Origin and Dispersal of Common Millet and Foxtail Millet*

By SADAO SAKAMOTO

Plant Germ-plasm Institute, Faculty of Agriculture, Kyoto University (Mozume, Muko, Kyoto, 617 Japan)

In Eurasia, both winter and summer cereal with nine bivalents at the MI of the PMCs4>. This crops have been cultivated over a wide range of indicates that these two species are biologically the environmental conditions. Winter cereals, such as same species and S. viridis is the probable ancestor wheat and barley, were domesticated in the Middle of domesticated foxtail millet. However, because East about 10,000 years ago. Rice is a representa­ of wide occurrence of S. vin'dis in Eurasia, it is tive of summer cereals originating in southern almost impossible to localize the place of domesti­ Asia, and it has been grown widely in Asian coun­ cation of foxtail millet mainly on the geographical tries. In addition to rice, several kinds of millets distribution of wild ancestral species. have been domesticated in Eurasia, some in east or central Asia and others in the Indian subcontinent. Among them, common millet (Panicum miliaceum Current theories on the geogra­ L.) and foxtail millet (Se/aria italica (L.) P. Beauv.) phical origin of common millet are thought to be the most anciently domesticated and foxtail millet cereals in Eurasia. In the present paper, a new 14 theory on the origin and dispersal of those two Vavilov > assumed that common millet and fox­ millets based on the results of recent field research tail millet originated in East Asia based on the fol­ on the two millets in southwestern Eurasia and lowing two lines of botanical evidence: (1) sharp those of the experimental works on landraces of increase in diversity in common millet toward the millets collected from the vast areas of Eurasia East Asia and an exceptional diversity of biological is discussed. characters of this millet in Mongolia and Manchu­ ria, an area of Chinese civilization, and (2) the Botanical origin of common principle center of the diversity of foxtail millet is millet and foxtail millet eastern Asia including China and Japan, where it is cultivated as a cereal crop for food. Vavilov's The botanical origin of common millet is still theory has been supported by the findings of unknown. The probable progenitor of this millet archaeological remains of these two millets from in the genus Panicum based on the genetical stud­ the Neolithic Yan-shao villages of China, which ies has not been found yet. On the other hand, were established about 4,000 BC. green foxtail (S. viridis (L.) P. Beauv.) is presumed Quite interestingly, the earliest palaeoethnobo­ a wild ancestor of foxtail millet, S. italica. The two tanical finds of common millet in the West came species are diploids with the same chromosome from the Neolithic sites of central Europe. Foxtail number (2n=18). The F, hybrid between them is millet has also been found in some Neolithic sites fertile and chromosome pairing in F, is normal in Europe, and this millet was chiefly cultivated in 7 the Bronze Age in alpine Europe ' . Carbonized * Contribution no. 46 from the Plant Germ-plasm Insti­ grains of common millet have also been reported tute, Faculty of Agriculture, Kyoto University, Kyoto, from the aceramic levels at Argissa-Maghilla in Japan. Greece (5,000-6,000 BC) and from the Jemdet Nasr 85

9 period in Mesopotamia dating around 3,000 BC i. form are very similar to the cultivated common The palaeoethnobotanical data obtained so far evi­ millet except for the shorter stature, more sparsely dently indicate that common millet and foxtail opened panicles, fewer spikelets per panicle, millet have been the staple food for the people smaller floral glumes and brittle spikelets. The since a very early period of agricultural develop­ average heading date of this weed grown in a 2 ment in Eurasia. Therefore, recently, Harlan > glasshouse was 39 days, which was very similar to suggested the possibility of independent domestica­ those of cultivated Romanian forms. tion of these two millets at least in China and Based on the resemblance of this form to the cul­ Europe. He discussed three different possibilities tivated one, it was classified as P. miliaceum sub­ on the domestication and dispersal of these two species ruderale (Kitag.) Tzelev 13>, which was found crops considering the archaeological findings from in Manchuria and originall y named as P. milia­ 5 8 Neolithic sites in China and Europe; (1) common ceum var. ruderale Kitagawa >. Popova > reported a millet and foxtail millet were domesticated in very similar readily shattering form of common China and dispersed to Europe before 4,000 BC, (2) millet from the Central-Boharian Oasis. This form they were domesticated in the West and dispersed is distributed not only in Manchuria but also in 13 to China before Yang·shao times, and (3) there was Siberia and Central Asia '. A similar plant is des­ more than one domestication. He believed that the cribed in the European weed flora. independent domestication was the most likely The origin of common millet is still unknown. A answer because no other known crop has such a more detailed study of this weed form in Romania distribution in that time range. He mentioned may provide some ideas on the botanical origin of further that in our present ignorance to these two this crop, although it is often difficult to distin­ millets, independent domestication appears to be guish the wild form from the weedy race which the most likely but new information would easily sometimes might possibly be derived from an lead to other conclusions. escaped cultivated form. However. judging from the wide occurrence of this weed form, from Man­ Our recent research results churia to East Europe through Central Asia, detailed genetic analysis of this form with the cul­ In 1978, 1979, 1980 and 1982, we made four tivated form is expected to provide some positive botanical expeditions to Afghanistan, Turkey, ideas on the botanical origin of common millet that Greece, Romania, France and Spain. During these has been completely ignored and unexplored. field research works, we collected many samples of common millet and foxtail millet together with 2) Traditional utilization methods information on traditional cultivation methods and If common millet and foxtail millet have been utilization for food. At the same time, during cultivated extensively in wide areas of Eurasia these years, we collected many different Jandraces since ancient times, there must be the traditional of these two millets from the vast areas of Eurasia. way of utilization for those two millets. During Using these materials we carried out an analysis our field research work, we gathered information of genetic variations and their geographical on this subject from local people in each region. distribution. Table 1 gives a summary of the results. The infor­ mation on China and Caucasia was obtained from 1) Finding of a weed form of common millet the literature. During our field research in 1980, we collected a Generally speaking, the utilization methods of weed form of common millet along the road located cereal grains are divided into two groups, one for in the western suburbs of Iasi, Moldavia Province foods and the other for drinks. The former is of Romania 11 >. This form occurred abundantly further divided into three categories, grain, meal forming a dense belt between the road and the and flour, and the latter non-alcoholic and alco­ maize field and partly invaded into the cultivation holic drinks. For preparation of food from the fields. The morphological characteristics of this grains of common millet and foxtail millet, seven 86 JARQ Vol. 21, No. 2, 1987

Table 1. Foods and drinks made from the grains of common millet and foxtail millet in Eurasia

Type of cooking Region Grain Meal Flour Drink Boiled . Dumpling Flour Non- . grain Gruel Mochi Porndge (dango) porridge Bread alcoholic Alcoholic Japan { non -waxy + + waxy + + + + Korea l non-waxy + waxy + + + Ch.ina { non-waxy + + + + waxy + + + . { non-waxy + Taiwan waxy + + + + Batan Islands + Halmahera Islands + India + + + + Afghanistan + + Caucasia + + Turkey + Bulgaria + + + Romania + + Italy + France + subcategories have been recorded, namely boiled To shed light on this problem, comparative stu· grain, gruel, mochi (cake made from waxy starch), dies of common millet using 43 samples from Af. porridge and dumplings (dango), flour porridge and ghanistan (15 samples). Romania (8), Turkey (2), bread. It is interesting to note, as seen in Table 1, Greece (2), Bulgaria (1), Czekoslovakia (2), Central that in East Asia preparation of boiled grain, gruel, Asia (7) and Japan (6), and 70 samples of foxtail mochi and alcoholic drinks is popular, while in the millet from Afghanistan (25), Europe (8), Central area from Southeast Asia and India to Europe, por­ Asia (7), India (4), Taiwan (2), Korea (6) and Japan ridge, flour porridge, bread and non-alcoholic (18), were carried out in Kyoto in 1981 10>. drinks are prevailing. All European and Central Asiatic strains of com· mon millet were very early varieties, while the 3) Morphological variation and its distri­ Afghan strains include both intermediate and late bution varieties. The Afghan strains showed a wide vari· No comparative studies of the characteristics of ation in plant height, the number of tillers and the these two millets covering all of Eurasia have been length of panicle in addition to the number of days made so far. There is only the taxonomic study on from sowing to heading and the number of leaves. common millet from European parts of USSR, Romanian strains were intermediate between Cen· Caucasia, Central Asia, Siberia, Mongolia and tral Asiatic and other European strains. Manchuria made by Lyssov6>. He lists 78 varieties Field observation of foxtail millet in Afghanistan classified into five subspecies. Thirteen varieties indicated that they are characterized by early of foxtail millet were reported in Georgia, western heading and short plant height. According to our '.Transcaucasia 1>. The collections of millets in Hin­ experiment, the Afghan strains were clearly char· dukush made by the German Hindukush Expedi· acterized by v':!ry early heading as were those of 2 tion in 1937 and 1938 were studied by Scheibe' '. European and Central Asiatic strains, which were He classified 103 collections of foxtail millet into quite different from the East Asiatic ones used for 15 varieties belonging to two subspecies. He also comparison. Another remarkable characteristic of classified 118 samples of common millet into seven the Afghan strains was the large number of tillers varieties. from 7 to 41. These characteristics are very sim· 87 ilar with those observed in S. viridis, a probable AC, BC and X by means of intraspecific hybrid ancestor of this millet. This indicates that the partial pollen sterility. Namely, the strains of type Afghan foxtail millet has conserved rather primi­ A, B, C were those which produced F1 hybrids hav­ tive features. Furthermore, the present compara· ing normal pollen fertility when crossed with test· tive studies suggest that there is a distinct er A, B, and C, respectively. When both F1 difference between the strains of the West and hybrids from the crosses with two testers, A and those of the East, Afghanistan and its vicinity C, or B and C, showed normal pollen fertility, the being a distributional boundary. strain was classified as type AC or BC. The strains whose F1 hybrids always showed pollen 4) Geographical distribution of landrace fertility of less than 75% in all three cross combina· groups of foxtail millet classified by hybrid tions were designated as type X. pollen sterility Most strains of type A are distributed in Japan, Through the dispersal process of a given crop Korea and China. The strains of S. italica in these from the place of its domestication, genetic differ· regions are thought to be closely related with one entiation adapted to various ecological conditions another and are clustered as a single landrace and agricultural practices occurs, and this process group. Another phylogenetic group might be dis­ produces specific landraces of the crop. A well­ tributed in south China because the strains from known example is Japonica, Indica and Javanica Hunan and Yunnan Provinces were not type A. type differentiation in cultivated Asian rice. Sim­ The strains of type B are narrowly distributed in ilar genetic differentiation might be expected in the mountainous areas of Taiwan and in the common millet and foxtail millet but there have southwestern part of the Nansei Islands (Okinawa been no experimental studies. Prefecture) of Japan in the neighborhood of Tai­ To clarify the genetic differentiation among local wan. This strongly suggests that the strains of landraces of foxtail millet, we produced 223 combi­ the southwestern part of the Nansei Islands were nations of intraspecific hybrids using 83 strains introduced from Taiwan. Type C is found espe­ collected from various parts of Eurasia. These cially in most European strains. This indicates strains were crossed with three tester strains, one that the European strains also form a distinct land· (tester A) from Japan, one (B) from Taiwan and race group. Type AC strains are distributed in one (C) from Europe3>. Among the 83 strains of Afghanistan and type BC in India. These types foxtail millet used in this experiment, 62 strains are thought to be less specialized genetically than successfully produced F1 hybrids with these tes­ type A, B or C. Type X might include more than ters. They could be classified into types A, B, C, one type, because different kinds of pollen fertility

USSR 8 ------Afghanistan Type AC~ Burma

"T~~~~uBC)

Indonesia

Fig. 1. Geographical distribution of six landrace groups of fox tail millet 31 Data based on Kawase & Sakamoto • 88 JARQ Vol. 21, No. 2, 1987 arrays in the three cross combinations are recog· Therefore, this form is assumed to be the progeni· nized among the strains classified as type X. The tor of common millet. type X strains from Lan Hsu Island of Taiwan and (5) Foxtail millet collected from Afghanistan was the Batan Islands of the Philippines might belong characterized by short plant height with quite a to a single landrace group, because these islands few tillers and consequently with many small pan­ are located in a close vicinity. Some strains of type icles as observed in green foxtail, S. viridis, a prob­ X might have been originated in different ways, able ancestor of this millet. T his indicates that since they are sporadically distributed in different Afghan foxtail millet has conserved rather primi· regions. To elucidate these problems, further ti ve features. cross experiments are needed. (6) From the analysis of partial pollen sterili ty Partial hybrid pollen sterility occurs as the observed in intraspecific hybrids of foxtail millet, result of genetic differentiation between strains. genetically less specialized landraces, i. e. type AC Therefore, these types are thought to reflect the and type BC, were found in those from Afghanis· phylogenetically differentiated landrace groups. tan and India. Each of these landrace groups showed a rather Based on these findings, a new theory on the limited geographical distribution, and was often geographical origin of common millet and foxtail found endemic to small areas as shown in Fig. l. millet is proposed: Common millet and foxtail The indigenousness of these six landrace groups millet were first domesticated within the area differentiating with one another implies a long his­ ranging from Central Asia and Afghanistan to tory of cultivation of foxtail millet in each region. India, and from there they were dispersed both The differentiation of a landrace group in foxtail westward to Europe and eastward to East Asia millet is thought to have occurred along with the being gradually differentiated genetically. They dispersal from the place of domestication as is the have expanded their cultivation southeastward to case for Asian rice. continental parts of Southeast Asia and finally to the Pacific islands of Southeast Asia through the A new theory on the geographi­ Indian subcontinent. The archaeological evidence of the Neolithic occurrence of those two millets cal origin of common millet and both in North China and Europe is in good agree­ foxtail millet ment with this theory. As mentioned before, the possibility of Chinese The foregoing discussion is summarized as origin of those two millets has been supported by follows: the archaeological remains from Neolithic Yan· (1) Archaeological remains of common millet and shao sites. The comparative studies made under foxtail millet were found in Neolithic sites not only experimental conditions in the present investiga· in East Asia and Europe but also in the central tion also demonstrated a large diversity in East part of Asia and Middle East. This indicates that Asian strains of both millets concerning some these two millets are the most anciently domesti· quantitative as well as qualitative characters. cated cereals in Eurasia, and that they have been Some authors have pointed out that the center of cultivated throughout most of Eurasia. diversity is not necessarily the center of origin. (2) Traditional utilization of those two millets can The large morphological diversity in East Asia as 14 still be found in various parts of Eurasia. observed by Vavilov > could have resulted merely (3) The diversity of botanical characters is from the intensive cultivation of the landraces of observed not only in East Asia but also in Central both millets for a long time. Therefore, East Asia Asia and its vicinities. might be a secondary center of diversity but not (4) The wide distribution of a weed form closely the place of domestication. related to cultivated common millet from North China to East Europe through Central Asia was evident. 89

9) Renfrew, J. M.: The archaeological evidence for the References domestication of plants: methods and problems. In The domestication and exploitation of plants and 1) Dekaprelevich. L. L. & Kasparian, A.: A contribu­ animals. ed. Ucko, P. J. & Dimbleby, G. W., Aldine tion to the study of foxtail millet (Setan·a italica P. 8. Pub. Co., Chicago, 149-172 (1969). maxima Alef.) cultivated in Georgia (western Trans­ 10) Sakamoto, S.: Origin and phylogenetic differentia­ caucasia). Bull. Af>P/. Bet. Genet. P/.-Breed., 19, 533- tion of cereals in southwestern Eurasia. In Domes­ 572 (1928). ticated plants and animals of the southwest Eur· 2) Harlan. J. R.: Crops and Man. Amer. Soc. Agro. & asian agro-pastoral culture complex. I. Cereals. ed. Crop Sci. Soc. Amer.. Madison (1975). Sakamoto. S .. Kyoto Univ., Kyoto, 1- 45 (1986). 3) Kawase. M. & Sakamoto, S.: Geographical distribu­ 11) Sakamoto, S. & Kobayashi, H.: Variation and geo­ tion of landrace groups classified by hybrid pollen graphical distribution of cultivated plants. their sterility in fox tail millet, Setaria italica (L.) P. Beauv. wild relatives and weeds native to Turkey, Greece Jpn. / . Breed., 37, 1-9 (1987). and Romania. In Preliminary report of comparative 4) Kihara. H. & Kishimoto, E.: Bastarde zwischen Se/a­ studies on the agrico-pastoral peoples in southwest­ ria italica und S. viridis. Bot. Mag. Tokyo, 56, 62-67 ern Eurasia, II. 1980. ed. Tani, Y., Kyoto Univ., (1942) [ln Japanese with German summary]. Kyoto, 41-104 (1982). 5) Kitagawa, M.: Contributio as cognitionem florae Man­ 12) Scheibe, A.: Die Hirsen im Hindukush; Ein Beitrag shuricae. X. Bot. Mag. Tokyo, 5 1 , 150- 157 (1937). zur Kenntnis von Kulturpflanzen in geographischer 6) Lyssov, V. N.: Proso-Panicum L. /11 Groat crops RUckzugs-position. Zeitschri/t f ~anzenztichl., 25, (buckwheat, millet, rice). ed., Krotov, A. S., Kolos, 392-436 (1943) [In German). Leningrad, 119- 236 (1975) [In Russian). 13) Tsvelev, N. N.: Grasses of the Soviet Union. Oxo­ 7) Netolitzky, F.: Die Hirse aus antiken Funden. Sit­ nian Press PVT, Ltd., New Delhi (1983). zimgsber. Kais/. Akad. Wiss. Wien, 123, 725-750 14) Vavilov, N. I.: Studies on the origin of cultivated (1914) [In German]. plants. Inst. Bot. Appl. Am el. Plant., Leningrad 8) Popova, G. M.: Essay on the cultivated vegetation of (1926). the Central Boharian oasis. Bull. Af>PI. Bet. Genet. P/. -Breed., 22, 3-46 (1926). (Received for publication October 6, 1986)