Amer. J. Bot. 71(4): 550-557. 1984.

SYSTEMATICS AND EVOLUTION OF CORACANA (GRAMINEAE)1

J. M. J. de W et,2 K. E. Prasada Rao,3 D. E. Brink,2 and M. H. Mengesha3 departm ent of Agronomy, University of Illinois, 1102 So. Goodwin, Urbana, Illinois 61801, and international Crops Research Institute for the Semi-arid Tropics, Patancheru,

ABSTRACT Finger ( (L.) Gaertn. subsp. coracana) is cultivated in eastern and southern Africa and in southern Asia. The closest wild relative of finger millet is E. coracana subsp. africana (Kennedy-O’Byme) H ilu & de Wet. W ild finger m illet (subsp. africana) is native to Africa but was introduced as a weed to the warmer parts of Asia and America. Derivatives of hybrids between subsp. coracana and subsp. africana are companion weeds of the crop in Africa. Cultivated finger are divided into five races on the basis of inflorescence mor­ phology. Race coracana is widely distributed across the range of finger millet cultivation. It is present in the archaeological record o f early African agriculture that m ay date back 5,000 years. Racial evolution took place in Africa. Races vulgaris, elongata., plana, and compacta evolved from race coracana, and were introduced into India some 3,000 years ago. Little independent racial evolution took place in India.

E l e u s i n e Gaertn. is predominantly an African tancheru in India, and studied morphologi­ . Six of its nine species are confined to cally. These include 698 accessions from the tropical and subtropical Africa (Phillips, 1972). Indian subcontinent and Sri Lanka, 648 acces­ One species occurs in , one sions from Africa, and 63 accessions of un­ species extends from Ethiopia into Arabia, and certain origin..Xwelve. quantitative and nine- one species is a pantropical weed but native to qualitative characters were recorded for each Africa and Asia. Finger millet, E . coracana (L.) accession. Data are derived from observations Gaertn., is cultivated in southern and eastern on at least ten in each accession. These Africa, and across most of southern Asia (Hilu observations were compared with field studies and de Wet, 1976a). The cereal is African in in Malawi (Africa) and the states of Andhra origin, was domesticated some 5,000 years B.P. Pradesh and Orissa in India, and with studies in eastern Africa, and introduced into India as of herbarium specimens filed at the Rijksher- a crop 3,000 years ago (Hilu, de Wet and Har- barium, Leiden (L) and Royal Botanic Gar­ lan, 1979). African and Indian cultivated com­ dens, Kew (K). Quantitative characters scored plexes Have been isolated from one another are days to flowering (DFL), height (PLH), until historical times. This separation provides number of basal tillers (NBT), number of in­ an opportunity to contrast racial evolution in florescences on the major culm (NIN), flag leaf Africa where the crop is sympatric with its wild length (FLL) and width (FLW), length of flag progenitor, with racial evolution in India where leaf sheath (LFS), peduncle length of terminal wild E . coracana subsp. africana (Kennedy- inflorescence (PDL), inflorescence length (INL), O’Byme) Hilu & de Wet is absent. length of longest inflorescence branch (LIB), width of major inflorescence branch (WIB), M a t e r i a l s a n d m e t h o d s — Germplasm col­ and number of inflorescence branches (NIB). lections of Eleusine coracana filed with the Accessions were classified into six major groups Genetic Resources Unit of the International based upon overall appearance, with the fol­ Crops Research Institute for the Semi-arid lowing sample sizes: Subsp. africana (wild and Tropics (ICRISAT) were planted near Pa- weed) N = 15; subsp. coracana, race coracana N = 206, race elongata N = 88, race plana N = 1 Received for publication 25 June 1983; revision ac­ 287, race compacta N = 202, race vulgaris N cepted 11 October 1983. = 635. Discriminant analysis (direct method, Research supported financially by a grant from the In­ Nie et al., 1975) was used to assess separation ternational Board for Plant Genetic Resources as part of of groups using the above characters. Quali­ its global interest in priority species of millets, and Grant AGR 58-319R-0-165 from the USDA Binational Pro­ tative characters studied are degree of plant grams (US-Spain). The use of facilities at ICRISAT is pigmentation,! growth habit, degree of glume appreciated. . prominence, degree of inflorescence compact-

5 5 0 April, 1984] de WET ET AL. — SYSTEMATICS OF ELEUSINE 551

ness, fruit color, degree of lodging at time of maturity, degree of senescence at time of ma­

turity, overall disease resistance, and yield po­ Os — O O tential. Data are filed as permanent descriptors C\ I t~- cs m so cm vi | Irt M —I-OO! | / — COI —I 00I —I —I for each of the Eleusine accessions in the gene — so ec CS in bank at ICRISAT. Specimens of a majority of P o ' S ' o ' ^ r~- o 'r?'£> m — collections studied are filed with the Crop Evo­ c*“> cs

Systematics—Finger millet is variable (Ta­ ble 1). This led to considerable controversy -?r about its origin and evolution. De Candolle fn — 0 0 o O' ItT (1886) and Cobley (1956) suggested that finger l — — — I — s© r - I I SO O I 1 I I O I millet was domesticated in India, and that its ,^_, ^r»1 — — 5-. ,-_, V , wild progenitor is the pantropical weed, E. in­ !/■> —< (S —>’ vO, C-i —Tf —' —_ ££ dica (L.) Gaertner. Porteres(1951,1958,1970) and Mehra (1963) proposed that this cereal Oao so'o Tf 3 ^ Oj ^ n o vo « ’t r^_ « « tivars. f s ^■OOOMMJiffi'OvovDvcO Phillips (1972) suggested that E. indica znd E. africana are conspecific, and recognized these taxa as subspecies of E. indica. Cytologically E. africana (2n = 36) and E. indica (In = 18) vn are distantly related, with little homology ob­ w is SO = 'S >/-> — © © O N 't vious between their basic genomes (Chenna- os l o t~- Os I P I U1NVOM---OS finger millet (Hilu and de Wet, 1976a). These —lOCSf-l 00 | |soo|cool©l 1—1 l—l — two taxa were combined, and E. africana was 'O — —— oe — — vo — <0 recognized as a subspecies of E. coracana by t--’ — ^ —'• v£_ r4_ —’ Hilu and de Wet (1976b). Stabilized, weedy m sovocvfn - o « co n s ■q:fnodrs'^i(Nr4rntsoo — *6 derivatives of crosses between wild and cul­ r -' — {<1 — — c s — — tivated E. coracana resemble either the cul­ tivated or wild complex in inflorescence mor­ phology. so r~>v-i 00 — so vi m © so 1. Eleusine coracana (L.) Gaertn. subsp. cor­ M O — — I — I I [——I acana Fruct. et Sem. 1: 8, t. 1. 1778. Type: soosi 1 ^ i m ^ rr I I — Illustration in Plukenet, Phytographia t. 91, f. O^fj-optJip^poeeov 0\ 5. 1691. coracan L., Syst. Nat. Ed. c^' O P~t ©_ O M m © © r^> 10, 2: 875. 1759. Cynosurus coracanus L., Sp. S so 00 oc ©' od 00 r^' —- so

cussa Fressen., Mus. Senck. 2:141.1837. Eleu­ ognition of five cultivated races with eleven sine coracana (L.) Gaertn. var. stricta (Roxb.) distinct cultivated complexes. Nees, FI. Afr. Austr. 1: 251. 1841. Eleusine luco Welw., in Apont. Phyto. Geog.: 591.1858, 2. Eleusine coracana subsp. africana (Ken- nom. nud. Eleusine dagussa Schimper in Re­ nedy-O’Byme) Hilu & de Wet, Econ. Bot. 30: gel, Garten Flora 21: 205. 1872. Eleusine cor­ 202.1976. Type-. South Africa, Cape Provinqe, acana (L.) Gaertn. vars. alba, atrajusca Koem., Kimberly district, Warrenton-on-Vaal, Wil- Handb. Getreid. 1: 329. 1885. Eleusine cor­ man H.K.1, March 1950 (K, Holotype). Eleu­ acana (L.) Gaertn. var. tocussa (Fresen.) sine africana Kennedy-O’Byme, Kew Bull. 12: Franch., Bull. Soc. Hist. Nat. Autun 8: 377. 65. 1957. subsp. africana 1895. Eleusine indica var. stricta (Roxb.) Chiov. (Kennedy-O’Bvme) S.M. Phillips, Kew Bull. in Nuov. Giora. Bot. Ital. 26: 83. 1919. Eleu­ 27: 259. 1972.' sine stricta (Roxb.) Chiov., vars. rufoabbre- viata, fuscoabbreviata, alboabbreviata, rufoe- Wild finger millet is a tufted annual, with longata, fuscoelongata, alboelongata, Cif., in slender and geniculately ascending culms that Atti 1st. Bot. Univ. Pavia ser. 5, 2: 172. 1944. branch at the lower nodes. Flowering culms Eleusine pilosa Gilli, Ann. Naturhist. Mus. are up to 135 cm tall, with leaf-blades up to Wien 69: 50. 1965. 36 cm long and 10 mm wide. Inflorescences are digitate, composed of 3-13 (rarely more) Subspecies coracana includes all cultivated slender and ascending branches, with one or a finger millets. Plants are annual, tufted, erect few branches often arranged some distance be­ or with geniculately ascending culms that are low the digitate cluster (Fig. 1). Inflorescence up to 165 cm high, and sometimes root from branches are 8-17 cm long and rarely more the lower nodes. Culms are commonly than 5 mm wide, with the spikelets arranged branched from, the upper nodes to produce sec­ in two rows on one side of the rachis. Spikelets ondary inflorescences. Leaf-blades are linear are 4-9 flowered and 5-8 mm long. Glumes to linear-lanceolate, up to 70 cm long and 20 are shorter than the spikelet, lanceolate-oblong mm wide. Inflorescences are digitate, often with in profile, rarely over 5 mm long and narrowly one or more racemes some distance below the winged along the keel. Lemmas are lanceolate main cluster of 4-19 branches. Inflorescence in profile and up to 6 mm long. The palea is branches are slender to robust, up to 24 cm distinctly winged along the keels. Grains are long, reflexed when slender, or incurved at the 1.0-1.8 mm long, elliptic in outline and the tip when robust, sometimes with secondary surface is shallowly ridged. branches. Spikelets are 6-9 flowered and 6-10 Eleusine coracana is predominantly self-fer­ mm long, overlapping and mostly arranged in tilized. Subspecies africana, however, does two rows along one side o f the rachis. Glumes cross occasionally with subsp. coracana to pro­ are unequal and shorter than the spikelet. The duce fully fertile hybrids. Derivatives of such grain is white, red, brown or black; up to 2 mm crosses are aggressive colonizers. Hilu and de long, more or less globose, with the surface Wet (1976a) recognized two weedy races of finely striated. finger millet. The common weed differs from Inflorescence shape is variable. The digi- its wild relative primarily in occupying habitats tately arranged branches may spread out and that are continuously being disturbed by man. become reflexed, or they may be erect and in­ This weed extends across the range of finger curved, often forming a fist-like structure. In­ millet cultivation in Africa. The other weed florescence shape, correlated with geographic race is characterized by a well developed ter­ distribution, allowed Hilu and de Wet (1976b) minal raceme ofloosely arranged spikelets with to recognize three races of cultivated finger mil­ as many as 15 subdigitately arranged branches let. Hussaini, Goodman and Timothy (1977) below it (Fig. 2). This weed occurs in cultivated extended this study to a larger collection and fields and has been collected around Dedza and distinguished twelve groups on the basis of Kota Kota in Malawi, and Iringa in Tanzania. principal component or canonical variate anal­ Plants that mimic cultivated finger millet in yses of a number of inflorescence and vege­ vegetative as well as inflorescence morphology, tative traits. Indian material showed clinal but with spikelets that disarticulate at maturity variation, with southern and eastern cultivars also occur as weeds in Malawi. Most of these forming phenotypic extremes. Cultivars from are first generation hybrids but some represent Ethiopia and Uganda appeared morphologi­ hybrid derivatives of such crosses. cally distinct from those of the rest of Africa The weedy E. indica and E. coracana subsp. or India. Our comparative morphological study africana are -widely sympatric in Africa, with of the ICRISAT collections resulted in the rec­ E. indica extending to Asia. In Africa, E. indica April 1984] e® w et et a l.—system atics of eleusine

Fig 1-6 Variation within Eleusine coracana. x 0.32. 1-2. Subsp. africana. 1. W ild i

I.E. 2312 (CEL). 554 AMERICAN JOURNAL OF BOTANY [Vol. 71 occurs primarily along the eastern and western cence a lily-like appearance (Fig. 11). Others coastal plains while subsp. africana is primar­ have short but spreading branches that are ily found along the eastern and southern high­ twisted to form a starfish-shaped inflorescence lands. Subspecies africana is commonly more (Fig. 12). The most widely distributed finger robust than E. indica, with the grain shallowly millets of this group are characterized by in­ ridged rather than obliquely striate (Phillips, florescence branches that curve inward 16 form 1972). a loosely to tightly clenched fist-like structure (Fig. 13). Racial evolution in finger m illet—-In­ Comparative morphology, coupled with dis­ florescence morphology is associated with grain criminant function analysis of quantitative yield, and is used by farmers to distinguish characteristics suggests five taxonomic groups complexes of cultivars (Fig. 3-13). The most of cultivated finger millet (Fig. 14). A relatively primitive cultivars are characterized by inflo­ low correspondence of 67% between a classi­ rescences with spreading branches that are fication of plants based upon qualitative inflo­ straight or slightly incurved at the tip when rescence morphology, and the results of the mature. Some of these cultivars resemble ro­ discriminant analysis classification routine, re­ bust specimens of subsp. africana, except that flects racial intergradation in all quantitative they lack the ability of natural seed dispersal characters studied (Table 1). Variation within (Fig. 3). These cultivars are common in Africa, each of the five groups is essentially continu­ but are also grown in southern and eastern ous. The first two discriminant functions ac­ India. This complex includes the African and count for 94% of the. variation. Characters con­ Indian highland races recognized by Hilu and tributing most to group separation on de Wet (1976b). Indian and African cultivars discriminant function 1 are length of the long­ can not consistently be separated on the basis est inflorescence branch, inflorescence length, of inflorescence morphology. A second group and days to flowering, with standardized ca­ of African and Highland Indian cultivars is nonical discriminant fiinction coefficient scores characterized by spreading inflorescence of 0.66, 0.33, and 0.24, respectively. Charac­ branches. Cultivars with 10-18 cm long ters most important in separating groups on branches are grown in eastern Africa and in discriminant-function 2 are - flag leaf width southern and eastern India (Fig. 4), while those (—0.62), days to flowering (—0.40) and number with branches up to 24 cm long are typically of major culm branches (0.36). The five races East African (Fig. 5). African collections have recognized are coracana (Fig. 3), elongata (Fig. more slender racemes than those of Indian cul­ 4-6), plana (Fig. 7-9), compacta (Fig. 10), and tivars, allowing the branches to become re­ vulgaris (Fig. 11-13). Inflorescence types with­ flexed. at maturity. Individuals in fields of these in races grade into one another so completely cultivars sometimes have spikelets arranged in that the recognition of subraces becomes im­ clusters along the rachis (Fig. 6). A distinct possible. j»roup_o£cuLtiYarsis.grownirom Ethiopia south to Zambia. Typical representatives of this Race coracana—Inflorescence morphology complex are characterized by spreading inflo­ in race coracana is similar to that of wild finger rescence branches with large spikelets arranged millet. Race coracana probably gave rise to the in two even rows along one side of the rachis other four races through selection under do­ (Fig. 7). The spikelets in these cultivars are mestication (Fig. 14). Race coracana resembles often over 10 mm long with the glumes nar­ wild subsp. africana in having 5-19 slender rowly lanceolate (Fig. 8). Similar to this group inflorescence branches that are 6-11 cm long, are cultivars from Africa and from southern digitately arranged, ascending, and with the and eastern India with large spikelets that axe tips straight, slightly incurved or somewhat re- irregularly arranged, often almost surrounding flexed at time of maturity (Fig. 3). Some ge­ the rachis (Fig. 9). The most advanced cultivars notypes differ from wild finger millet primarily are characterized by highly proliferated inflo­ in being unable to disperse their spikelets with­ rescence branches that are clumped together out the help of man. They often resemble weedy to form a large fist-like structure (Fig. 10). These genotypes (Fig. 2) in having a well developed cultivars have been collected in Uganda, Sri central inflorescence branch. Race coracana is Lanka, southern India, and Nepal, but are widely, but sporadically grown across the range probably more widely grown. The most com­ of finger millet cultivation in Africa and India. monly grown finger millets in Africa as well as It is particularly well adapted to agriculture in India are characterized by relatively small in­ the East African highlands, and the Western florescences. Some cultivars have short Ghats of India. Higher yielding cultivars have branches that curve down giving the inflores­ the primary inflorescence branches divided near April, 1984] PE WET ET AL. — SYSTEMATICS OF ELEUSINE 555

Fig. 7-13. Variation within Eleusine coracana subsp. coracana. *0.32. 7-9. Race plana. 7. Malawi, Choweaten, I.E. 2622 (CEL). 8. Malawi, Nchisa, I.E. 2644 (CEL). 9. Malawi, Chinguluwe, I.E. 2673 (CEL). 10. Race compacia. India, Tamil Nadu, CEL 4437.11-13. Race vulgaris. 11. India, Uttar Pradesh, CEL 4165.12. India, Maharashtra, I.E. 2289 (CEL). 13. Malawi. Mastade, I.E. 2622 (CEL).

the base. Some cultivars are drought tolerant Eastern Ghats of India. These cultivars are often and compete aggressively with weeds or other difficult to distinguish from robust specimens minor cereals under conditions of traditional of race coracana, except that the inflorescence agriculture. In India it is often sown as a sec­ branches are longer and always reflexed at ma­ ondary crop in fields of Pennisetum ameri- turity. Indian and African cultivars cannot canum (L.) Leefce (pearl millet) or Sorghum consistently be separated on the basis of mor­ bicolor (L.) Moench (sorghum). phology. They may have evolved indepen­ dently in India and Africa under similar en­ Race elongata—This race is morphologically vironmental conditions from race coracana. the most distinct of the five races of finger Since finger millet is an African crop, however, millet. It is characterized by long slender in­ it seems more likely that the Indian genotypes florescence branches that are 10-24 cm long, were introduced from Africa. Individuals with digitately arranged, spreading and curved out­ spikelets arranged in clusters along the rachis ward at time of maturity. A cultivax grown only occur within fields of race elongata across its in Malawi and adjacent Zambia has 15-24 cm range of cultivation in Africa. Branches are long inflorescence branches (Fig. 5) and is lo­ commonly reflexed giving the inflorescence a cally referred to as “elephant foot” finger mil­ distinctive appearance (Fig. 6). let. The more common cultivars have 10-18 cm long inflorescence branches (Fig. 4) and are Race plana—This race is primarily African grown along the east African highlands and the in distribution, is grown in Ethiopia and Ugan- 556 AMERICAN JOURNAL OF BOTANY [Vol. 71

AFRICANA Race vulgaris—This is the common finger WILD & WEED millet of Africa and Asia. It is grown from 4 - • southern Africa to Ethiopia and Uganda, and from India to Indonesia. It is variable in size and growth habit. Three cultivar complexes are recognized on the basis of inflorescence mor­ phology. Members of these complexes are fre­ quently grown together in the same field, but some fanners maintain them as distinct. In­ florescence branches are reflexed (Fig. 11), twisted (Fig. 12) or incurved (Fig. 13), with all possible intermediates occurring where these complexes are grown together in the same field. In India, race vulgaris frequently follows irri­ gated rice (Oryza sativa) as a dry season crop. Some cultivars are drought tolerant while oth­ ers are well adapted to areas of high rainfall. *COMPACTA In the Eastern Ghats of India some genotypes are sown in nursery beds and transplanted to - 1 0 1 2 3 4 fields with the first rains of the season, to m a­ ture about 45 days later. Grains are cooked as OF 1 rice, or ground into flour to make a porridge Fig. 14. Separation of group centroids for six Eleusine or unleavened bread. Stems and leaves are taxaon the first two functions from Discriminant Analysis. commonly fed to livestock. Arrows indicate putative origins of cultivated races. Dom estication of finger m illet— The species was domesticated in Africa. Its closest wild relative, Eleusine coracana subsp. afri­ da, and to some extent in the Eastern and West­ cana is native_to Africa. It is common along ern Ghats of India. It is characterized by large the highlands of East Africa and the grasslands spikelets (8.0-15.0 mm long) that are arranged of the southern African plateau, and extends in two, more or less even rows along the rachis, into the coastal plains of East Africa and the giving the inflorescence branch a flat ribbon- forests of West Africa (Phillips, 1972). It was like appearance (Fig. 7). Some cultivars are introduced to tropical and.subtropical Austra­ characterized by spikelets 10.0-15.0 mm rath­ lia, America, and South Asia as a weed. Sub­ er than 8.0-11.0 mm long (Fig. 8), and in others species africana crosses with subsp. coracana the fertile florets are so numerous that they (finger millet) to produce fertile hybrids. It rep­ almost surround the rachis at maturity (Fig. resents the progenitor of the cultivated finger 9). This latter group resembles race compacta, millet, and is still harvested as a wild cereal in except that the inflorescence branches are the African savanna during times of drought. straight or reflexed, rather than incurved. Finger millet is cultivated on the highlands of East Africa and the plateau of southern Af­ Race compacta—Members of this race are rica. It is not adapted to the tropical forests of commonly referred to as cockscomb finger mil- Zaire or West Africa, and rarely grown in the lets in both Africa and India. Spikelets are com­ savanna west of Chad. In the West African posed of nine or more florets, with the inflo­ savanna, finger millet is replaced as a minor rescence branches divided at the base, ascending cereal by the fonios, exilis (Kippist) and incurved at the tip to form a large fist-like Stapf and D. iburua Stapf (Stapf, 1915; Por- inflorescence (Fig. 10). Indian cultivars almost teres, 1955, 1976). It was probably domesti­ always have an inflorescence branch located cated in an area extending from western Ugan­ some distance below the 4-14 branches in the da to the highlands of Ethiopia (Harlan, 1971). terminal cluster on the primary axis. African Wild finger millet is particularly abundant in cultivars often lack this lower inflorescence this region and domesticated finger millet is branch. Race compacta is grown in Ethiopia grown extensively there. and Uganda, and in northeastern India. It re­ The antiquity of cereal cultivation in Africa sembles cultivars of race vulgaris with fist-like south of the Sahara is not known. Harlan, de inflorescences, but inflorescences of compacta Wet and Stemler (1976) suggested that do­ are larger, and the lower raceme is rarely di­ mestication of hative African food plants start­ vided in vulgaris. ed at the beginning of the present dry period April, 1984] DE WET ET AL. — SYSTEMATICS OF ELEUSINE 557 of the Sahara, some 5,000 years ago. Hilu et H a r l a n , J. R. 1971. Agricultural origins: centers and al. (1979) indicated that this cereal could have noncenters. Science 174: 468-474. been grown in Ethiopia at that time. Archae­ ------, J. M . J. d e W e t , a n d A. B. L. S t e m l e r . 1976. Plant domestication and indigenous African agricul­ ological material excavated at Axum resembles ture. In J. R. Harlan, J. M. J. de Wet, and A. B. L. race plana which is the principal finger millet Stemler teds.], Origins of African plant domestication, still grown in Ethiopia. If this archaeological pp. 3-19. Mouton Press, The Hague. material dates from 3,000 B.C., as suggested H il u , K. W., a n d J. M. J. d e Wet. 1976a. Domestication by Hilu et al. (1979), the cereal must be sub­ o f Eleusine coracana. Econ. Bot. 30: 199-208. stantially older as a crop in Africa, since plana ------, a n d ------. 1976b. Racial evolution in Eleusine coracana ssp. coracana (Finger millet). Amer. J. Bot. is a highly evolved race. The cereal reached 63: 1311-1318. India during the first millenium B.C. (Vishnu------, ------, a n d J. R. H a r l a n . 1979. Archaeobo- Mittre, 1968). It became widely distributed in tanical studies of Eleusine coracana ssp. coracana southern Africa during the expansion of iron (Finger millet). Amer. J. Bot 66: 330-333. working technology (Summers, 1958). Sugges­ H u s s a in i, S. H ., M. M. G o o d m a n , a n d D . H . T im o t h y . tions of Davies and Gordon-Gray (1977) that 1977. Multivariate analysis and the geographical dis­ tribution o f the world collection o f finger millet. Crop finger millet, sorghum and pearl millet were Sci. 17: 257-263. cultivated in South Africa by the “latter part J a m e s o n , J . D. 1970. Agriculture in Uganda. Oxford of the third millenium B.C.” are probably not University Press, Oxford. correct (Oliver Davies, pers. comm.). A more K e n n e d y -O ’B y r n e , S. 1957. Notes on African grasses. likely date for this cereal assemblage in south­ XXIX. A new species of Eleusine from Tropical and ern Africa is early iron age, some 800 years South America. Kew Bull. 11: 65-72. M e h r a , K. L. 1963. Differentiation of the cultivated and ago. wild Eleusine species. Phyton (Argentina) 20: 189— Finger millets in Africa and India are similar 198. in adaptation and morphology. This is not sur­ N ie , N. H., C. H . H u l l , J . G. J e n k in s , K. S teinbrenner , prising. African and Indian cultivars of finger a n d D. H . B e n t . 1975. Statistical package for the millet are grown in similar habitats. Finger social sciences. 2nd ed. McGraw-Hill, New York. millets grown in the Ghats of India and high­ P h il l ip s , S. M. 1972. A survey of the genus Eleusine in Africa. Kew Bull. 27: 251-270. lands of East Africa, or the coastal plains of P o r t e r e s , R. 1951. Eleusine coracana Gaertn. Cereale East Africa and tropical South India belong to des humanites des pays tropicaux. Bull. Inst. Fr. Afr. the same races, respectively. Similarly, guinea Naire 12: 1-78. sorghums grown in high rainfall areas of the ------. 1955. Les cereales mineurs du genre Digitaria Eastern Ghats of India resemble those widely en Afrique et . J. Agric. Trop. Bot. Appl. 2: cultivated in the fog belt of the East African 349-675. ------. 1958. Lemillet£7ewsmede-rindeetdel’Afrique highlands (de Wet, 1978). Racial evolution in Oriental (E. coracana). J. Agric. Trop. Bot. Appl. 5: finger millet occurred in Africa before this ce­ 463-486. real was introduced into India. — :— . 1970. Primary cradles of agriculture in the Af­ rican continent. In J. D. Fage and R. A. Oliver [eds.l, Papers in African pre-history, pp. 43-58. Cambridge LITERATURE CITED University Press, England. — r— . 1976. African cereals: Eleusine, Fonio, Black C hennaveeraiah , M . S ., a n d S. C . H jr e m a t h . 1 9 74. Fonio, Teff. Brachiaria, Paspalum, Pennisetum, and Genome analysis of Eleusine coracana (L.) Gaertner. African rice. In J. R. Harlan, J. M. J. de Wet, and A. Euphytica 23: 489—495. B. L. Stemler [eds.], Origins of African plant domes­ C o b l e y , L. S. 1956. An introduction to the botany of tication, pp. 409-452. Mouton Press, The Hague. tropical crops. Longmans, Green and Co., London. S t a f f , O. 1915. Iburu and Fundi, two cereals of Upper D a v ie s , O., a n d K . G o r d o n -G r a y . 1977. Tropical Af­ Guinea. Kew Bull. 1915: 381-386. rican cultigens from Shangweni excavations, Natal. J. S u m m e r s , R. 1958. Inyanga. Cambridge University Press, Archaeol. Sci. 4: 153-162. England. D e C a n d o l l e , A. 1886. Origin ofcultivatedpIants(1967 V a v il o v , N. I. 1951. The origin, variation, immunity reprint). Hafher, New York. and breeding of cultivated plants. Chronica Bot. 13: d e W e t , J. M. J. 1978.' Systematics and evolution of 1-366. Sorghum sect. Sorghum (Gramineae). Amer. J. Bot. V is h n u -M it t r e . 1968. Prehistoric records of agriculture 65: 477-484. in India. Trans. Bose Res. Inst, 31: 87-106.