Wheat is regarded as one of the most important West Asian domesticates that were introduced into Late Neolithic/Early Bronze Age China. Despite a growing body of archaeological data, the timing and routes of its dispersal remain controversial. New radiocarbon (14C) dating evidence from six archaeological sites in the Shandong and Liaoning Peninsulas and Bayesian modelling of available 14C data from China suggest that wheat appeared in the lower Yellow River around 2600 Before Common Era (bce), followed by Gansu and Xinjiang around 1900 bce and finally occurred in the middle Yellow River and Tibet regions by 1600 bce. These results neither support long-standing hypotheses of a progressive spread of wheat agriculture from Xinjiang or Gansu to eastern China nor suggest a nearly synchronous appearance in this vast zone, but corroborate transmission to lower Yellow River elites as an exotic good through cultural interactions with the Eurasian steppe along north–south routes.
Wheat (Triticum spp.) was domesticated in West Asia from at least 8500 Before Common Era (bce)1. Its westward spread into Europe as a sign of Neolithization has been well reconstructed2. However, debates about its spread into East Asia continue3, centring on timing, dispersal routes and the effect of wheat on prehistoric societies in China4,5. Another unresolved issue is the motives of people moving wheat over such long distances6,7, especially given that agriculture was already well established in China based on locally domesticated cereal crops, notably rice (Oryza sativa) and millets (Panicum miliaceum and Setaria italica)8.
So far, the appearance of wheat in China dates to the Late Neolithic/Early Bronze Age9. This period is often associated with increasing interaction among different groups in Eurasia10. Wheat was possibly introduced together with barley (Hordeum vulgare), sheep (Ovis aries), goat (Capra aegagrus ssp. hircus), cattle (Bos taurus) and bronze metallurgy11. It is believed that these new cultural elements broadened subsistence strategies and revolutionized material culture that supported a rapid change towards higher social complexity in northern China12. Following traditional views, the onset of the Bronze Age signifies the establishment of states13. Thus, the spread of wheat is closely linked to one of the most hotly debated questions in Chinese archaeology addressing the timing of the birth of the Chinese civilization.
Different dispersal routes have been suggested hypothesizing that wheat cultivation either gradually spread (1) along the course of the later Silk Road via Xinjiang and the Hexi Corridor in modern Gansu to the middle and lower Yellow River regions14 or (2) via the Mongolian steppe to the Hexi Corridor and from there into western (Xinjiang) and eastern China (for example, Shaanxi, Henan and Shandong)3,4 or appeared (3) in multiple places throughout northern China at approximately the same time5,14. A comprehensive review5 evaluating these most prominent hypotheses summarized their weak points and put forward an alternative model arguing against ‘a wave of wheat-farming colonists’. In a more recent study15, grain size data of prehistoric and early historic wheat remains from different parts of Eurasia have been used to unravel the diffusion of wheat in ancient China. Based on the spatiotemporal decrease in grain size from West Asia towards eastern China inferred from a data set of carbonized wheat grain dimensional (length and width) measurements, the authors suggest a discontinuous, west-to-east dispersal of this crop. Despite a growing body of studies, there is still only poor knowledge about the timing and patterns of wheat introduction into prehistoric China.
In this paper, we report a new set of 14C dates obtained from carbonized wheat grains from the lower Yellow River region (Fig. 1a). The size of the obtained caryopses was determined and compared with other available size data of grains from different regions. Bayesian modelling was applied to a systematic collection of published and newly obtained wheat-based dates to reconstruct the chronology of wheat appearance in different regions of China. Dispersal routes and motives behind the adoption of wheat are discussed in the context of the derived chronology.
Samples and 14C dating
Abundant remains of several cultivated crops (including two types of millet, rice and wheat) were recovered from six key archaeological sites in Shandong and on the Liaondong Peninsula (Fig. 1a) using standard water flotation16. Plant macrofossils were identified with reference to the archaeobotanical collection at Shandong University. The six sites, including Dinggong, Dongpan, Shilipubei, Wangjiacun, Wutai and Zhaojiazhuang, date to the Late Neolithic or Early Bronze Age cultural complexes17 (Table 1 and Supplementary Information 1). The final excavation report has not been published yet for any of these sites, but according to preliminary information, Dinggong is the most prominent site. It is a settlement founded during the Dawenkou cultural period17 (Table 1). During the early Longshan period (prior to ca. 2300 bce), it was fortified with walls, which were substantially expanded in the middle Longshan time18. Salt production and trade might have been the base for its development and growing wealth19. One of the two known earliest scripts carved in pottery sherds19 was found within the walls; even though it has not yet been deciphered, it indicates the particularity of the site.
We selected ten well-preserved caryopses of bread wheat (Triticum aestivum) (Table 2) along with other representative crops for accelerator mass spectrometry (AMS) 14C dating in the 14C laboratory at Poznan University (Methods). Three of the dated wheat samples (Poz-82358, Poz-82362 and Poz-82352) from the Dinggong and Zhaojiazhuang sites have a median calibrated age older than 2000 bce. Another three wheat samples (Poz-82365, Poz-82367 and Poz-82349) date to 1200–500 bce (calibrated median). The four samples (Poz-82348, Poz-82368, Poz-82354 and Poz-82356) from the Dongpan, Wangjiacun and Wutai sites are evidently too young for their context, dating to, respectively, 661 Common Era (ce), post-1950 ce, 1827 ce and 1483 ce. We assume that these much younger seeds reflect redeposition from upper cultural layers downwards into relatively older ones, which is not an uncommon phenomenon20.
Wheat grain size
We compiled a data set of wheat grain size (Fig. 2 and Supplementary Table 2) based on the measurements of length and width dimensions, including those of the intact (n = 8) caryopses presented and dated in the current study and other data reported in the literature. The measuring results of the well-preserved grains from Shandong and Liaoning show a large range in length (2.6–4.5 mm) and width (2.0–2.9 mm) dimensions. Reported wheat grain size data15 are available as average values (mm) for different regions including West Asia (length = 4.8; width = 3.0), the upper Yellow River region (length = 4.1; width = 2.9) and the middle Yellow River region (length = 3.5; width = 2.4). Further measurements are available from a small number of grains from three Bronze Age sites in Central Asia, including site 1211 (ref. 21) in southern Turkmenistan and Tasbas22 and Begash23 in eastern Kazakhstan. These grains from Central Asia also illustrate a wide array of dimensions, ranging from 2.0 to 5.0 mm and from 1.9 to 4.3 mm for length and width, respectively.
Chronological modelling of wheat-based 14C dates from prehistoric China
The data set (Supplementary Table 3) of directly dated wheat remains from China is compiled from available publications (n = 100) and also includes the set of seeds presented in this study (n = 10). We assessed their quality and selected 73 for Bayesian chronological modelling in OxCal v.4.3 (see Methods; Supplementary Information 1). These 73 data of early wheat are from modern Gansu, Qinghai, Henan, Shaanxi, Shandong and Liaoning provinces and Tibet and Xinjiang autonomous regions.
The constructed Bayesian model consists of five sub-models (Supplementary Information 1), each named after a geographical region in China: upper Yellow River (that is, Gansu and northwestern Qinghai provinces; 35 data points), middle Yellow River (that is, Shaanxi, Shanxi and Henan provinces; 11 data points), lower Yellow River (that is, Shandong province and southern Liaoning province; 6 data points), Tibet (7 data points) and Xinjiang (14 data points). These five regions are well established in Chinese archaeology, revealing their own characteristic trajectories of cultural development and displaying a distinct typology of cultural materials19.
The modelled age for the appearance of wheat in the upper reaches of the Yellow River is ca. 2000–1700 bce (95% probability range; median: 1900 bce; Fig. 3 and Supplementary Information 1). One 14C date from the region, BA92101 (4,230 ± 250 14C before present (bp), calibrated median ca. 2840 bce, measured in 1993)24, which has been widely cited in the literature3 as one of the earliest wheat remains in China, predates the modelled start by almost 1,000 years (Supplementary Information 1). Owing to its anomalously old age in view of the remaining 34 dates from the same region, as indicated by a box-and-whiskers plot (Fig. 4a), this date was downweighted by our model (Fig. 4b). Additional doubt on the reliability of this date arises from its large-scale measurement uncertainty (250 years) and the much younger age of the archaeological site (ca. 1600 bce) confirmed by a recent re-investigation25.
Following the 95% probability range, wheat appeared in Xinjiang ca. 2100–1700 bce (median: 1900 bce) at more or less the same time as in the upper Yellow River (ca. 2000–1700 bce; median: 1900 bce). Along the middle reaches of the Yellow River and in Tibet, wheat existed since ca. 1800–1500 bce (median: 1600 bce) and ca. 2000–1400 bce (median: 1600 bce), respectively. The earliest presence of wheat is suggested for the lower reaches of the Yellow River dating to ca. 3600–2200 bce (median: 2600 bce). This is supported by the modelled order matrix; the difference between geographical regions of China shows a significantly high probability that the wheat appeared earlier in the lower Yellow River than in the other regions (Tables 3 and 4). The order is followed by the upper Yellow River and Xinjiang, whereas the middle Yellow River and Tibet demonstrate the latest wheat appearance among the five analysed regions.
Bayesian modelling applied to a systematic collection (Supplementary Table 3) of published and newly obtained 14C dates directly derived from charred seed remains suggests that archaeological sites along the lower Yellow River comprise, so far, the earliest reliable evidence of wheat in China. The first appearance of wheat in this region dates to around 2600 bce. Our findings corroborate the reliability of a seemingly too old 14C age (BA061052, calibrated median ca. 2386 bce; Supplementary Table 3) based on a single wheat grain from Shandong published in China26, which has been questioned by other authors27 and largely ignored in studies concerned with uncovering the dispersal of early wheat cultivation across China. Some important, although indirect, support to our findings is provided by toponymic evidence, which reveals the lower Yellow River as a primary source area of wheat in China28. Many towns in Shandong are named by one of the two characters, ‘来’ ([lái]; modern meaning: come) or ‘莱’ ([lái]; modern meaning: goosefoot or cropland), which are considered interchangeable with ‘麦’ ([mài]; modern meaning: wheat or barley) in the more pictorial writing style of oracle bone inscriptions dating to the Shang Dynasty (ca. 1600–1046 bce)29,30. These town names are considered to be toponymic heritage of an early arrival of wheat in Shandong before the formalization of the character script around 1000 bce28.
Many Chinese scholars have long argued that wheat domestication was established locally, perhaps in the lower Yellow River region12, which is the largest producer of this crop in China today. However, this ‘Chinese origins’ theory seems unlikely according to botanical and genetic data that reveal a lack of possible wild progenitors and a limited modern genetic diversity of the Triticum genus in the region31. Wheat grown in the lower Yellow River seems to have been introduced as an already fully domesticated, hexaploid form of bread wheat, which is a hybrid from cultivated tetraploid Triticum turgidum (emmer wheat) and a diploid wild grass Aegilops squarrosa31. Neither progenitor is native to China1.
Today, an external origin of wheat is increasingly accepted among Chinese and international scholars31. So far, there have been different routes suggested along which wheat agriculture spread within China, including the course of the later Silk Road via Xinjiang and the Hexi Corridor (Fig. 1a), a sea route along the coastal areas of South Asia and Southeast Asia, and different northern routes via the Eurasian steppe5. Besides the recently suggested hypothesis3,4 that wheat was first adopted in the Gansu region subsequently spreading towards the west (Xinjiang) and east (Henan and Shandong), the most favoured hypothesis14 to explain the introduction of wheat into China has been a gradual west-to-east dispersal within the narrow Xinjiang–Gansu–Shandong zone. This assumption had a revival in a recent study15 that is based on the spatial patterns of grain size derived from prehistoric charred wheat grains. The key argument for the proposed west-to-east dispersal is the temporal decrease in grain size (especially length) from West Asia towards central and eastern China.
The dimensional measurements (Fig. 2) of the directly dated wheat grains from Shandong and Liaoning presented in the current study illustrate a major variability in grain size that does not agree with the suggested west-to-east progressive decline theory. Further evidence against it is found in the size analysis of the Bronze Age charred bread wheat grains from archaeological sites in southern Turkmenistan and eastern Kazakhstan. Similar to the grains from Shandong and Liaoning, the available measurements strongly scatter in the xy plot (Fig. 2). In addition, most of those Central Asian grains are smaller than the average size derived for the prehistoric upper Yellow River region wheat, thus contradicting with the postulated decline in grain size along a west–east transect. A wide variation in morphology and size is also emphasized for the wheat grains (n > 10,000) extracted from site 1211 (ca. 1400 bce) in southern Turkmenistan21, which is represented by four individual specimens shown in photographs. They clearly demonstrate that the size of wheat caryopses may vary significantly at a single site and over a narrow time window, hampering a clean division of populations or varieties. Consequently, size parameters should be treated carefully when used for interpreting grain populations from different geographical regions. Furthermore, there are different factors, including the carbonization process32 and depositional environments, which may overprint the initial size and shape of grains in different ways that may be barely estimated.
Given our modelling results, it seems most plausible that wheat was introduced into the lower Yellow River region via the Eurasian steppe (Fig. 1b). In addition to the archaeobotanical record, there is an increasing amount of evidence supporting an early linkage between eastern China and the Eurasian steppe with regard to early metallurgy33. Similarities in style and chemical composition of bronze objects found in prehistoric cultural layers in China and the northern steppe exist, dating back to at least ca. 2500 bce34. Newly excavated or reported archaeological sites, such as the fortified Shimao settlement site (ca. 2300–1800 bce; Fig. 1a)35, filled a spatial gap between the heartland of the Chinese civilizations and the Eurasian steppe. At Shimao, bronze knives and stone moulds were found36, and bones of cattle, sheep/goat and horse dominate the zooarchaeological record37. Despite the lack of archaeobotanical data from sites located at China’s northern borders, their archaeological assemblages are indicative of north–south connection routes between the steppe zone and the lower Yellow River that make plausible the first appearance of wheat in Shandong. These ancient routes might have passed through the steppe and the forest steppe of central and eastern Mongolia via the foothills of the Great Xing’an Range and the Liao River basin (Fig. 1), rich in pastures, hunting grounds and in archaeological sites dating back to over 2500 bce38. Another possible transit route goes along the Mongolian Altai (Fig. 1) towards the southern Mongolian Plateau and further to Shandong4,14,31. Although both suggested routes still need securely dated archaeological evidence to further test, they corroborate a theory that binds the dispersion of wheat and barley into eastern Asia to the spread of pastoralists of the Eurasian steppe and their domestic animals5. Genetic evidence39 locates the highest mitochondrial haplotype diversity of eastern Asian sheep breeds in the Mongolian Plateau, making it the most likely entry area and a hub for further dispersal of sheep and goats (and possibly wheat and bronze technology) across northern China.
The age modelling results (Fig. 3) suggest that wheat reached the upper Yellow River region by 1900 bce. Pottery and jade objects from Gansu/Qinghai found in the Dawenkou and Longshan sites in Shandong and the Ordos Plateau proved contacts between the upper and lower Yellow River via the Ordos bent. It seems plausible that wheat may have spread this way from Shandong into Gansu (Fig. 1b) and from there into Xinjiang, as implied by the derived chronology. Conversely, direct contacts between the agro-pastoralists from Central Asia and oases in and around the Taklamakan Desert (Fig. 1) cannot be excluded31. However, today, archaeobotanical records from China are still limited in amount and are spatially imbalanced8. To verify the outlined hypotheses on wheat dispersal routes during the Late Neolithic and Early Bronze Age, more data based on directly dated archaeobotanical material are needed.
Another question arising from our findings is why wheat was introduced into the lower Yellow River earlier than in other regions of China. Some scholars5 link the access to wheat and the relative abundance of it to social elites and prestige economy. The appearance of wheat in the lower Yellow River falls roughly into the Neolithic/Bronze Age transition18,19,38. This period is marked by the co-existence of groups characterized by diversified material cultures and the development of social hierarchy and societal complexity17. Is it the introduction of the West Asian cultural elements that underpinned these fundamental changes at the onset of the East Asian Bronze Age? With regard to wheat, this question is associated with the effect of this domesticate on the existing subsistence economy and the impetus for people to introduce starchy cereals40. After the introduction of wheat in the upper Yellow River region (ca. 1900 bce according to our results), the crop seems to have become a staple comparatively quickly. Isotopic signals in human bone remains allow tracing back of a transformation from a C4-based diet to a mixed C4/C3-based diet through a period of ca. 500 years (ca. 2000–1500 bce)9,41. Both wheat and barley are C3 plants, whereas common and foxtail millets, the traditional crops in northern China that were domesticated prior to ca. 5000 bce, are C4 plants42. An increasing importance of wheat and barley versus millets is also evident from archaeobotanical records reflecting the same period of time43,44. This demonstrates that wheat became an important element of agricultural production in the upper Yellow River region by at least ca. 1500 bce.
Despite its earlier appearance in Shandong, archaeobotanical and isotopic evidence9 suggests that wheat did not become an important crop, partly replacing millets in the middle and lower reaches of the Yellow River45. Some authors9 have even suggested that wheat did not become a staple in the region before ca. 600 ce. During the Late Neolithic and Early Bronze Age, wheat was probably not cultivated as a high-production subsistence base but for alternative purposes. One explanation for this is the suggested close relationship between power, prestige and ‘exotic’ crops6,7. Ethnological and archaeological investigations revealed that elites may take advantage of the rarity and luxury of exotic crops to impress their communities and to differentiate themselves from other social strata, as a way of maintaining social power and controlling hierarchies46. In the lower Yellow River region, a social elite exhibiting wealth and individual status in exceptionally costly furnished tombs emerges during the middle Dawenkou period and becomes a prominent feature during Longshan times18. There was a strong demand for and supply of luxury goods, such as delicate jade ornaments and elegant dishes. In particular, the many elaborate vessels for serving and consuming food and drink are interpreted as signs of feasting during funeral ceremonies19. In this regard, it is conceivable that wheat or products made of wheat might also have been served or presented during social events, such as feasts or religious rituals. This may well explain the modelled early appearance of wheat and the composition of archaeobotanical assemblages showing no significant increase in wheat remains over a long period of time. Except for Taosi19 (Fig. 1a), such power display by elites has not been documented yet for contemporaneous societies of the middle and upper Yellow River regions.
There are also other aspects, which highlight the specific role of the lower Yellow River within Chinese prehistory and may explain the early adoption of wheat. The region seems to have long been a centre of experimenting with the production of exotic crops and goods. Although farming in Shandong was predominated by millets throughout the Neolithic and Bronze Age, early remains of rice were found in the region dating to ca. 6000 bce47. There is also evidence that Shandong was one of the venues of early millet domestication48 and thus probably one of the first regions that saw the transition from foraging to farming in East Asia. Later, during the terminal Neolithic Longshan cultural complex (ca. 2700–1900 bce), Shandong and southern Liaoning are considered to have been one of the main centres of cultural diversity in China38, with well-developed and specialized production systems and exchange networks49. The potter’s wheel was invented and widely used, and the technology of controlled firing at high temperatures was mastered to produce extremely thin-walled and hard-fired ceramic vessels and to smelt copper18. Some of the earliest bronze objects in China were found in the lower Yellow River region, being as old as those found in the Hexi Corridor33,50. These early bronze objects may be considered as direct evidence for contacts of Eurasian steppe cultures with archaeological cultures of northern China, including Shandong and Liaoning.
The new 14C dates and modelling results presented in this study highlight the role of the lower Yellow River region as an early centre of agriculture and of cultural interaction in the Late Neolithic and Early Bronze Age. The data neither support a simplistic, west-to-east spread of wheat cultivation across China nor a nearly synchronous appearance of wheat agriculture in the vast zone between Xinjiang and the Yellow Sea coast, but corroborate the Eurasian steppe dispersal route framework. Considering the high economic potential and technological lead compared to middle and upper Yellow River societies, the evolving Shandong elites had a demand for exotic goods, including crops, and they had something to offer in exchange, for example, salt. Thus, the area might have been a favoured target for inhabitants of the Eurasian steppe.
AMS 14C dating and calibration
Samples were pre-treated with an acid–base–acid method, followed by combustion and graphitization prior to AMS measurement51. Conventional ages from 14C dates were calibrated to the calendar age based on the IntCal13 calibration curve52 using OxCal v.4.3 (ref. 53).
Wheat grain size data compilation
To verify an earlier hypothesis15 of wheat dispersal across East Asia that was built on changes in grain size, we compiled a data set of grain size described by length and width measurements. Biometric measurements of the wheat specimens from the lower Yellow River region (Fig. 2) were taken from intact seeds showing no or minor damage. The two dimensions were measured with the dorsal side of the seed facing up. This has been also done for the four wheat grains from site 1211 (ca. 1400 bce) in southern Turkmenistan21. These four grains are representative for the large size variation within the number (n > 10,000) of grains mostly extracted from caches. Measurements for the single grain from the Begash site and the average value for the three grains from the Tasbas site (both in eastern Kazakhstan) were extracted from the primary sources22,23. As for the categories West Asia as well as the upper and middle Yellow River regions, we extracted the average values from a secondary source15 listed by province. Uncertainty ranges (1σ) for these three average size values15 are shown.
14C data selection for data set construction
In addition to directly 14C-dated wheat grains, available publications often refer the age of wheat remains to typological dates and 14C dates based on stratigraphically correlated material (for example, carbon-rich soil and charcoal)54. Such correlations involve a high risk of introducing additional errors, including those resulting from post-depositional processes55 or the ‘old wood effect’56, which may lead to ages that differ from the actual time of deposition by hundreds or even thousands of years55. Thus, only wheat-grain-based 14C dates were incorporated into the data set.
Bayesian chronological modelling
Accepted dates were used to construct a Bayesian chronological model, using the built-in phase model57 in OxCal as a building block (that is, a sub-model). These sub-models were combined together as one integrated overlapping multi-phase model to allow for the calculation of their relative chronological order. The overlapping multi-phase model also allows independence of all sub-models; the results calculated in any sub-model are not influenced by other sub-models. Each of the sub-models integrated all accepted dates from one geographical region representing an archaeological period/phase during which wheat started to appear in that region. Thus, the upper and lower boundaries of a modelled phase indicate the start and the end ‘event’ of a particular archaeological phase, respectively58. There is the necessity to introduce constraints to the lower boundaries in these sub-models. In China, the 14C dating technique has been less frequently applied to more recent historic periods than to prehistoric periods. This is owing to the fact that the age control of historic periods can be referred to written sources59. Although the existence of wheat should be increasingly visible following its gradual establishment as a staple in northern China5, the amount of available 14C dates directly derived from wheat remains may not necessarily increase accordingly. In other words, lower boundaries should not be allowed to extend without constraints into the historic period. Otherwise, the difference in research paradigm between prehistoric and historic chronologies may introduce a bias to the modelling results. Thus, we set up a consistent threshold value, 1000 bce, as a prior for the lower boundary of all sub-models. This threshold value was also used to rule out 14C dates with calibrated medians younger than 1000 bce. The value of 1000 bce was proposed and widely acknowledged in published sources25, as written records of wheat in China are available from oracle bones dating between the late Shang Dynasty (ca. 1300 bce) and the early Eastern Zhou Dynasty (ca. 700 bce). Furthermore, a short time lag date outlier OxCal command was applied to each of the accepted data points17, as, in a statistical sense, 1 out of 20 14C dates represents an outlier60. This command allowed for a more robust model. Because of the application of this outlier command, we did not manually remove 14C dates whose agreement index was lower than 60% from the model60. OxCal built-in order and difference commands were used to identify the relative order of all sub-models, which represented the relative order of wheat appearance in different geographical regions of China. To avoid unrealistically long-interval estimates, the difference command was set using a prior between −10,000 and 10,000 years.
Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.
The OxCal code in the study is available in Supplementary Information 1.
The 14C data first reported in this study are given in Table 2. The size measurement data of archaeological wheat grains reported in this study are given in Supplementary Table 2. The compiled 14C data set from literature sources and this study is given in Supplementary Table 3.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Research grants from the German Archaeological Institute, Gerda Henkel Stiftung (grant no. AZ 06/F/17), Deutsche Forschungsgemeinschaft (grant no. LE 3508/1-1), National Basic Research Program of China (grant no. 2015CB953803), Shandong University and the National Social Science Fund of China (grant no. 11AZD116, 12&ZD151 and 12&ZD194) are gratefully acknowledged. We appreciate support from Robert Spengler (Max Planck Institute for the Science of Human History) and Fengshi Luan (Shandong University). Thanks are also extended to colleagues at Poznan Radiocarbon Laboratory, Shandong Institute of Cultural Relics and Archaeology, and Yantai Museum.
About this article
Nature Plants (2018)