly (Nesbitt 1997: 186). The only time range as the earliest known dates on pearl available experimental data on charring pearl millet is found millet at the site of Dhar Tichitt, Mauritania in Abdel-Magid (1989: 303) who suggests that grain thick- 1).At this locality, organic remains in ness and breadth increase by 7-8%. Although some pa- (FIGURE ranieters of this experiment are unclear, the results are pottery containing impressions of pearl millet considered acceptable (Nesbitt 1997: 183). grains and spikelets have produced a determi-

thickness (mm) breadth (mm) grain source max. min. mean max. min. mean Birimi archaeological 1.7 0.6 0.8 1.9 0.6 0.9 TABLE 3. Birimi and nlodern pearl millet* 2.5 1.2 NA 3.2 1.6 NA published grain modern shibras* 2.0 1.0 NA 2.2 1.0 NA dimensions for pearl modern wild" 1.0 0.6 NA 1.5 1.0 NA millet, wild pearl miIlef and skbrus. * metrical data published in Brunken (1977: 165, 169, 170)

nation of 3500f100 BP (1936-1683 cal BC). inhabitants were cultivating pearl millet un- Ancient pastoralists of Saharan West Africa, such der highly variable environmental conditions. as those occupying Dhar Tichitt, were cultivating It is difficult to be more specific about the ef- pearl millet that was genetically isolated from fects of these climatic changes on the inhabit- wild forms by at least 3600 BP (Amblard 1996: ants of Birimi, given the lack of palaeoecological 423). The Birimi samples have provided the data from northern Ghana. first indication that this crop was present in There is general agreement that pearl millet sub-Saharan regions in association with was domesticated somewhere along the south- Kintampo peoples living about 1000 km fur- western fringes of the Sahara (Harlan 1971: 470- ther south.’ Evidently, pearl millet was utilized 71; Brunken et al. 1977: 173; Tostain & Marchais over a relatively widespread region of West 1993: 41-7). A Saharan origin for this domes- Africa during this time, implying an earlier do- ticate may indicate that it was introduced to mestication event. The pearl millet preserved Kintampo peoples from the north. This theme at Birimi and Kursakata4 (Neumann et al. of northern origins raises once again the ques- 1996:433, figure 2; Klee and Zach 1999:86-871, tion of sources of influence on the Kintampo may represent one or more small-grained vari- complex. Earlier workers believed that the eties in existence at the early stages of domes- Kintampo developed as the result of popula- tication. If so, it can be postulated that changes tion movements out of the Sahara in response in caryopsis shape preceded grain size increases to increasing aridity after 4500 BP, based on in the domestication of this species. Although similarities in material remains to cultures oc- thickness measurements of some Birimi grains cupying the Tilemsi region (Davies 1962: 301; fall into established ranges for modern domes- Flight 1976: 216). More recent thinking envi- ticated and shibra forms, a significant propor- sions the Kintampo as a fusion of northern tion have breadth dimensions smaller than these imported and local elements, based on conti- extant representatives (FIGURE5). A possible nuity in cultural development evident in the explanation for this anomaly may lie in the archaeological record (Stahll985: 145-7). Pearl remarkable variation observed in grain char- millet can now be added to the list of northern acters in modern landraces, caused by elements that were incorporated into what was introgression with regional populations of the primarily an indigenous Kintampo cultural wild subspecies, as well as geographic isola- milieu. The cultivation of this highly drought- tion and human selection over millennia resistant crop would have been an effective (Brunken et al. 1977:173).This picture of vari- adaptation to the wet/dry seasonal climatic ation is even further complicated if environ- regime of the West African savannas (Casey 1998: mental alteration through time is taken into 53-7; D’Andrea & Casey in preparation). account. Palaeoecological evidence based on The Birimi pearl millet finds have expanded water level fluctuations in Lake Bosumtwi, the database on Kintampo subsistence which southern Ghana (FIGUREI), indicates that a pe- until now, has been dominated by tropical for- riod of extreme aridity culminating at 3750 BP est margin species, such as oil palm, incense is immediately followed by conditions signifi- tree and hackberry, recovered from sites in cen- cantly more humid than today which peak at tral and southern Ghana. These species also ap- approximately 2420 BP (Talbot eta].1984: 186, pear in later West and Central Africa at ceramic 190). Occupation of Birimi spans the end of LSA sites dating to the 1st millennium BC: (Stahl this arid phase as well as the ensuing rapid 1993: 272-3; Mbida et al. 2000: 155-8) indicat- increase in precipitation levels, indicating that ing a widespread tropical West African LSA adaptation to forest-margin environments, 3 Davies (1980: 220) reports the possible occurrence of emphasizing the exploitation of oleaginous tree pearl millet ‘cob’ impressions in pottery at Ntesero. How- fruits. These species, however are notable by ever, he does not advance this as evidence for cereal agri- their absence in Birimi deposits which are domi- culture in the absence of a definitive identification, which nated by pearl millet. Macrobotanical assem- has not taken place. There also is some uncertainty sur- rounding thc context of the potsherds. blages including pearl millet and wild grasses 4 It is not possible to provide coinparative figures be- are more typical of sites in the Sahara and Sahel cause grain sizc data are not available. regions farther north (Amblard & Pernks 1989;

Klee & Zach 1999; Neumann 1999). This ap- of West Africa. This underscores the impres- parent dichotomy can be explained to some sive range of ecological knowledge Kintampo degree by an examination of the present-day peoples must have had about survival in the ecology of the region. Birimi exists today in varied landscapes at their disposal. dry savanna grassland, more than 500 km north of the extent of oil palm cultivation. Shea but- Acknowledgements. This paper is dedicated to the memory ter tree (Vitellaria paradoxa), not oil palm, is of the late Kivilwura Yahaya Seidu of Accra, Ghana. the significant oleaginous species exploited in The Gambaga Archaeological Project was funded by a northern Ghana, the remains of which have not Social Sciences & Humanities Research Council of Canada Research Grant to JC in 1995. We would like to thank the yet been recovered from Birimi deposits. Also, Ghana Museums and Monuments Board and the Ghana large-scale pearl millet cultivation is limited Forestry Service for permission to carry out the excava- by precipitation levels to northern Ghana be- tions. Birimi macrobotanical remains were identified us- cause this crop requires a pronounced dry sea- ing botanical reference collections housed at the Seminar 1985: 25). fir Vor- und Fruhgeschichte, Archaologie und Archaobotanik son (Appa Rao It appears that regional Afrikas at the University of Frankfurt and personal collec- resource specialization was in existence dur- tions of ACD. In addition, Pennisetum specimens held at ing the Kintampo as it is today, with southern the Botanischer Garten und Botanisches Museum Berlin populations practising a forest-margin adapta- Dahlem were consulted. We are grateful to: Katharina tion, while further north in wooded savanna Neumann (University of Frankfurt) for analytical advice; Patrick Hogue (University of British Columbia) and Nicole grasslands, inhabitants were integrating pearl Jackman (Simon Fraser University) for laboratory assist- millet into their subsistence regimes. ance; Ann Butler (University College London) who com- Results presented herein offer support to pleted the SEM work; Shannon Wood, Cheryl Takahashi earlier speculations that domesticated cereals and Greg Ehlers (Simon Fraser University) for assistance formed a component of Kintampo subsistence. with illustrations; Dorothy Godfrey-Smith (Dalhousie Uni- versity) for assistance with charcoal dates; and Larry Pavlish A picture is now emerging of Kintampo peo- (University of Toronto), Diane Lyons (Simon Fraser Uni- ples partaking in a broad range of subsistence versity) and two anonymous reviewers for providing valu- activities that cross-cut the complex ecology able comments on various drafts of this manuscript.

References ABDEL-MAGID,A. 1989. Plant domestication in the Mddle Nile ment. Ph.D thesis, Department of Anthropology, Univer- Basin: an archaeobotanicol case study. Oxford: British sity of Toronto. Archaeological Reports. International series S523iCam- 1998. The ecology of food production in West Africa, in G.

bridge Monograph in African Archaeology 35. ~ Connah (ed.), Transformations in Africa: Essays on Afri- AMBLAKD,S. 1996. Agricultural evidence and its interpreta- ca's laterpast: 46-70. London: Leicester University Press. tion on the Dhars Tichitt and Oualata, southeastern Mali, CASEY,J., R. SAWATSKY,D. GODFREY-SMITH.N. QIJICKERT, A.C. in Pwiti & Soper Led.): 421-8. D'ANDREA,M. WOLLSTONECROF~'& A. HAWKINS.1997. AMMLARD,S. & J. PERNES.1989. The identification of cultivated Report of investigations at the Birimi Site in northern pearl millet (Pennisetum) amongst plant impressions on Ghana, NJiarne Akuma 48: 32-8. pottery from Oued Chebbi (Dhar Oualata, Mauritania], CLAYTON,W.D. & S.A. RENVOIZE.1982. Flora oJ Tropical East African Archaeological Review 7: 117-26. Africa, Gramineae [Part 3): 672-3. Rotterdam: A.A. ANDAH,B.W. 1993. Identifying early farming traditions of West Balkema. Africa, in Shaw et al. (ed.): 240-54. D'ANDREA,A.C. & J. CASEY.In preparation. Pearl millet and ANQUANDAH,J. 1993. The Kintampo Complex: A case study of Kintampo subsistence. early sedentism and food production in sub-Sahelian West DAVIES,0. 1960. The Neolithic revolution in tropical Africa, Africa, in Shaw eta]. (ed): 255-60. Transactions of the Historical Societyof Ghana 4(2): 14-, AWA RAO, s.,M.H. MENCESHA& D. SHAKMA. 1985. Collection 20. and evaluation of pearl millet (Pennisetum americanum) 1962. Neolithic cultures of Ghana, in G. Mortelmans & J. germplasm from Ghana, Economic Botany 39(1): 25-38. Nenquin (ed.), Actes du IVe Congres Panafricoin de BAKAKAT,H. & A.G. FAHMY.1999. Wild grasses as 'Neolithic' Prkhistoire et de ]'Etude du Quaternoire, Section Ill. Prk- food resources in the Eastern Sahara: A review of the et protohistoire: 291-302. Tervuren: Musee Royal de evidence from Egypt, in M. van der Veen (ed.), The ex- 1'Afrique Central. ploitation of plant resources in ancient Africa: 33-46. 1980. The Ntesero Culture in Ghana, in B.K. Swartx & R.A. New York (NY): Kluwer Academic. Dumett (ed.),WestAfrican Culture Dynamics: 205-25. The BRUNKEN,J.N. 1977. A systematic study of Pennisetum Sect. Hague: Mouton. Prnnisetum (Gramineae), American Journal of Botany DOMBROWSKI,J.C. 1980. Earliest settlements in Ghana: the 64(2): 161-76. ' Kintampo industry, in R.E. Leakcy & B.A. Ogot (ed.),Pro- BRUNK~~N,J.N., J.M.J. DEWET& J.R. HARLAN.1977. The mor- ceedings of the VlIIth Panafrican Congress of Prehistory phology and domestication of pearl millet, Economic and Quaternary Studies: 261-2. Nairobi: International Botony 31: 163-74. Louis Leakey Memorial Institute for African Prehistory. CASEY, J. 1993. The Kintampo Complex in northern Ghana: FLIGHT,C. 1968. Kintampo 1967, West African Archaeological Late Holocene human ecology on the Gambaga Escarp- Newsletter. 8: 25-20.

1976. The Kintampo Culture and its place in the economic PWITI,G. & R. SOPER (ed.). 1996. Aspects of african archaeol- prehistory of West Africa, in J.R. Harlan, J.M.J.deWet & ~gy.Harare: University of Zimbabwe Publications. A.B.L. Stemler (ed.),Origins of African plant domestica- SHAW,T., P. SINCLAIR,B. ANDAH& A. OKPOKO(ed.). 1993. The tion: 211-21. The Hague: Mouton. archaeology of Africa: food, metals and towns. London: HARLAN,J.R. 1971. Agricultural origins: centres and non-cen- Routledge. tres, Science 174: 468-474. SOWUNMI,M.A. 1981. Late Quaternary environmental changes 1989.Wild grass harvesting in the Sahara and Sub-Sahara, in Nigeria, Pollen et Spores 13(1]: 125-48. in D.R. Harris & G.C. Hillman [ed.), Foraging and Farm- STAHL, A.B. 1985. Reinvestigation of Kintampo 6 rock shelter, ing: 79-98. London: Unwin Hyrnan. Ghana: implications for the nature of culture change, KLEE,M. & B. ZACH.1999. The exploitation of wild and do- African Archaeological Review 3: 117-50. mesticated food plants at settlement mounds in north- 1986. Early food production in West Africa: rethinking the east Nigeria, in van der Veen (ed.):81-8. role of the Kintampo Culture, Currenf Anthropology27(5): MBIDA,C.M., W. VAN NEER, H. DOUTRELEPONT& L. VRYDAGHS. 532-6. 2000. Evidence for banana cultivation and animal hus- 1993. Intensification in the West African Late Stone Age, in bandry during the first millennium BC in the forest of Shaw et al. (ed.): 261-73. southern Cameroon, Journal of .-lrchaeological Sciencp TALBOT,M., D.A. LIVINGSTONE,P.G. PARKER,J. MALEY,J.M. 27(2): 151-62. MELACK,G. DELIBRIAS& S. GIJLLIKSEN.1984. Preliminary NESBITT,R.M.A. 1997. Archaeobotanical identification of Near results from sediment cores from Lake Bosumtwi, Ghana, Eastern grass caryopses. Ph.D thesis, University College Palaeoecology of Africa 16: 173-92. London. TOSTAIN.S. & L. MARCHAIS.1993. Evaluation de la diversite NEUMANN,K. 1999. Early plant food production in the West ghnetique des mils (Pennisetum glaucum (-.) R. BR) au African Sahel: new evidence, in van der Veen (ed.1: 73- moyen de marqueurs enzymatiques et relations entre 81. formes sauvages et cultivees, in S. Hamon [ed.), Le Mil NEUMANN,K., A. BALLOUCHE& M. KLEI!. 1996. The emergence en Afrique: 33-56. Paris: Editions de I’ORSTOM. of plant food production in the West African Sahel: new VAN DER VEEN,M. (ed.). 1999. The exploitation ofplant resources evidence from northeast Nigeria and northern Burkina in ancient Africa. New York [NY):Kluwer Academic. Faso, in Pwiti & Soper [ed.): 441-8. WASYLIKOWA,K. 1997. Flora of the 8000 years old archaeo- POSNANSKY,M. 1984. Early agricultural societies in Ghana, in logical site E-75-6 at Nabta Playa, Western Desert, south- J.D. Clark & S.A. Brandt (ed.), From hunters to farmers: ern Egypt, Act0 Paloeobotanica 37(2): 99-205. 147-51. Berkeley (CAI: University of California Press. WHITE,F. 1983. The vegetation of Africa. Paris: UNESCO.