Earliest domestication of common millet (Panicummiliaceum) in East Asia extended to 10,000 years agoHouyuan Lua,1, Jianping Zhanga, Kam-biu Liub, Naiqin Wua, Yumei Lic, Kunshu Zhoua, Maolin Yed, Tianyu Zhange,Haijiang Zhange, Xiaoyan Yangf, Licheng Shene, Deke Xua, and Quan Lia

aKey Laboratory of Cenozoic Geology and Environment, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China;bDepartment of Oceanography and Coastal Sciences, Louisiana State University, Baton Rouge, LA 70803; cGraduate University of the Chinese Academy ofSciences, Beijing 100049, China; dInstitute of Archaeology, Chinese Academy of Social Sciences, Beijing 100710, China; eCulture Museum of Cishan, Wuan,Hebei Province, 036302, China; and fInstitute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China

Edited by Dolores R. Piperno, Smithsonian Tropical Research Institute and National Museum of Natural History, Washington, DC, and approvedMarch 17, 2009 (received for review January 8, 2009)

The origin of millet from Neolithic China has generally beenaccepted, but it remains unknown whether common millet (Pani-cum miliaceum) or foxtail millet (Setaria italica) was the first speciesdomesticated. Nor do we know the timing of their domesticationand their routes of dispersal. Here, we report the discovery of huskphytoliths and biomolecular components identifiable solely ascommon millet from newly excavated storage pits at the NeolithicCishan site, China, dated to between ca. 10,300 and ca. 8,700calibrated years before present (cal yr BP). After ca. 8,700 cal yr BP,the grain crops began to contain a small quantity of foxtail millet.Our research reveals that the common millet was the earliest dryfarming crop in East Asia, which is probably attributed to itsexcellent resistance to drought.

Holocene � origins of agriculture � phytoliths � Neolithic � Cishan

Foxtail millet (Setaria italica) and common millet (or broom-corn millet; Panicum miliaceum) were among the world’s

most important and ancient domesticated crops. They werestaple foods in the semiarid regions of East Asia (China, Japan,Russia, India, and Korea) and even in the entire Eurasiancontinent before the popularity of rice and wheat (1–4), and arestill important foods in these regions today (5, 6).

Thirty years ago, the world’s oldest millet remains, dating toca. 8,200 calibrated years before present (cal yr BP), werediscovered at the Early Neolithic site of Cishan, northern China.The site contained �50,000 kg of grain crops stored in thestorage pits (7–9). Until now, the importance of these findingshas been constrained by limited taxonomic identification withregard to whether they are from foxtail millet (S. italica) orcommon millet (P. miliaceum), because the early reported S.italica identifications are not all accepted (4, 9–12). This articlepresents the phytoliths, biomolecular records, and new radio-carbon dating from newly excavated grain crop storage pits at theCishan site. Large modern reference collections are used tocompare and contrast microfossil morphology and biomolecularcomponents in different millets and related grass species (13).The renewed investigations show that common millet agriculturearose independently in the semiarid regions of China by 10,000cal yr BP. Our findings contribute to our knowledge of agricul-tural origins across the globe and have broader implications forunderstanding the development of human societies.

The Cishan site (36°34.511� N, 114°06.720� E) is located nearthe junction between the Loess Plateau and the North ChinaPlain at an elevation of 260–270 m above sea level (Fig. 1). Thearchaeological site, containing a total of 88 storage pits withsignificant quantities (�109 m3) of grain crop remains, wasexcavated from 1976 to 1978 (7, 8). Each storage pit included0.3- to 2-m-thick grain crops, which were well preserved andfound in situ in the 3- to 5-m-deep loess layer (9). All grainremains have been oxidized to ashes soon after they wereexposed to air. Archaeological excavations also revealed theremains of houses and numerous millstones (Fig. S1), stone

shovels, grind rollers, potteries, rich faunal remains, and plantassemblages including charred fruits of walnut (Juglans regia),hazel (Corylus heterophylla), and hackberry (Celtis bungeana)(7–9). Only 2 14C dates of charcoal from previously excavatedH145 and H48 storage pits yielded uncalibrated ages of 7355 �100 yr BP and 7235 � 105 yr BP, respectively (8). Theseremains represent the earliest evidence for the significant useof dry-farming crop plants in the human diet in East Asia. Theyalso suggest that by this time agriculture had already beenrelatively well developed here.

Early identification assumed these grain crop remains to befoxtail millet (S. italica). This preliminary identification wasmainly based on a characteristic of very small sizes of grain cropash (no charred grains)—often �2–3 mm in length—that re-sembled the foxtail millets (Fig. S2), but without any spodogramevidence for the grain crop remains in the Cishan site (7–10).These have been reported as the world’s oldest foxtail millet inthe literature (6, 9, 10, 14). However, the millet identification hasbeen questionable (4, 11, 12, 15), because the macro (ash)remains were too friable to be observed under the microscope.Furthermore, very little study has been conducted on thespodograms or phytoliths of modern millets, so no clear diag-nostic feature has been used to distinguish foxtail millet fromcommon millet (12, 16). Previous study has also considered atsome length how the charred dehusked grains of various nativemillet species might have been systematically misidentified (3).Thus, questions remain regarding whether the Cishan graincrops are from foxtail millet or common millet, or both, becausethey cannot be distinguished from the ash vestige (11, 12).

ResultsPhytoliths and biomolecular components have provided sub-stantial empirical evidence demonstrating the considerable an-tiquity of food production and crop dispersals in many regionsof the world (17–20). Based on our observation and statistics ofthe variation of anatomy and silicon structure patterns in theglumes, lemmas, and paleas occurring in modern cultivatedmillets and related grass species collected from different regionsin China (13), we found that 5 key diagnostic characteristics inphytolith morphology could be used to distinguish betweenfoxtail millet and common millet (13): (i) a cross-shaped typephytolith is formed in the lower lemma and glume of S. italica,

Author contributions: H.L. and N.W. designed research; H.L., J.Z., Y.L., K.Z., M.Y., T.Z., H.Z.,L.S., D.X., and Q.L. performed research; H.L. and Y.L. contributed new reagents/analytictools; H.L., J.Z., K.-b.L., N.W., Y.L., and X.Y. analyzed data; and H.L., K.-b.L., and N.W. wrotethe paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

See Commentary on page 7271.

1To whom correspondence should be addressed. E-mail: houyuanlu@mail.iggcas.ac.cn.

This article contains supporting information online at www.pnas.org/cgi/content/full/0900158106/DCSupplemental.

www.pnas.org�cgi�doi�10.1073�pnas.0900158106 PNAS � May 5, 2009 � vol. 106 � no. 18 � 7367–7372

AN

THRO

POLO

GY

SEE

COM

MEN

TARY

Dow

nloa

ded

by g

uest

on

July

26,

202

0

whereas a bilobe-shaped type is formed in those of P. miliaceum;(ii) regularly arranged papillae on the surface of the upperlemma and paleas are peculiar to S. italica; (iii) the epidermallong cell walls are �-undulated in S. italica (Fig. 2H), and�-undulated in P. miliaceum (Fig. 2 B, D, and F); (iv) the endingsstructures of epidermal long cells are cross-wavy type in S. italicaand cross-finger type in P. miliaceum; the R value (ratio of thewidth of endings interdigitation to the amplitude of undulations)

is higher (0.79 � 0.12; n � 3,303) in P. miliaceum than in S. italica(0.33 � 0.11; n � 2,774); and (v) surface ridgy line sculpture ofthe upper lemma is peculiar to S. italica. These 5 diagnosticcharacteristics used together give the only reliable way ofdistinguishing foxtail millet from common millet when onlypowder remains are available. In addition, a species-specificidentification of phytoliths is possible for S. italica and P.miliaceum because they have typically well-defined silica skele-

Fig. 1. Locality map. (A) Map showing the location of Cishan and other important Early Neolithic millet agricultural sites in the semiarid region of northeasternChina. (B) The Cishan site is located on a terrace of the Ming River (Inset). A detailed plan of the west area of Cishan site excavated in 1976–1978, showing theoutlines of the serried storage pits and the 5 newly excavated storage pits, CS-I to CS-V (numbers in brackets), in an area to the northwest of the earlier excavationis presented. (C) A photograph of the newly excavated storage pit CS-V, found on the cliff of the northern terrace. (D) Close-up photograph of the loose layerof grain crop remains in storage pit CS-III found in situ in the loess layer.

A B

DC

E F

HG

Fig. 2. Scanning microscopic interferometer photographs. (A and B) Phytoliths from CS-I storage pit (A), compared with modern �-I type husk phytoliths fromP. miliaceum (B). (C and D) Phytoliths from CS-II storage pit (C), compared with modern �-II type husk phytoliths from P. miliaceum (D). (E and F) Phytoliths fromBWG (E), compared with modern �-III type husk phytoliths from P. miliaceum (F). (G and H) Phytoliths from CS-II storage pit (G), compared with modern �-II typehusk phytoliths from S. italica (H). (I) Bivariate biplot showing coordinates of the 3,303 measurements from epidermal long cells of P. miliaceum and 2,774measurements from those of S. italica, plotted along axis W (width of endings interdigitation of dendriform epidermal long cells) and axis R (ratios of W toundulations amplitude of dendriform epidermal long cells), and their classification into 2 groups corresponding to 2 species (P. miliaceum and S. italica) (13).Also plotted are the fossil samples of husk phytoliths from CS-I-V and BWG, interpreted to be of P. miliaceum origin, dated between ca. 10,300 and ca. 7,500yr BP. The CS-II, V, and BWG samples contained 0.4–2.83% �-type husk phytoliths, interpreted to be of S. italica origin, dated to less than ca. 8,700 yr BP.

7368 � www.pnas.org�cgi�doi�10.1073�pnas.0900158106 Lu et al.

Dow

nloa

ded

by g

uest

on

July

26,

202

0

tons that are distinguishable from those in Panicum bisulcatum,Setaria viridis, and Setaria plicata, which have no such demon-strable patterns (13).

To determine the taxa of foxtail millet and common millet, weanalyzed the phytoliths of 46 grain crop samples stored in 5 newlyexcavated storage pits (CS-I to CS-V) at the Cishan site, and 1sample (BWG) preserved in a storage bottle from the CultureMuseum of Cishan. These grain crop samples are dated betweenca. 10,300 and ca. 7,500 cal yr BP based on new 14C datingmeasurements (Fig. 3, Table S1).

All 47 archaeological samples we analyzed contained abun-dant diagnostic husk phytoliths that can be divided into 2 groupsaccording to their phytolith assemblages and 14C dating results.The first group, including 27 samples from CS-I, III, and IV, isdated between ca. 10,300 and ca. 8,700 cal yr BP. All of the huskphytoliths present are diagnostic of common millet based ontheir shapes, � patterns (Fig. 2 A, C, and E), and average R value(�0.7) (Fig. 2I). The second group, including 20 samples fromCS-II, V, and BWG, is dated between ca. 8700 and ca. 7,500 calyr BP. More than 97% of the husk phytoliths are also diagnosticof the common millet (97.2% for CS- II, n � 1,273; 97.5% forCS-V, n � 1,000; 99.6% for BWG, n � 1,000), but a smallquantity (0.4–2.8%) of the husk phytoliths in the second groupcan be attributed to S. italica (Fig. 2 G and I).

Fig. 4 shows that the CS-I storage pit contains �1.5-m-thickgrain crop remains composed of 3 prominent layers of lemmaand palea from common millets alternating with 3 glume plusreed layers. This storage manner indicates that prehistorichumans had known how to preserve large volumes of grains insecure storage. They did so by digging deep storage pits in thedry loess strata and by covering the floors with thick mats ofmillet glumes and reed (Phragmites australis) leaves.

Identification of biomolecular components was mainly basedon our modern reference collection. A recent study has used thebiomolecular components of P. miliaceum, the only miliacin-exclusive producer reported, as a basis for identification (20).However, because of the lack of comparable data derived fromS. italica, previous investigators are still not clear how to usebiomarkers to distinguish P. miliaceum from S. italica. In thisstudy, we examined biomolecular components in 6 samples of

modern Paniceae (see Materials and Methods). The results showthat biomolecular components can be used to distinguish be-tween P. miliaceum and S. italica based on the presence orabsence of 5 biomarkers—miliacin, �-amyrin methyl ether(ME), and 3 pentacyclic triterpene methyl ethers (PTMEs),although the structures of the 3 PTMEs have yet to be confirmed(Fig. 5, Fig. S3). The total ion current trace shows the wholedistribution of aromatic hydrocarbons and ethers extracted fromP. miliaceum (Fig. 5A) and S. italica (Fig. 5B) to be in the 54- to62-min analysis time range. The significant relative abundance ofmiliacin (compound 2) (Fig. 5D) was found in both modernspecies (89.0 � 1.64% for P. miliaceum and 33.8 � 22.2% for S.italica); the miliacin relative abundances in the aromatic hydro-carbons and ethers were estimated by measuring the area of themiliacin peak on the m/z 189 � 204 � 218 ion-specific chro-matogram. However, compounds 1, 4, and 5 are peculiar to P.miliaceum, and compound 3 is peculiar to S. italica.

Fig. 5 shows that the prominent compound products of thearchaeological samples (BWG, CS-V-03) are compounds 1, 2, 4,and 5, which are also present in P. miliaceum. However, compounds1, 4, and 5 are notably absent from S. italica, which containcompounds 2 and 3 only. Thus, the distribution pattern of majorbiomarkers of BWG and CS-V-03 grain crop remains is similar tothat of modern common millet (Fig. 5A). In 2 archeologicalsamples, including BWG and CS-V-03, the relative abundance ofmiliacin reaches 88.5% and 88.2%, respectively, which is also verysimilar to that of common millet. These results provide furthersupport to our conclusion and suggest that common millet is animportant source of grain crops stored in the storage pits.

DiscussionAccording to archeobotanical research, the early charred grainsof common millet occurred during the initial stages of variousEarly Neolithic sites (Fig. 1), including Dadiwan (ca. 7.8–7.35 calkyr BP) (21), Xinglonggou (ca. 8.0–7.5 cal kyr BP) (22), and

Fig. 3. Carbon-14 dates and chronology-corrected dates of samples exca-vated at Cishan site. Lab no: GZ, Laboratory of Peking University AcceleratorMass Spectrometry and Key Laboratory of Isotope Geochronology and Geo-chemistry, Guangzhou, Chinese Academy of Sciences; CNL, Radiocarbon Lab-oratory of the Institute of Geology and Geophysics, Chinese Academy ofSciences; ZK, Laboratory of the Institute of Cultural Relics of the ChineseBureau of Cultural Relics (8). Red box, calendar 68% range, by CalPal, Univer-sity of Cologne Radiocarbon Calibration Program Package (www.calpal.de/).G, grain crops; C, charcoal. Fig. 4. Percentage diagram of 4 phytolith types in CS-I, showing 3 prominent

layers of lemma and palea phytoliths from common millets alternating with 3layers of mixed common millet glumes, reed, and panicoid types. Threeradiocarbon dates were obtained from the grain crops. Ages shown aredendro-corrected calendar years BP. Minor age reversal is attributed to theintrusion of roots from the modern grasses growing on the cliff foot face.

Lu et al. PNAS � May 5, 2009 � vol. 106 � no. 18 � 7369

AN

THRO

POLO

GY

SEE

COM

MEN

TARY

Dow

nloa

ded

by g

uest

on

July

26,

202

0

Yuezhuang (ca. 7.87 cal kyr BP) (23) in North China, but foxtailmillet was barely present during these stages. Lee et al. (24) havespeculated that the Early Neolithic predominance of broomcornover foxtail millet at Xinglonggou and Yuezhuang ca. 6000 calB.C. might be a regional phenomenon, implying that broomcornmillet might have been domesticated earlier than foxtail millet.Our analytical results of both phytoliths and biomolecularcomponents have established that the earliest cereal remains

stored in the Cishan Neolithic sites, during ca. 10,300–8,700 calyr BP, are not foxtail millet, but only common millet. After 8,700cal yr BP, the grain crops gradually contained 0.4–2.8% foxtailmillet. Our study also suggests that common millet was used asa staple food significantly earlier than foxtail millet in northernChina. It provides direct evidence to show that, by 10,000 cal yrBP, the early people in northern China had developed variousmethods of maintenance and multiplication of millet seeds for

A

B

C

D

Fig. 5. Total ion chromatogram of extracted aromatic hydrocarbons and ethers from common millet (A), foxtail millet (B), and grain crops from BWG andCS-V-03 (C). Peaks labeled 1, 2, 3, 4, and 5 correspond to PTME-1 [M� 440, m/z, 425, 393, 257, 218, 204, 189, 161, 135, 109 (100%), 95, 69] (Fig. S3A), PTME-2 [miliacin,olean-18-en-3�-ol ME, M� 440, m/z, 425, 393, 257, 218, 204, 189 (100%), 177, 161, 135, 109, 95, 69], PTME-3 [M� 440, m/z, 425, 397 (100%), 365, 261, 229, 218,204, 189, 175, 161, 135] (Fig. S3B), PTME-4 [�-amyrin ME, urs-12-en-3�-ol ME, M� 440, m/z, 393, 259, 218 (100%), 203, 189, 161, 135, 109, 95] (Fig. S3C), and PTME-5[M� 440, m/z, 425, 408, 393, 257, 221, 203, 189 (100%), 147, 135, 121, 109, 95] (Fig. S3D), respectively. Mass spectra of PTME-2 (miliacin, olean-18-en-3�-ol ME)are presented in D.

7370 � www.pnas.org�cgi�doi�10.1073�pnas.0900158106 Lu et al.

Dow

nloa

ded

by g

uest

on

July

26,

202

0

the next generation, and had known how to store crops of staplefood in secure, dry places of storage pits during the EarlyNeolithic epoch.

Common millet has the lowest water requirement among allgrain crops; it is also a relatively short-season crop, and couldgrow well in poor soils (5, 6, 25). The geographical distributionof both foxtail millet and common millet in China (Fig. S4) showsthat foxtail millet is more common in the semiwet eastern areas,and its optimal growth occurs at mean annual temperature(MAT) from 8 to 10 °C and mean annual precipitation (MAP)from 450 to 550 mm. However, common millet is more adaptedto the drier interior areas, and its optimal growing conditionsoccur at MAT from 6 to 8 °C and MAP from 350 to 450 mm (5,6). The origin and dispersal of millet agriculture is a key problemclosely related to the history of human impact on the environ-ment and transformation of natural vegetation.

Paleoenvironmental data from the Weinan section (26–29)(Fig. 1) in the southern part of the Loess Plateau between theCishan and Dadiwan sites are crucial for understanding the earlystage of the forager–cultivator transition. The early Holocenewas a period of significant environmental change marked by dryclimate conditions as inferred from sediment texture (26, 28),magnetic susceptibility (26, 28), pollen (27), phytoliths (28), andmollusk assemblages (29). These proxy records show an envi-ronmental transition from cold–dry (ca. 11,000–8,700 cal yr BP)to warm–wet (ca. 8,700–5,500 cal yr BP) conditions. Manylacustrine and loess records from the Chinese Loess Plateau toCentral Asia also support the scenario of a dry climate during theearly Holocene (30–34). Under the drier climate conditions, soildevelopment was slowed, and the soil developed on the under-lying older and coarser loess of the glacial period was poor innutrients (28). This raises the possibility that common millet wasmore significant than foxtail millet in the early stages of foodproduction in North China because it was more adaptable thanfoxtail millet to the dry condition prevailing during the earlyHolocene. The common millet cultivation may involve complexselection by natural forces and human activities, although noclear evidence has been documented in this region for thetransitions from gathering to cultivation and/or from a wildancestor to domesticated common millet (1, 2, 5).

ConclusionsOur research indicates that the earliest significant commonmillet cultivation system was established in the semiarid regionsof China by 10,000 cal yr BP, and that the relatively dry conditionin the early Holocene may have been favorable for the domes-tication of common millet over foxtail millet. Our study showsthat common millet appeared as a staple crop in northern China�10,000 years ago, suggesting that common millet might havebeen domesticated independently in this area and later spread toRussia, India, the Middle East, and Europe. Nevertheless, likeMesopotamia, where the spread of wheat and barley to the fertilefloodplains of the Lower Tigris and Euphrates was a key factorin the emergence of civilization, the spread of common millet tothe more productive regions of the Yellow River and its tribu-

taries provided the essential food surplus that later permitted thedevelopment of social complexity in the Chinese civilization.

Materials and MethodsPhytolith Extraction from Archaeological Material. The Cishan site is located onthe north terrace of the Ming River (Fig. 1). The geography of this regionconsists of alluvial terraces covered by loess deposits. Five storage pits (CS-I toCS-V) are situated on the edge of the terrace. The terrace is composed of thicksediments (4–5 m thick for CS-I to CS-IV, 11 m thick for CS-V) that consist ofalternating layers of loess and soil. Each of the storage pits (CS-I to CS-V)contains �1- to 2-m-thick grain crop remains, being well preserved in the 3- to5-m-deep loess layer (Table S1).

A total of 46 samples from 5 storage pits (CS-I to CS-V) at the Cishan site and1 sample (BWG) from the Culture Museum of Cishan were analyzed. Of these,16 continuous samples were taken at 10-cm intervals from CS-I storage pit. Thesamples were prepared according to the procedure slightly modified fromPiperno (35, 36) and Lu et al. (37). It consists of sodium pyrophosphate(Na4P2O7) deflocculation, treatment with 30% hydrogen peroxide (H2O2) andcold 15% hydrochloric acid (HCl), zinc bromide (ZnBr2; density, 2.35 g/cm3)heavy liquid separation, and mounting on a slide with Canada balsam. Phy-tolith counting and identification were performed by using a Leica microscopewith phase-contrast and microscopic interferometer at 400 magnification.

Millet Molecule Extractions from Modern Plants and Archaeological Material.We examined modern millet molecules from 6 domesticated samples ofPaniceae, including 3 samples of Setaria italica L. Beauv. (foxtail millet) and 3samples of Panicum miliaceum L. (common millet). Two archaeological sam-ples of grain crop remains, including 1 sample from CS-V storage pit and 1sample (BWG) in a storage bottle from the Culture Museum of Cishan, wereanalyzed.

One gram of each of dried sample (smaller than 100 mesh) was ultrasoni-cally extracted for 3 times with a mixture of acetone and pentane at 1:1(vol/vol). After filtration, the mixture was eluted with n-hexane to givealiphatic hydrocarbons (N). Using dichloromethane in n-hexane passingthrough the column resulted in satisfactory separation of aromatic hydrocar-bons and ethers (A). The extracted subfractions (A) were then analyzed by gaschromatography coupled with mass spectrometry (GC/MS) .

GC-MS was conducted by using an Agilent 6890 N GC-5973 N MSD massselective detector system (EI 70 eV; ion source temperature, 230 °C) fitted witha fused silica chemically bonded capillary column (J&W DB-5; 0.25 mm indiameter, 30 m long, 0.25 �m film thickness). Each sample was injected ontothe column at 280 °C in the splitless mode. After a 1-min isothermal hold at50 °C, the column temperature was increased by 30 °C/min to 120 °C; and then3 °C/min to 290 °C, successively, with a 30-min isothermal hold at 290 °C. Theflow rate of the helium carrier gas was 1.2 mL/min (40 cm/sec).

Compounds were identified by their retention time within the gas chro-matograph, their fragmentation pattern within the mass spectrometer, andby matching their mass spectra to reference spectra from the literature or fromlibraries (NIST02L). The miliacin relative abundances were estimated by mea-suring the area of the miliacin peak on the m/z 189 � 204 � 218 ion-specificchromatogram. Note that this value may vary under different GC/MS condi-tions, and there will be a small change in the abundance of this ion.

ACKNOWLEDGMENTS. We are grateful to Dolores R. Piperno for criticallyreading the original manuscript and her helpful comments in improving thismanuscript. We also greatly appreciate the valuable comments from twoanonymous reviewers. We are grateful to Jacob Jeremy and A. Gerasimenkofor their help on biomolecular analyses, and D. Q. Fuller and L. Qin for theirdiscussions on phytolith analyses. We thank C. D. Shen, Z. Y. Gu, and B. Xu fortheir help in radiocarbon age measurement. This work was supported by theChinese Academy of Sciences (100 Talents Program, kzcx2-yw-117), the Na-tional Natural Science Foundation of China (40771216; 40325002), ChineseCivilization Origin projects (2006BAK21B02), and the U.S. National ScienceFoundation (BCS-0623514, ATM-0402475).

1. Bellwood P (2005) First Farmers: The Origins of Agricultural Societies (Blackwell,Malden, MA).

2. Zohary D, Hopf M (2000) Domestication of Plants in the Old Word: The Origin andSpread of Cultivated Plants in West Asia, Europe and the Nile Valley (Oxford UnivPress, Oxford).

3. Fuller DQ (2006) Agricultural origins and frontiers in South Asia: A working synthesis.J World Prehist 20:1–86.

4. Crawford G (2005) East Asian plant domestication. Archaeology of Asia, ed Stark MT(Blackwell, Malden, MA), pp 77–95.

5. You XL (1993) The question for origin and spread in both foxtail millet and commonmillet. Agric Hist China 12:1–13.

6. Li Y, Wu SZ (1996) Traditional maintenance and multiplication of foxtail millet (Setariaitalica. (L.) P. Beauv.) landraces in China. Euphytica 87:33–38.

7. Handan Relics Preservation Station, Education Team of Cishan Archaeology (1977)Preliminary excavation of Neolithic sites in Cishan, Hebei Province. Kaogu (Archaeol-ogy) 6:361–372.

8. CPAM Hebei Province, Handan Relics Preservation Station (1981) The Cishan site inWu’an, Hebei Province. Acta Archaeologica Sinica 3:303–338.

9. Tong WH (1984) Original agricultural relics of Cishan sites and related problems. AgricArchaeology 1:194–207.

10. Huang QX (1982) Spodograms study and its application in archaeology. Kaogu (Ar-chaeology) 4:418–420.

Lu et al. PNAS � May 5, 2009 � vol. 106 � no. 18 � 7371

AN

THRO

POLO

GY

SEE

COM

MEN

TARY

Dow

nloa

ded

by g

uest

on

July

26,

202

0

11. Zhao ZJ (2006) Domestication of millet—paleoethnobotanic data and ecological per-spective. Archaeology in China and Sweden, eds Institute of Archaeology ChineseAcademy of Social Sciences and the Institute of Archaeology Swedish National Heri-tage Board (Science Press, Beijing), pp 97–104.

12. Harvey EL, Fuller DQ (2005) Investigating crop processing using phytolith analysis: Theexample of rice and millets. J Archaeol Sci 32:739–752.

13. Lu HY, et al. (2009) Phytoliths analysis for the discrimination of foxtail millet (Setariaitalica) and common millet (Panicum miliaceum). PLoS ONE 4:e4448.

14. Fukunaga K, Ichitani K, Kawase M (2006) Phylogenetic analysis of the rDNA intergenicspacer subrepeats and its implication for the domestication history of foxtail millet,Setaria italica. Theor Appl Gene 113:261–269.

15. Hunt HV, et al. (2008) Millets across Eurasia: Chronology and context of early recordsof the genera Panicum and Setaria from archaeological sites in the Old World. VegetHist Archaeobot 17(Suppl 1):S5–S18.

16. Nasu H, Momohara A, Yasuda Y, He JJ (2007) The occurrence and identification ofSetaria italica (L.) P. Beauv. (foxtail millet) grains from the Chengtoushan site (ca. 5800cal B.P.) in central China, with reference to the domestication centre in Asia. Veget HistArchaeobot 16:481–494.

17. Pearsall DM, et al. (1995) Distinguishing rice (Oryza sativa Poaceae) from wild Oryzaspecies through phytolith analysis: Results of preliminary research. Econ Bot 49:183–196.

18. Piperno DR, Weiss E, Holst I, Nadel D (2004) Processing of wild cereal grains in the UpperPalaeolithic revealed by starch grain analysis. Nature 430:670–673.

19. Lu HY, et al. (2005) Millet noodles in Late Neolithic China. Nature 437:967–968.20. Jacob J, et al. (2008) Millet cultivation history in the French Alps as evidenced by a

sedimentary molecule. J Archaeol Sci 35:814–820.21. Liu CJ, Kong ZC (2004) The morphological comparison of grains between foxtail and

broomcorn millet and its application for identification in archaeology remains. Kaogu(Archaeology) 3:76–83.

22. Zhao ZJ (2005) Palaeoethnobotany and its new achievements in China. Kaogu (Ar-chaeology) 7:42–49.

23. Crawford GW, Chen XX, Wang JH (2006) Houli Culture rice from the Yuezhuang site,Jinan. Orient Archaeol 3:247–251.

24. Lee GA, Crawford GW, Li L, Chen XC (2007) Plants and people from the Early Neolithicto Shang periods in North China. Proc Natl Acad Sci USA 104:1087–1092.

25. Nesbitt M (2005) Grains. Cultural History of Plants, eds Prance G, Nesbitt M (Routledge,London), pp 45–60.

26. Guo ZT, et al. (1996) High frequency pulses of East Asian monsoon climate in the lasttwo glaciations: Link with the North Atlantic. Clim Dyn 12:701–709.

27. Sun XJ, Song CQ, Wang FY, Sun MR (1997) Vegetation history of the loess plateau ofChina during the last 100,000 years based on pollen data. Quat Int 37:25–36.

28. Lu HY, Wu NQ, Liu KB, Jiang H, Liu TS (2007) Phytoliths as quantitative indicators forthe reconstruction of past environmental conditions in China II: Palaeo-environmentalreconstruction in the Loess Plateau. Quat Sci Rev 26:759–772.

29. Lu HY, Zhang JP (2008) Neolithic cultural evolution and Holocene climate in theGuanzhong basin, Shaanxi, China. Quat Sci 28:1050–1060.

30. Feng ZD, Tang LY, Wang HB, Ma YZ, Liu K-b (2006) Holocene vegetation variations andthe associated environmental changes in the western part of the Chinese Loess Plateau.Palaeogeogr Palaeoclimatol Palaeoecol 241:440–456.

31. Feng ZD, An CB, Tang LY, Jull AJT (2004) Stratigraphic evidence of a megahumidclimate between 10,000 and 4000 years B.P. in the western part of the Chinese LoessPlateau. Glob Planet Change 43:145–155.

32. Boomer I, Aladin N, Plotnikov I, Whatley R (2000) The palaeolimnology of the Aral Sea:A review. Quat Sci Rev 19:1259–1278.

33. Fowell SJ, Hansen BCS, Peck JA, Khosbayar P, Ganbold E (2003) Mid to late Holoceneclimate evolution of the Lake Telmen Basin, North Central Mongolia, based on pa-lynological data. Quat Res 59:353–363.

34. Feng ZD, An CB, Wang HB (2006) Holocene climatic and environmental changes in thearid and semi-arid areas of China: A review. Holocene 16:119–130.

35. Piperno DR (1988) Phytolith Analysis. An Archaeological and Geological Perspective(Academic, San Diego).

36. Piperno DR (2006) Phytoliths: A Comprehensive Guide for Archaeologists and Paleo-ecologists (Alta Mira Press, Oxford).

37. Lu HY, et al. (2006) Phytoliths as quantitative indicators for the reconstruction of pastenvironmental conditions in China I: Phytolith-based transfer functions. Quat Sci Rev25:945–959.

7372 � www.pnas.org�cgi�doi�10.1073�pnas.0900158106 Lu et al.

Dow

nloa

ded

by g

uest

on

July

26,

202

0