First and second millennium a.d. agriculture in Rwanda
First and second millennium A.D. agriculture in Rwanda: archaeobotanical finds andradiocarbon dates from seven sites
Author(s): John D. Giblin and Dorian Q. Fuller
Source: Vegetation History and Archaeobotany, Vol. 20, No. 4 (July, 2011), pp. 253-265
Published by: Springer
Stable URL: https://www.jstor.org/stable/23885179
Accessed: 16-04-2019 07:08 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a widerange of content in a trusted digital archive. We use information technology and tools to increase productivity andfacilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available athttps://about.jstor.org/terms
Springer is collaborating with JSTOR to digitize, preserve and extend access to Vegetation
History and Archaeobotany
This content downloaded from 146.230.187.153 on Tue, 16 Apr 2019 07:08:03 UTC
All use subject to https://about.jstor.org/terms
Veget Hist Archaeobot (2011) 20:253-265 DOI 10.1007/s00334-011-0288-0 ORIGINAL ARTICLE First and second millennium a.d. agriculture in Rwanda: archaeobotanical finds and radiocarbon dates from seven sites John D. Giblin • Dorian Q. Fuller Received: 8 September 2010/Accepted: 12 March 2011/Published online: 25 March 2011 © Springer-Verlag 2011 Abstract This article presents the results from a pro gramme of bulk soil sampling and flotation of first and second millennium a.d. early farming, 'Iron Age', archaeological sites in Rwanda conducted in 2006-2007 alongside a new set of associated radiocarbon dates, which contribute toward the development of a chronology of plant use for the region. This research has identified the earliest examples of pearl millet (Pennisetum glaucum), finger millet (Eleusine coracana) and sorghum (Sorghum bicolor) in Great Lakes Africa and thus this article also discusses the significance of these finds within the later archaeology of the region and presents a brief synthesis of the direct archaeological evidence for finger millet in sub-Saharan Africa. Keywords Palaeoethnobotany ■ Eleusine cornearí ■ Sorghum bicolor ■ Pennisetum glaucum ■ Bantu •Urewe Communicated by M. van der Veen. Electronic supplementary material The online version of this article (doi:10.1007/s00334-011-0288-0) contains supplementary material, which is available to authorized users. J. D. Giblin School of Global Studies, University of Gothenburg, Gothenburg, Sweden e-mail: john.giblin@globalstudies.gu.se D. Q. Fuller (El) Institute of Archaeology, University College London, 31-34 Gordon Square, London WC1H OPY, UK e-mail: d.fuller@ucl.ac.uk Introduction For many regions of Africa archaeobotanical research is minimal, and this is true of Rwanda. The prehistory of agriculture in such regions is therefore unknown, or has been inferred from other sources such as historical lin guistics. For example, in much of eastern Africa the appearance of early farming has been suggested to be associated with the expansion by migration of Bantu speaking agriculturalists, with sorghum (Sorghum bicolor) and pearl millet (Pennisetum glaucum) obtained from Central Sudanian (Nilo-Saharan) speaking groups some where around the northern area of Great Lakes Africa (southern Sudan, western Kenya, Uganda) (e.g. Phillipson 1993; Schoenbrun 1993; Ehret 1998). Many traditional agriculturalists also cultivate finger millet (Eleusine com earía), which is often attributed to an origin in the Ethio pian highlands and dispersal southwards initially with Cushitic speaking groups (Ehret 1998). However, increas ing archaeological research has called into question some of the hypothesized straightforward correlations between artefact styles, the presence of agriculture and thus the Bantu-farming expansion hypothesis (e.g. Robertshaw and Collet 1983; Lane et al. 2007). One line of evidence which is needed to better under stand the prehistory of African subsistence and its rela tionship to various material culture traditions is systematic archaeobotanical research from which crops can be directly associated with particular cultures and tied to a radiocarbon chronology. Thus, the present paper reports a small but significant archaeobotanical dataset obtained by flotation of samples from seven sites in Rwanda dating to between the early first millennium a.d. through to the late second mil lennium a.D., together with associated radiocarbon dates. These provide the first direct archaeobotanical evidence Ö SpringerThis content downloaded from 146.230.187.153 on Tue, 16 Apr 2019 07:08:03 UTCAll use subject to https://about.jstor.org/terms
254 Veget Hist Archaeobot (2011) 20:253-265 from the region for sorghum, pearl millet, cowpea, together with several sites with finger paper also provides a discussion of the known or history of the discovered crops in eastern Africa, particular emphasis on finger millet (E. coracana). Prior archaeobotanical recovery Southern Rwanda around was an important pre-colonial the first and second millennia a.d. Central Rwanda, around Gitarama, a more arid region on Rwanda's central plateau that was to become the core political region of the Nyiginya Kingdom, "The Ante cedent to Modern Rwanda" (Vansina 2004), in the late second millennium a.d. Northern Rwanda, around Ruhengeri a lacustrine region situated at the foot of the Virunga volcanoes that was within the periphery of the central African forest until the late second millennium a.d. (Vansina 2004). Wetterstrom (1991) conducted one of the first systematic efforts to recover archaeobotanical remains from archaeo logical contexts in eastern Africa at Gogo Falls but did not identify any cereal or seed crops, only wild plant remains. Similarly, efforts by Young (1999; Young and Thompson 1999) to recover crop remains in Uganda at Münsa were unsuccessful, which led her to suggest that the particular crop-processing practices applied to finger millet The potential vegetation of southern and central Rwanda (E. coracana) might operate against archaeological pres- is considered to be evergreen bushland, with numerous ervation. Another example of this approach in eastern Acacia spp. (White 1983, p. 182), and associated short Africa is Walshaw's (2005, 2010) work at two Swahili sites grasses such as Themeda triandra, Loudetia spp. and on Pemba where she employed systematic bulk soil sam- Brachiaria spp. (Rattray 1960, p. 121), although today the pling and bucket flotation and identified both wild and area is heavily cultivated with cash-crops, beans and some domesticated plant remains, including sorghum, pearl millets alongside other subsistence crops such as sweet millet and finger millet. potatoes and cassava. In these areas, although there is some Such evidence, however, is very limited from this rainfall throughout the year, there are region. A single grain of finger millet was recovered from a seasons with peaks in precipitation flotation sample at the "Iron Age" site of Deloraine in November (FAO 2010). In northern Rwanda Kenya from cal. a.d. 800 (Ambrose et al. 1984), and a nga volcanoes, the potential woodland is single grain of wild finger millet (E. africana) was reported etation (White 1983, p. 169) which transitions from the lowest level at Gogo Falls (Lange 1991). While elevations into the evergreen bushland pollen data from the archaeological site of Kabuye in south, although today the region around southeast Rwanda have been referred to as "cf. Eleusine" volcanoes is heavily deforested. This northern and "cf. Sorghum" (Van Grunderbeek and Roche 2007), main rainy season with a peak in rainfall species level identifications of grass pollen must be taken precipitation in other months (FAO 2010). cautiously because there is no comparative pollen mor- In total, excavations at eight sites, employing phological (i.e. ultrastructural) study known to the authors units of 1 x 2 m2 to 2 x 3 m2, encountered that would allow the differentiation of these crops from the archaeological deposits, with sampling for majority of African grasses based on pollen morphology, carried out at all sites (Fig. 1). Most sites although for much more recent periods pollen and phyto- shallow stratigraphy so the total number liths are useful for tracking the spread of maize (Zea mays) relatively small with 32 samples taken (e.g. Lamb et al. 2003; Kiage and Liu 2009). (Table 1). The present paper reports the archaeobotanical Thus within this regional context, where evidence is finds and radiocarbon dates from these sites, scarce, the limited assemblage of palaeobotanical remains from Rwanda (presented below) makes a significant Ceramic associations empirical contribution to our understanding of subsistence in Great Lakes Africa in the first and second millennia a.d. All of the palaeobotanical samples presented here come from archaeological periods that are regarded as Iron Age Environmental and archaeological context in older chronological terminology. Herein we discuss them in terms of their associated ceramic phases. A program of archaeological survey and small-scale Classic Urewe users are represented at two sites, Kabusanze excavation was conducted in Rwanda, during 2006-2007, and Masangano. Three dates, all ca. cal. a.d. 400, came from with the aim of better documenting the later prehistory of three features (two pits, one burial) at Kabusanze, which the country and its relationship to adjacent parts of Africa were associated with Classic Urewe ceramics. A single date, (Giblin 2010). also cal. a.d. 400 came from an occupation deposit at <£) SpringerThis content downloaded from 146.230.187.153 on Tue, 16 Apr 2019 07:08:03 UTCAll use subject to https://about.jstor.org/terms
Veget Hist Archaeobot (2011) 20:253-265 255 DEMOCRATIC I UGANDA rT The aPPearance REPUBLIC OF ( / \ been widely associated with the appearance of Classic CONGO muuJi.^n/a j Urewe ceramics (Posnansky 1973). However, this sim plistic "one-to-one" association has been questioned by more recent work (e.g. Lane et al. 2007). Classic Urewe ceramics are generally of a high quality, with a broad vessel form range, limited fabric range and commonly display bevelled rims, dimple-bases and complex incised motifs (Van Grunderbeek 1988). A critical estimate for the existence of Classic Urewe is approximately 500 cal b.c. to cal. a.D. 800 in Great Lakes Africa (Clist 1987). Posnansky (1967, 1973; Posnansky et al. 2005) was the TANZANIA first to identify a transitional, or devolved, Urewe at Lolui Island, Uganda and since then it has also been identified at various other sites in Uganda (Ashley 2005, 2010). Although absolute dates are rare, the dates that do exist Fig. 1 Map showing excavated sites from Rwanda (site position place transitional Urewe in the terminal first millennium indicated by +) a.d. Transitional Urewe has less effort invested than Classic Urewe, including a significant Masangano with a mixed ceramic assemblage that included frequency Classic Urewe ceramics. plex incised decorations, and a narrower form range but Transitional Urewe users are represented by a single broader fabric range. This period in Great date, ca. cal. a.d. 700, from the site of Karama in central been associated with climatic insecurities Rwanda that came from a pit fill containing transitional patterns of resource exploitation Urewe ceramics. Taylor et al. 2000). Roulette-decorated users are represented by five dates, Roulette-ca. cal. a.d. 1100, from four cave sites in northern Rwanda that were generated from samples taken from the strati- although graphically earliest archaeological deposits in each cave. lematic, The ceramic assemblages from all the caves were entirely Grunderbeek made up of roulette-decorated ceramics. period coincided with growing subsistence complexity, DEMOCRATIC UCAMHA republic of uganda CONGO Mwe NguruO ■lusanie "tesanganoV^fe, KaramaT*KIGALI RjN A N qTa' \ Kabusanze» ^ m BURUNDI TANZANIA 1 Map showing excavated sites from Rwanda (site position indicated by +) Table 1 Excavated and sampled Rwandan sites Location Period Site type Palaeobotanical results Kabusanze Southern Rwanda 1°38'25"S, 29°52'44"E Mid-1st mill. a.d. Open hilltop Identifiable seed remains Masangano Northern Rwanda 1°44'28"S, 29°38'60"E Mid-lst mill. a.d. Open site at junction of two rivers Wood charcoal only Karama Central Rwanda 2°1'41"S, 29°55'30"E Late 1st mill. a.d. Open hilltop Identifiable seed remains Musanze II Northern Rwanda 1°35'38"S, 29°50'29"E 2nd mill. a.d. Cave at base of Virunga volcanoes Identifiable seed remains Musanze III Northern Rwanda 1°37'15"S, 29°50'51"E 2nd mill. a.d. Cave at base of Virunga volcanoes Identifiable seed remains Musanze IV Northern Rwanda 1°30'50"S, 29°36'91"E 2nd mill. a.d. Cave at base of Virunga volcanoes Non-identifiable seed remains Nguri Northern Rwanda 1°25'7"S, 29°53'52"E 2nd mill. a.d. Cave at middle elevation on Muhabura volcano Identifiable seed/fruit remains Mweru cave Northern Rwanda 1°28'3"S, 29°53'2"E 2nd mill. a.d. Cave on shores of Lake Bulera Wood charcoal only Site name Location Period Site type Palaeobotanical results D. D. mill. A.D. D. D. D. A.D. <£) SpringerThis content downloaded from 146.230.187.153 on Tue, 16 Apr 2019 07:08:03 UTCAll use subject to https://about.jstor.org/terms
256 Veget Hist Archaeobot (2011) 20:253-265 including the development of specialised pastoralism agriculture, increasing socio-political centralisation Ntusi, Reid 1996, and Münsa, Robertshaw 1997) eventually the establishment of kingdoms throughout region in the late second millennium a.d. (2003). Roulette-decorated ceramics are generally utilitarian than Classic Urewe, with a limited broad fabric range, simple rims and frequent oration. Overall there is less effort invested decorated ceramics than in Classic Urewe. However, examples of high quality roulette-decorated exist, which have been associated with the establishment elites and the development of specialised potters Kingdom Era (e.g. Desmedt 1991; Vansina The archaeobotanical results discussed in the section represent some of the first macrobotanical for plant use in the first millennium a.d. associated Classic Urewe and transitional Urewe ceramics Lakes Africa and make a limited contribution understanding of plant use by roulette-decorated using forest cave dwellers in the second millennium Results Materials and methods Radiocarbon chronology All recognized archaeological features and distinct AMS radiocarbon dates archaeological strata were sampled. Bulk soil samples of from sample contexts between 6 and 30 1 were collected from each context. A All the radiocarbon standard of 12 1 was set, based on transportation and pro- chronology based cessing logistical constraints in Rwanda, but where very assignment of these small contexts were encountered smaller samples were first and second millennia taken, and where the most 'significant' contexts were below by period and encountered, such as the Classic Urewe burial at Kabu- Table 2. Dates were sanze, the sample size was increased. It was not intended which they were that the results of the sample analysis would be statistically Urewe, Transitional analysed, instead sampling was undertaken as a 'presence/ mic phases (Fig. absence' investigation, to assess the potential for archaeo- with OxCal. 3.10 botanical research in this region, and to recover charcoal for IntCal 2009 curve radiocarbon dating. Flotation was carried out using bucket wash-over flotation (Helbaek 1969, p. 385; Pearsall 2000, Seed taxa identified p. 31). The light fraction samples were collected with a set of sedimentological sieves with mesh sizes of 2, 1 and The full range of 0.3 mm. The remains collected in the sieves were dried Table 3. Seed taxa that separately on muslin material indoors. All remains were level are described picked through with tweezers and all organic remains were taxonomic authorities separated, collected and stored in muslin wraps. A non- addition a number discretionary approach was employed at this stage and all or seed coats, which non-geological material was collected for more detailed widespread finds were analysis and identification in London. finds of pulse seeds, additional wild seeds, and at least one In London each dot fraction was sorted by the second tree fruit. Whole or nearly whole seeds, in particular those author under a binocular microscope at magnifications of likely cultivars, were measured on digital photomicro from x6 to x40. Larger wood charcoal pieces were taken graphs using a basic image capture program (Leica EZ), for radiocarbon dating from archaeological contexts that and these measurements are collected in ESM Table 1. <£) SpringerThis content downloaded from 146.230.187.153 on Tue, 16 Apr 2019 07:08:03 UTCAll use subject to https://about.jstor.org/terms
Veget Hist Archaeobot (2011) 20:253-265 257 Table 2 Radiocarbon dates from the excavated Rwandan sites Site (sample origin) Lab. no Associated ceramic 14C (years b.p.) Cal age (a.d., 2-<r range) Kabusanze (pit fill) OxA-19517 Classic Urewe 1610 ± 26 Kabusanze (burial fill) OxA-19518 Classic Urewe 1630 ± 26 417-554 Kabusanze (pit fill) OxA-19583 Classic Urewe 1694 ± 37 263-538 Masangano (occupation deposit) OxA-19520 Classic Urewe 1698 ± 27 Karama (pit fill) OxA-19519 'Transitional' Urewe 1291 ± 25 688-877 Musanze II (occupation deposit) Roulette-decorated 956 ± 26 Musanze III (occupation deposit) OxA-19522 Roulette-decorated 996 ± 25 1028-1152 Nguri cave (occupation deposit) OxA-19523 Roulette-decorated 956 ± 26 Mweru cave (burnt floor) OxA-19524 Roulette-decorated 951 ± 25 1041-1202 Mweru cave (occupation deposit) Roulette-decorated 940 ± 26 1045-1214 Site (sample origin) Corr. "C age D., t 425-573 266-534 11 (occupation OxA-19521 1040-1201 1042-1206 (burnt floor) cave (occupation deposit) OxA-19811 ; A few other seed remains could be assigned probable - Sequence Rwanda'lronAge'{A-101.6%} ^ , family or genus identifications Sequence remained indeterminate. A single nutlet apparently of the ¡¡¡¡unjan: class« Uivtrf —family Polygonaceae (cf. Persicaria), came from Kabu phase classic Urewe sanze. Individual finds of biomineralised Celtis sp. OxA-19517 96 3% (Pig- 3c), and Boraginaceae type seeds from Musanze II are probably ancient, but could be modern intrusions. OxA-19518 112.6% A ~ J, ' Charred fragments of parenchyma OxA-19583 104.6%— ''A — recovered OxA-19520 101.3% when modern analogue materials are available from the Boundary region, nevertheless the presence of parenchyma suggests the use of some roots or tubers (cf. Hather 2000). The cereal crops pearl millet (Pennisetum glaucum R. Br.) and sorghum (Sorghum bicolor [L.] Moench. bicolor of probable race caudatum of Harlan and Phase RouiieiiedDecoration 1972) were identified from the mid-first millennium OxA-19521 98.7% a.D. site of Kabusanze (Figs. 3a, 4, 5b). The club-shaped OxA-19322 95.9% Í profile of pearl millet suggests domesticated grain mor phology (D'Andrea et al. 2001), while its size (1.75 mm wide, by 1.07 mm thick) is comparable to domesticated populations from the first millennium a.d. at Jarma (Libya) and Qasr Ibrim (Egyptian Nubia), as well as "Iron Age" Boundary ÉÉMN Indian measurements (Fuller 2007, Fig. 15; Fuller et al. —i I , , , i 1 i , , i I , i . i ! . . i ^ L_ rPhase Transitional Unerve OxA-19523 99.0% Ox.4-age BC'AD 500AD Urcwc Rouicttcd 2010, Fig. 3). It is also consistent with the most widespread varieties of the pearl millet group, race typhoides Brunken et al. 1977), and the better-preserved Pennisetum glaucum — . . , , ~ ^ i • * , Sorghum bicolor — gram Vigna cf. unguiculata —— as race CQudfltlWl. Eleusine coracana Legume seed fragments or cotyledons were found at , . , , , Kabusanze associated with the pearl millet and sorghum Fig. 2 Bayestan analysis of the calendncal ages of the sites based on . the available radiocarbon ages. Dates have been grouped into ceramic 8rams- Thelr ovold< barrel-like shape phases, grey shading indicates the suggested focal periods of Classic plumules about half the length of the seed Urewe user, transitional Urewe user and roulette-decorated ceramic being Vigna (Figs. 3h, 5a). Their small user, occupations on these sites. The chronological distribution of rved hilum means that they show overlap crops identified is indicated at the base of the chart. Calibrations were \ . . performed with OxCal 3.10 (Bronk-Ramsey 2009) using the updated Vigna spp. including the wild calibration curve of IntCal09 (Reimer et al. 2009) (V. unguiculata [L.] Walp. ssp. dekindtiana (Ö Springer Sequence Rwanda' Iron Age' {A= 101.6%} Sequence Boundary: Classic Urewe Phase Classic Urewe OxA-19517 96.3% OxA-19518 112.6% OxA-19583 104.6%. OxA-19520 101.3% Boundan• r Phase Transitional Urewe OxA-19519 99.4% Boundan' Phase Routletted Decoration OxA-19521 98.7% OxA-19522 95.9% OxA-19524 98.6% OxA-19811 99.5% Boundary J . L_J I 1 I . I I .—» .— Calibrated age BC/AD 500AD 1000 1500 Crops present Urewe Rouletted Pennisetum glaucum Sorghum hicolor Vigna cf. unguiculata Eleusine coracana 2 Bayesian analysis of the calendrical ages of the sites based on available radiocarbon ages. Dates have been grouped into ceramic grey shading indicates the suggested focal periods of Classic Urewe user, transitional Urewe user and roulette-decorated ceramic user, occupations on these sites. The chronological distribution of identified is indicated at the base of the chart. Calibrations were performed with OxCal 3.10 (Bronk-Ramsey 2009) using the updated calibration curve of IntCal09 (Reimer et al. 2009)This content downloaded from 146.230.187.153 on Tue, 16 Apr 2019 07:08:03 UTCAll use subject to https://about.jstor.org/terms
258 Veget Hist Archaeobot (2011) 20:253-265 Table 3 Samples with identified seeds and radiocarbon Site (site code) Sample Context type (no) [sample] Plant remains (seed/frag, quantity) Assoc. I4C date (a.d., approx. modal calibration) 12 Pit Fill (53) [6] Vigna cf. unguiculata (2) ca. 420 12 Pit (ill (53) [7] Vigna cf. unguiculata (3) Charred ca. 420 12 Pit fill (53) [8] Wood charcoal only ca. 420 30 Grave fill (57) [9] Polygonaceae (1) Charred ca. 420 30 Grave fill (57) [10] Wood charcoal only ca. 420 30 Pit fill (59) [11] Pennisetum glaucum (1) Charred ca. 380 24 Pit fill (53) [12] Sorghum bicolor (1) Charred ca. 420 Masangano (MSG) 24 Deposit (70) [13] ca. 380 Karama (GPS014) 12 Pit fill (39) [3] coracana (3) ca. 750 12 Deposit (32) [4] Eleusine coracana (1) Charred 12 Pit fill (41) [5] Seed fragment Charred Musanze II (M2) 12 Eleusine coracana (7) Biomineralised 12 (76) [15] Parenchyma (1), Boraginaceae (1), Celtis (1) Biomineralised 12 (77) [16] Eleusine coracana (1) ca. 1030/1110/1150 Musanze III (M3) 12 coracana (5) Charred 6 [18] (80) Eleusine coracana (4) Charred 12 None 12 Deposit (85) [20] ca. 1020 12 (84) [21] None Musanze IV (M4) 12 (3) [1] Seed fragments (1) Charred 12 (5) [2] Parenchyma (1) Nguri cave (RPS003) 12 [28] None 12 Eleusine coracana (3), Parenchyma (1), indet. fragment (1) 12 Deposit (119) [30] Eleusine coracana (2) 12 Eleusine coracana (3), seed (1) ca. 1030/1110/1150 12 Deposit (121) [32] None Mweru cave (RPS015) 6 12 12 12 Burnt floor (95) [22] Deposit (98) [23] Wood charcoal only None None None ca. 1030/1110/1150 12 Deposit (101) [26] Parenchyma (1) Charred 12 Deposit (102) [27] Site (site code) size (1) remains frag, quantity) Preservation D., Kabusanze (BPS 036) (2) Charred fill (53) [7] cf. unguiculata (3) (53) [12] bicolor (1) (MSG) Wood charcoal only (GPS014) fill (39) [3] Eleusine coracana (3) Charred Deposit (75) [14] Deposit (76) [15] Eleusine coracana (1), (1), (1), Celtis (1) Deposit (77) [16] Charred III (M3) Deposit [17] (79) Eleusine coracana (5) Deposit [18] (80) Deposit [19] (81) Wood charcoal only Deposit (84) [21] Deposit (3) [1] (1) Deposit (5) [2] (1) Charred cave (RPS003) Deposit (117) [28] Deposit (118) [29] Charred (119) [30] (2) Charred Deposit (120) [31] coracana (3), cf. Lablab (2), fruit seed (1) Charred Deposit (99) [24] Deposit (100) [25] 26] 1) None All samples with seeds also contained wood charcoal which is widespread across Africa. The UCL reference The Kabusanze specimens would fit with a small-seeded collection is incomplete for wild African Vigna spp., of form of cultivated cowpea (Vigna unguiculata [L.] Walp. which many species are reported from the region (there are ssp. unguiculata), which seems a plausible interpretation 30 species in southeast Africa, Mackinder et al. 2001), and given the presence of other crops at this time. The only a few wild taxa could be excluded based on our ref- domesticated form of cowpea probably originated in wes erence collection (V. radiata (L.) R. Wilczek var. sublo- tern Africa, Ghana or thereabouts, which is supported both bata (Roxb.) Verde., V. luteola (Jacq.) Benth., V. vexillata by archaeobotany, genetics and linguistics (Fuller 2003; (L.) A. Rich.). Blench 2006; D'Andrea et al. 2007). However, our £) SpringerThis content downloaded from 146.230.187.153 on Tue, 16 Apr 2019 07:08:03 UTCAll use subject to https://about.jstor.org/terms
Veget Hist Archaeobot (2011) 20:253-265 Fig. 3 SEM images of selected specimens: a Sorghum bicolor cf. race caudatum, ventral view (damaged). Kabusanze context BPS036 (56); b example of parenchyma tissue, indet. type, Musanze IV context M4 (2); c Celtis sp., seed, Musanze II context M2 (77); d Eleusine coracana, dorsal view, Musanze II context M2 (77); e close-up of pusticulate surface of E. coracana, Musanze II context M2 (77); f cf. Lablab, interior view of cotyledon, Nguri Cave context RPS003 (120); g nutlet, cf. Persicaria dorsal view, Kabusanze context BPS036 (56); h Vigna cf. unguiculata, interior view of cotyledon, Kabusanze context BPS036 (56) Fig. 4 Drawings of Kabusanze cereals. Sorghum bicolor (above) and Pennisetum glaucum (below) showing (from left to right) the dorsal view, longitudinal section, ventral view, and cross-section (of Pennisetum). Note that both specimens are quite damaged. Sorghum has undergone some 'popcorn' like puffing on the dorsal surface, and the approximate outline of the scutellum is indicated by dashed line specimens are attributed to Vigna cf. unguiculata to indi- material from South India (Fuller et al. 2004). However, a cate the remaining uncertainty over whether these are wild wider range of comparative material of legume species of or domesticated cowpea, although they do overlap with east Africa has not been consulted. Hyacinth bean is wild small, cultivated cowpea varieties. and cultivated in eastern Africa (Maass et al. 2005), and At Musanze II a single large legume fragment repre- has been previously reported archaeologically sented another pulse (Fig. 3f). This fragmentary cotyledon (first century cal. a.D.) from Geduld Rock Shelter has a size, shape and plumule comparable to that of the wild plant foods and evidence for early pastoralism hyacinth bean (Lablab purpureas [L.] Sweet), but in the and Jacobson 1995). It was also a crop of Late absence of preserved testa or remnants of a hilum it is not Nubia (early centuries a.d.) at Qasr Ibrim (possible to definitively identify this to species. The Hya- Rowley-Conwy 2007). The appearance of cinth bean specimen from Musanze II is also similar to alongside other crops of African origin in India large quantities of Lablab in various states of fragmenta- second millennium b.c. (Fuller 2003) implies tion observed by the second author in late Neolithic tion began somewhere as yet undetermined in <£) SpringerThis content downloaded from 146.230.187.153 on Tue, 16 Apr 2019 07:08:03 UTCAll use subject to https://about.jstor.org/terms
260 Veget Hist Archaeobot (2011) 20:253-265 Fig. 5 Photomicrographs of selected specimens: a Vigna ulata section and interior view of cotyledon, Kabusanze BPS036 (56); b Pennisetum glaucum, ventral context BPS036 (59); c Eleusine coracana, dorsal view, context M3 (85); d Eleusine context M2 (85); e Eleusine scutellum. and lateral view, Musanze coracana, dorsal and lateral view, Africa (perhaps Ethiopia) a few millennia before to southeast Africa. Stemler 1990). Whilst the Rwandan material _ .„ . , leaves no doubt as to its specific identity as finger millet, it Finger millet (Eleusme coracana) , , . _ . . . . . . ... ... should be noted that Rwanda is progenitor form, E. africana (1979) and Phillips (1972) provide distinguishing these taxa, but ulations of E. ssp. africana have the possibility of tracking archaeobotanically of domestication syndrome traits varieties needs to be better developed. Of the later samples, all of the identifiable cereal remains were finger millet (E. coracana [L.] Gaertn.). This came in the form of both carbonized grains (Fig. 5c, d) and apparently untransformed grains, some of which appear quasi-mineralised (Fig. 5e, f). We interpret these uncar bonized examples as ancient and preserved due to the locally dry and probably mineral rich conditions inside the excavated cave site of Musanze II. In the discussion that follows we use the short binomials of E. coracana (L.) Gaertn. and E. africana Kennedy O'Byrne for the domesticated and wild form, respectively, rather than the more cumbersome ssp. coracana and ssp. africana (of the taxonomy of Hilu and De Wet 1976a). Eleusine coracana has distinctive sub-globose grains, with somewhat angular edges, a short embryo on the dorsal surface, and a characteristically finely beaded (pusticulate) pericarp surface (Fig. 6; Hilu and De Wet 1976a; Fuller 2003), as illustrated in Fig. 3e; and carbonized examples have often puffed giving them somewhat distorted and exaggerated length and breadth measurements. However, as explored elsewhere there have been numerous misi dentified reports of "finger millet" in the archaeology of prehistoric India (Hilu et al. 1979; Fuller 2003). Similarly, early (un-illustrated) reports of finger millet impressions from Neolithic pottery of the Sudan may be regarded as ITI . . Fig- 6 Drawing of an Eleusine coracana specimen trom Musanze lit probably inaccurate (Klichowska 1984), given that sub- (19), in dorsal and basal view with a sequent specialist study and scanning electron microscopy of this specimen preserves fine •£) Springer Fig. 6 Drawing of an Eleusine coracana specimen from (79), in dorsal and basal view with a longitudinal section. of this specimen preserves fine pusticulaeThis content downloaded from 146.230.187.153 on Tue, 16 Apr 2019 07:08:03 UTCAll use subject to https://about.jstor.org/terms
Veget Hist Archaeobot (2011) 20:253-265 261 2.5 2 E E i? 15 CD O 1 0.5 * modem E. coracana • E. africana ♦ Eleusine (carbonized) ■ Eleusine (mineralized) are responsible for their rarity (having already ruled out archaeological collection strategies). In essence, because E. coracana is a free-threshing cereal, and as such seasonal threshing is likely be carried out off site, and on-site de „ * husking is unnecessary, there may be few opportunities for x * * • " charring. Similar issues will affect the likelihood of char X X+ X X ♦ ^ *4 . , » •—» ring in free-threshing varieties of Pennisetum and Sorghum I W * (such as race caudatum). Despite the identification of E. * * v * coracana during this study, the occurrence of this species XXX in very low frequency (1999) findings. 0.5 1 1.5 2 2.5 Discussion Grain Length mm These data constitute Fig. 7 Scatter plot of length and width measurements on archaeo- . „ . , _ , ... logical grains (raw data in ESM Table 1). Carbonized and mineral- remalns ln Rwanda and ized finger millet grains are distinguished. Shown for comparison are throughout the first the ranges for modern domesticated finger millet (E. coracana), based nological research on our measurements (ESM Table 2) and its wild progenitor (yan Grunderbeek and (E. africana), based on the ranges reported in the literature ...... ., r , ■ ■ r , (Kennedy-O'Byrne 1957; Phillips 1972; Hilu and De Wet 1976b) Provlde deflnit|ve evidence cated grain crops with secure species identifications. The absence of the wild form in our reference material Cereal and pulse agriculture of (at UCL) has meant that we relied on published illustra- period tions and descriptions for determining domestication status. These indicate that E. africana has a much more ridged The mid-first millennium surface pattern (approaching that of E. indica with which it Kabusanze, found in has sometimes been classified) and generally thinner more ceramics suggest that Classic elongated seed forms than E. coracana, with length-width agriculture, which included ratios of around 1.5 (Kennedy-O'Byrne 1957; Phillips cowpea. Historical linguistic 1972; Hilu and De Wet 1976a; Hilu et al. 1979). In con- Central/Eastern Sudanic speakers trast, the Rwandan archaeobotanical material overlaps with pearl millet to the region, modern domesticated populations, and has globose length- Bantu speakers (Schoenbrun to-width ratios of around 1 (Fig. 7). Although some millet has a wide climatic archaeological specimens appear larger than the modern little as 200 mm rainfall up reference material measured by us, we have only measured grown on the lower half of a limited number of modern landraces, which cannot be some distribution of water regarded as wholly representative of the size range of all Some varieties can mature cultivars. In addition, some of the carbonized grains appear 1980; Bourlag et al. 1996). to have puffed during charring. Nevertheless, on morpho- more moisture but is tolerant logical grounds the material identified here appears to be varieties adapted to a wide entirely domesticated E. coracana. ghum and pearl millet tend to be short-Ethnoarchaeological work by Young (1999) in Bunyoro, after the longest days of summer, Uganda, considered the possibility that the scarcity of to day length are also known domesticated E. coracana in archaeological deposits was regions with two peaks in rainfall, due to a lack of charring during processing. However, she are often sown during the first found that "traditional" practices in Bunyoro did involve 3 months later (e.g. March/heating and the charring of grains and there was a high easily be intercropped with probability that such processing would probably leave (Westphal 1974, p. 230). From archaeological remains; suggesting that either a different would suggest that cultivation processing practice took place in early farming communi- the April peak in rains with ties in Great Lakes Africa or poor preservation conditions September. <£) Springer 2.5 £ tj ^ 15 c 2 CD * * * ♦ x *© * * ° -*♦-*- -*■ > X *• X X B% XXX ■ * * -I -L «_ 0.5 1 1.5 2 2.5 Grain Length mm Fig. 7 Scatter plot of length logical grains (raw data in ESM ized finger millet grains are distinguished. the ranges for modern domesticated on our measurements (ESM Table 2) and its wild progenitor (E. africana), based on the ranges reported in the literature (Kennedy-O'Byrne 1957; Phillips 1972; Hilu and De Wet 1976b)This content downloaded from 146.230.187.153 on Tue, 16 Apr 2019 07:08:03 UTCAll use subject to https://about.jstor.org/terms
262 Veget Hist Archaeobot (2011) 20:253-265 While both caudatum sorghum and pearl millet widespread as traditional crops in eastern Africa, regarded as having considerable time depth torical linguistics (e.g. Philippson and Bahuchet 1996; gallus), sheep (Ovis aries) and goat (Capra hircus). At Ehret 1998; Blench 2006), they originated in different parts Nguri cave a single cotyledon comparable to hyacinth bean of Africa. Pearl millet is accepted as western African in (cf. Lablab) was recovered, and indicates that legumes origin, with the earliest evidence of the domesticated type were also consumed and probably cultivated. Hyacinth from northeast Mali dated to 2400-2000 cal b.c. (Manning bean is frequently grown in upland areas, intercropped with et al. 2011). Pearl millet was fairly widespread as a crop in millets, and can tolerate rainfall of 900 mm as well as drier western Africa by 1700-1500 cal b.c. (Fuller 2007; Kahl- conditions (Westphal 1974, p. 102). heber and Neumann 2007). Sorghum, by contrast, is usu ally regarded as originating in northeast Africa, such as in The origin and spread of finger millet eastern Chad or Sudan (Harlan 1992). Initial domesticated (Eleusine coracana) sorghum would have been race bicolor, with a persistent hull, whereas the material here probably represents an Finger millet is often grown at high elevations advanced free-threshing sorghum with a larger grain. The (1,000-2,000 m a.s.l.) in Africa, where rainfall is above race caudatum is usually regarded as having evolved later 900 mm, and crops can mature in 3.5-6 months (Hülse in western Africa (Harlan and Stemler 1976). However, et al. 1980). It is normally a short-day plant meaning that it recent genetic evidence suggests that caudatum types of was probably planted after the April rains and harvested southeast Africa form a distinct group raising the possi- around August. The variety of regional terms applied to bility of a parallel origin of caudatum morphotypes in finger millet (e.g. Ehret 1998, p. 129) suggests that this eastern Africa or India (cf. De Alencar Figueiredo et al. crop was adopted later than sorghum and pearl millet, 2008). What is clear, however, is that both pearl millet and although glottochronological estimates from linguistics sorghum initially dispersed separately in the northern parts remain too old by comparison to current archaeobotany. of Africa. However, the origins and early dispersal of finger millet The present evidence indicates that they had been inte- remain poorly understood. grated in east-central African agriculture by at least the Harlan's (1971) inference mid-first millennium a.d. Later sites of the late first or somewhere between second millennium a.d. did not produce evidence for either although the distribution of these crops. However, the absence of these crops should much of Africa (not necessarily be regarded as a change in the focus of 1984). There are no reports agriculture because these later sites occur in different early cultivation or the regions. In addition, archaeobotanical sampling across traits, although Lange Rwanda has been very limited. africana seed from a deposit with Oltome (i.e. Kansyore, pre-Classic Urewe) ceramics. Finger millet cultivation from the late The date for the arrival of finger millet in India provides first millennium a.d. a minimal age for the start of cultivation in its African homeland, but in the absence of better documentation, The late first millennium a.d. results from Karama, identified larger quantities and direct AMS-dating, this minimum age alongside a transitional Urewe ceramic, combined evidence could be anywhere from 2500 to 500 b.c. By early historic for finger millet cultivation with a mixed domestic and wild times in India, from 300 b.c., finds are more numerous. The zooarchaeological assemblage made up of cattle (Bos taurus), earliest finds from India, once probable misidentifications leopard (Panthera pardus), antelope (Antilopinae) and giant- are removed, consist of only a few sites as old as 1000 cal forest-hog (Hylochoerus meinertzhageni). These data suggest b.c. (Fuller 2003), including examples represented by a that hunting remained an important economic component solitary specimen such as from Hallur (Fuller et al. 2004), alongside fanning during this period. now associated with direct AMS dates on other crops from The evidence from the northern caves at Musanze, the same context (Fuller et al. 2007). The earliest claimed Mweru and Nguri suggests that mixed strategies continued Indian evidence comes from Rojdi between 2500 and throughout the second millennium a.d. It is believed that 1700 cal b.c. (Weber 1991), but it remains disputed (Weber these caves were occupied intermittently until the late 19th 2001; Fuller 2003; 2006, p. 37). There are no directly dated century based on local histories and the excavated assem- examples from India. blages (Nenquin 1967). Within the caves domesticated Within Africa, there are now 17 sites or site phases with finger millet was found in association with roulette- reported evidence for finger millet (Fig. 8; ESM Table 3). £) SpringerThis content downloaded from 146.230.187.153 on Tue, 16 Apr 2019 07:08:03 UTCAll use subject to https://about.jstor.org/terms
Veget Hist Archaeobot (2011) 20:253-265 263 Fig. 8 Distribution of archaeobotanical reports of finger millet or possible finger millet (including 'cf. coracana' grains and more dubious 'cf. Eleusine' pollen). These are shown in relation to an approximation of the distributional limits of modern finger millet cultivation (after Hilu and De Wet 1976b). Sites numbered: 1 Kursakata; 2 Ona Nagast; 3 Axum; 4 Gobedra Cave; 5 Kibuye; 6 Karama; 7 Musanze; 8 Deloraine; 9 Pemba sites—Tumbe, Chwaka; 10 Kadzi; 11 Ziwa (Inyanga); 12 Magogo; 13 Shongweni The earliest find comes from a single possible specimen in Adoption of a basic sampling strategy involving routine the upper "Iron Age" levels of Kursakata near Lake Chad, flotation from all contexts allowed a range of wild and which is estimated as first century a.d. (Klee et al. 2000). domesticated species to be identified. Although these Other finds from the start of the first millennium remains are in very low quantities they make an important a.d. include those from Ona Nagast near Axum in Ethiopia empirical contribution to our understanding of food-ways (D'Andrea 2008). The material reported here, and that and agriculture in the later prehistory of Great Lakes Africa, found elsewhere in eastern Africa (excluding the claimed Our results highlight the need for more research projects in pollen identification), is congruent with a later spread of the region to incorporate archaeobotanical sampling strate finger millet southwards in the middle of the first millen- gies despite the likelihood of poor preservation because even nium a.d., although archaeobotanical sampling has been so small pieces of the puzzle can contribute toward the devel scarce in these parts of Africa that any regional pattern is opment of more detailed subsistence histories, as demon currently tenuous. strated by the Rwandan material presented here. Conclusion Acknowledgments We thank our reviewers for their constructive comments on this text. The Arts and Humanities Research Council funded the first author's PhD research at University College London, with additional fieldwork assistance from the Institute of Archaeology Awards, University College London Graduate School, the Central This project has successfully demonstrated that archaeobo- Research Fund of the University of London. Assistance tanical remains can be identified in first and second millen- provided in Rwanda by the Institute of National Museums . , ■ . „t , . . (INMR) and the British Institute in Eastern Africa. Andrew Reid mum a.d. archaeological contexts in Great Lakes Africa .■' . , , , ,, , D supervised the first author s doctoral research, and the radiocarbon despite the poor potential for the preservation of plant dates were obtained through remains in this region (Young and Thompson 1999). awarded to A. Reid. <£) Springer Fig. 8 Distribution of archaeobotanical reports of finger millet or possible finger millet (including 'cf. coracana' grains and more dubious 'cf. Eleusine' pollen). These are ShongweniThis content downloaded from 146.230.187.153 on Tue, 16 Apr 2019 07:08:03 UTCAll use subject to https://about.jstor.org/terms
264 Veget Hist Archaeobot (2011) 20:253-265 References Abdel-Magid A (1989) Plant domestication in the Middle Nile Basin: an archaeoethnobotanical case study. British Archaeological Reports, Oxford Ambrose SH, Collett D, Modida, Marshall F (1984) Excavations at Deloraine, Rongai, 1978. Azania 19:79-104 Ashley CZ (2005) Ceramic variability and change: a perspective from Great Lakes Africa. Dissertation, University College London Ashley CZ (2010) Towards a socialised archaeology of ceramics in Great Lakes Africa. Afr Archaeol Rev 27:135-163 Blench R (2006) Archaeology, language and the African past. AltaMira Press, Oxford Bourlag N, Axtell J, Burton G, Harlan J, Rachie K (1996) Lost crops of Africa, vol 1, Grains. National Academy Press, Washington DC Bronk-Ramsey C (2009) Bayesian analysis of radiocarbon dates. Radiocarbon 51:337-360 Brunken J, De Wet JMJ, Harlan JR (1977) The morphology and domestication of pearl millet. Econ Bot 31:163-174 Chretien J-P (2003) The Great Lakes of Africa: two thousand years of history (translated by Scott Straus). Zone Books, New York Clapham A, Rowley-Conwy P (2007) Rewriting the history of African agriculture. Planet Earth 2006:24—26 Clist B (1987) A critical reappraisal of the chronological framework of the early Urewe Iron Age industry. Muntu 6:35-62 D'Andrea AC (2008) T'ef (Eragrostis tef) in ancient agricultural systems of highland Ethiopia. Econ Bot 62:547-566 D'Andrea AC, Klee M, Casey J (2001) Archaeobotanical evidence for pearl millet (Pennisetum glaucum) in sub-Saharan West Africa. Antiquity 75:341-348 D'Andrea AC, Kahlheber S, Logan AL, Watson DJ (2007) Early domesticated cowpea (Vigna unguiculata) from Central Ghana. Antiquity 81:686-698 De Alencar Figueiredo LF, Calatayud C, Dupuits C, Billot C, Rami J-F, Brunei B, Perrier X, Courtois B, Deu M, Glaszmann J-C (2008) Phylogeographic evidence of crop neodiversity in sorghum. Genetics 179:997-1.008 De Wet JMJ, Prasada Rao KE, Brink DE, Mengesha MH (1984) Systematics and evolution of Eleusine coracana (Gramineae). Am J Bot 71:550-557 Desmedí C (1991) Poteries anciennes décorées ä la roulette dans la region des Grands Lacs. Afr Archaeol Rev 9:161-196 Ehret C (1998) An African classical age: eastern and southern Africa in world history, 1000 b.c. to a.d. 400. University Press of Virginia, Charlottesville FAO (2010) Interpolated estimated Dekadal rainfall—by region/ province. Food and Agriculture Organization of the United Nations website, http://www.fao.org/giews/english/index.htm. Accessed 25 July 2010 Fuller DQ (2003) African crops in prehistoric South Asia: a critical review. In: Neumann K, Butler A, Kahlheber S (eds) Food, fuel and fields: progress in African archaeobotany. Heinrich-Barth Institute, Köln, pp 239-271 Fuller DQ (2006) Agricultural origins and frontiers in South Asia: a working synthesis. J World Prehist 20:1-86 Fuller DQ (2007) Contrasting patterns in crop domestication and domestication rates: recent archaeobotanical insights from the Old World. Ann Bot 100:903-924 Fuller DQ, Korisettar R, Vankatasubbaiah PC, Jones MK (2004) Early plant domestications in southern India: some preliminary archaeobotanical results. Veget Hist Archaeobot 13:115-129 Fuller DQ, Boivin N, Korisettar R (2007) Dating the Neolithic of South India: new radiometric evidence for key economic, social and ritual transformations. Antiquity 81:755-778 Fuller DQ, Allaby RG, Stevens C (2010) Domestication as innova tion: the entanglement of techniques, technology and chance in the domestication of cereal crops. World Archaeol 42:13-28 Giblin JD (2010) Re-constructing the past in post-genocide Rwanda: an archaeological contribution. Dissertation, University College London Harlan JR (1971) Agricultural origins: centers and noncenters. Science 174:468-474 Harlan JR (1992) Indigenous African agriculture. In: Cowan CW, Watson PJ (eds) The origins of agriculture: an international perspective. Smithsonian Press, Washington, pp 59-70 Harlan JR, De Wet JMJ (1972) A simplified classification of cultivated sorghum. Crop Sei 12:172-176 Harlan JR. Stemler R (1976) The races of sorghum in Africa. Harlan JR, De Wet JMJ, Stemler ABL (eds) Origins of African plant domestication. Mouton Publishers, Illinois, pp 465-478 Hather JG (2000) Archaeological parenchyma. Archetype Publica tions, London Helbaek H (1969) Plant collecting, dry farming and irrigation agriculture in prehistoric Deh Luran. In: Hole F, Flannery KV, Neely JA (eds) Prehistory and human ecology of Deh Luran Plain. Memoirs of the Museum of Anthropology. University Michigan 1, Ann Arbor, pp 383^126 Hilu KW, De Wet JMJ (1976a) Domestication of Eleusine coracana. Econ Bot 30:199-208 Hilu KW, De Wet JMJ (1976b) Racial evolution in Eleusine coracana ssp. coracana (finger millet). Am J Bot 63:1,311-1,318 Hilu KW. De Wet JMJ, Harlan JR (1979) Archaeobotanical studies Eleusine coracana ssp. coracana (finger millet). Am J Bot 66:330-333 Hülse JH, Laing EM, Pearson OE (1980) Sorghum and the millets: their composition and nutritive value. Academic Press, London Kahlheber S, Neumann K (2007) The development of plant cultivation in semi-arid West Africa. In: Denham T, Iriarte J, Vrydaghs L (eds) Rethinking agriculture: archaeological and ethnoarchaeo logical perspectives. Left Coast Press, Walnut Creek, pp 318-344 Kennedy-O'Byrne J (1957) Notes on African grasses: XXIX, a new species of Eleusine lfom Tropical and South Africa. Kew Bull 12:65-72 Kiage L. Liu K (2009) Palynological evidence of climate change and land degradation in the Lake Baringo area, Kenya, East Africa, since A.D. 1650. Palaeogeogr Palaeoclimatol Palaeoecol 279:60-72 Klee M, Zach B, Neumann K (2000) Four thousand years of plant exploitation in the Chad Basin of northeast Nigeria: the archaeobotany of Kursakata. Veget Hist Archaeobot 9:223-237 Klichowska M (1984) Plants of the Neolithic Kadero (Central Sudan): a palaeoethnobotanical study of the plant impressions on pottery. In: Krzyzaniak L, Kobusiewicz M (eds) Origin and early development of food producing cultures in north-eastern Africa. Polish Academy of Sciences, Poznan, pp 321-326 Lamb H, Darbyshire I, Verschuren D (2003) Vegetation response to rainfall variation and human impact in central Kenya during the past 1100 years. Holocene 13:285-292 Lane P, Ashley A, Setsonen O, Harvey P, Mire S, Odede F (2007) The transition to farming in eastern Africa: new faunal and dating evidence from Wadh Lang'o and Usenge 3, Kenya. Antiquity 81:62-81 Lange G (1991) Appendix V, a seed of wild finger millet from Gogo Falls. In: Robertshaw P. Gogo Falls. Azania 26:191-192 Maass BL, Jamnadass RH, Hanson J, Pengelly BC (2005) Determin ing sources of diversity in cultivated and wild Lablab purpureus related to provenance of germplasm by using amplified fragment length polymorphism. Genet Resour Crop Evol 52:683-695 SpringerThis content downloaded from 146.230.187.153 on Tue, 16 Apr 2019 07:08:03 UTCAll use subject to https://about.jstor.org/terms
Veget Hist Archaeobot (2011) 20:253-265 265 Mackinder B, Pasquet R, Polhill R, Verdcourt B (2001) Zambesiaca, vol 3, pt 5, Leguminosae. London, Royal Botanical Gardens at Kew. http://apps.kew.org/efloras/fz/intro.Accessed 25 July 2010 Manning K, Pelling R, Higham T, Schwenniger J-L, Fuller 4500-year old domesticated pearl millet (Pennisetum glaucum) from the Tilemsi Valley, Mali: new insights into an alternative cereal domestication pathway. J Archaeol Sei 38:312-322 Nenquin J (1967) Contributions to the study of the prehistoric of Rwanda and Burundi. Ann Musée Royal de L'Afrique Centrale, Tervuren Pearsall D (2000) Palaeoethnobotany: a handbook of procedures, edn. Academic Press, New York Philippson G, Bahuchet S (1996) Cultivated crops and migrations in central and eastern Africa: a linguistic approach. In: Sutton JEG (ed) The growth of farming communities Africa from the equator southwards. Azania 29:103-120 Phillips SM (1972) A survey of the genus Eleusine (Gramineae) in Africa. Kew Bull 27:251-270 Phillipson DW (1993) African archaeology, 2nd edn. Cambridge University Press, Cambridge Posnansky M (1967) The Iron Age in East Africa. In: Clark JD (eds) Background to evolution in Africa. University Chicago Press, Chicago, pp 629-649 Posnansky M (1973) Terminology in the early Iron Age Africa with particular reference to the dimple-based wares Lolui Island, Uganda. In: PanAf congress for prehistory associated studies, vol 577-579 Posnansky M, Reid A, Ashley C (2005) Archaeology on Lolui Island, Uganda 1964-5. Azania 40:1-29 Rattray J (1960) The grass cover of Africa. FAO, Rome Reid DAM (1996) Ntusi and the development of social complexity in southern Uganda. In: Pwiti G, Soper R (eds) Aspects of African archaeology. University of Zimbabwe Press, Harare, pp 621-628 Reimer PJ, Baillie MGL, Bard E, Bayliss A, Beck JW, Weyhenmeyer CE (2009) Intcal09 and Marine09 radiocarbon age calibration curves, 0-50,000 years cal B.P. Radiocarbon 51:1,111-1,150 Robertshaw P (1997) Münsa earthworks: a preliminary report. Azania 32:1-20 Robertshaw P, Collet D (1983) A new framework for the study of early pastoral communities in East Africa. J Afr Hist 24:289-301 Schoenbrun DL (1993) We are what we eat: ancient agriculture between the Great Lakes. J Afr Hist 34:1-31 Schoenbrun DL (1998) A green place, a good place: agrarian change, gender, and social identity in the Great Lakes region to the 15th century. James Currey, Oxford Smith AB, Jacobson L (1995) Excavations at Geduld and the appearance of early domestic stock in Namibia. S Alf Archaeol Bull 50:3-14 Stemler AB (1990) A scanning electron microscope analysis of plant impressions in pottery from sites of Kadero, El Zakiab, Urn Direiwa and El Kadada. Arch Nil Moyen 4:87-106 Taylor D, Robertshaw P, Marchant RA (2000) Environmental change and political-economic upheaval in pre-colonial western Uganda. Holocene 10:527-536 Van Grunderbeek M-C (1988) Essai d'étude typologique de cérami que urewe dans la region des collines au Burundi et Rwanda. Azania 23:11-55 Van Grunderbeek M-C (1992) Essai de délimitation Chronologie de L'Age du Fer Ancien au Burundi, au Rwanda et dans la région des Grands Lacs. Azania 27:53-80 Van Grunderbeek M-C, Roche E (2007) Multidisciplinary evidence of mixed farming during the early Iron Age in Rwanda and Burundi. In: Denham T, Iriate J, Vrydaghs L (eds) Rethinking agriculture: archaeological and ethnoarchaeological perspec tives. Left Coast Press, Walnut Creek, pp 299-319 Vansina J (2004) Antecedents to modern Rwanda: the Nyiginya Kingdom. Currey, Oxford Walshaw SC (2005) Swahili urbanization, trade, and food production: botanical perspectives from Pemba Island, Tanzania, a.D. 700-1500. Dissertation, Washington University in St. Louis Walshaw SC (2010) Converting to rice: urbanization, Islamization and crops on Pemba, a.D. 700-1500. World Archaeol 42:137-154 Weber SA (1991) Plants and Harappan subsistence: an example of stability and change from Rojdi. Oxford & IBH, New Delhi Weber SA (2001) Seeds of urbanism revisited. Antiquity 75:413-414 Westphal E (1974) Pulses in Ethiopia: their taxonomy and agricul tural significance. Centre for Agricultural Publishing and Doc umentation, Wageningen Wetterstrom W (1991) Appendix IV, Plant remains from Gogo Falls. In: Robertshaw P (ed) Gogo Falls. Azania 26:180-191 White F (1983) The vegetation of Africa: a descriptive memoir to accompany the UNESCO/AETFAT/UNSO vegetation map of Africa. UNESCO, Paris Young R (1999) Finger millet processing in East Africa. Veget Hist Archaeobot 8:31-34 Young R, Thompson G (1999) Missing plant foods? Where is the archaeobotanical evidence for sorghum and finger millet in East Africa? In: Van der Veen M (ed) The exploitation of plant resources in ancient Africa. Kluwer/Plenum Publishers, New York, pp 63-72 SpringeThis content downloaded from 146.230.187.153 on Tue, 16 Apr 2019 07:08:03 UTCAll use subject to https://about.jstor.org/terms