Introduction

Social complexity constitutes a crucial analytical framework in the study of early civilizations1. The extent of social sophistication exhibits significant variation across different regions and is influenced by diverse subsistence patterns2. These differences and complexities are evident in the interactions among individuals of different distinct cultural groups, social strata, and genders3.

The loess-desert transition zone in northern China primarily encompasses the sandy and wind-eroded regions of the Ordos Plateau and the sandy loess area of the Loess Plateau4. The administrative division includes central-southern Inner Mongolia, northern Shanxi Province, and Shaanxi Province. This area is distinguished by its limited water resources and ecologically fragile environment, which exhibits high sensitivity to environmental changes5. Despite these challenges, around 4000 BP, a remarkable civilization center emerged in this region, known as the Shimao Culture (石峁) (Fig. 1), and rapid population growth occurred6. The wide distribution of defensive facilities, such as stone-walled cities and ring trenches6, suggests high social complexity in the loess-desert transition zone during this period. Archaeological evidence suggests that settlements exhibited characteristics similar to those of the Shimao site7. The size of the settlements exhibited clear hierarchical differences: the Shimao site encompassed an area exceeding 4 million square meters, the sites had stone walls with areas of 500,000 to 1 million square meters, and some sites with stone walls were smaller than 500,000 square meters. Furthermore, numerous undefended settlements existed6,8, forming a hierarchical settlement system.

Fig. 1: Site locations and site pictures.
figure 1

a Location of Sharitala and related sites. This paper covers the following sites: 1. Shimao, 2. Bicun, 3. Zhaimao, 4. Zhaimaoliang, 5. Hengshan, 6. Sharitala, 7. Dakou, 8. Yadi, 9. Zhukaigou, 10. Shangyuan, 11. Shangshen, 12. Shengedaliang, 13. Jijiazhuang, 14. Muzhuzhuliang, 15. Xinzhang, 16. Huoshiliang. Other relevant sites: 17. Houchengzui, 18. Taosi, 19. Lushanmao, 20. Xiajiadian. The red star represents Beijing. The blue dotted line indicates the 400 mm rainfall contours. The red circle highlights ranges with similar cultural profiles7. The gray shading delineates the loess-desert transition zone. b The remains of house F1 at the Sharitala site in the 2023 excavation. c The burial remains M4 at the Sharitala site in the 2023 excavation. d, e The modern environment around the Sharitala site.

Full size image

The pattern of livelihood serves as the foundation for social development9. There is a significant correlation between the degree of agricultural intensification and the complexity of societal structures10. Relatively similar pottery assemblages were excavated from the Shimao site and nearby sites7, suggesting that the dietary structure and eating habits in the region were likely consistent11. The livelihood mode of occupants in the region was based on millet agriculture, complemented by the raising of pigs(Sus domesticus Linnaeus,1758), cattle(Bos taurus Linnaeus,1758), and sheep/goat(Ovis aries Linnaeus,1758/Capra hircus Linnaeus,1758)12,13,14,15,16,17,18,19,20,21,22,23,24,25,26.

Before the Shimao culture formed, occupants from south-central Inner Mongolia had already traveled southward into northern Shaanxi27, and the development of the archaeological culture and subsistence mode exemplified by Shimao in the loess-desert transition zone region was closely related to occupants from south-central Inner Mongolia. Previous research on subsistence practices in this region has predominantly focused on northern Shaanxi, with relatively limited attention devoted to the south-central part of Inner Mongolia12,18,28,29,30,31. Additionally, limited studies have examined the variations in agricultural development across sites of different grades within this region. What are the similarities and differences in crop structure and field management between non-stone-walled sites and stone-walled sites? How do these similarities and differences influence the subsistence economy and the level of social complexity of a region’s societies? Investigating these questions is essential for understanding both the adaptive strategies of human populations to harsh climatic conditions and the development of social structures and complexity within the region.

In this study, integrated bioarchaeological methods were applied to provide new evidence from the Sharitala site, to examine the subsistence strategies in Inner Mongolia, a key location in the loess-desert transition zone dating to approximately 4000 BP. Additionally, flotation and stable isotope published results from plants, animals, and humans across 15 sites within the region spanning 4200–3700 BP were synthesized to provide robust scientific evidence for understanding the relationship between the agricultural characteristics and the regional social structure during this remarkable period, with a significant increase in social complexity.

Methods

Sites and archaeological background

The Sharitala site (109°52′E,39°25′N, 1364 a.s.l) is located in the village of Sharitala, Ejin Horo Banner, Ordos city, Inner Mongolia Autonomous Region, in south-central Inner Mongolia. The site is approximately 45 kilometers southwest of the Zhukaigou site, 100 kilometers northwest of the Shimao site, 20 kilometers west of the Ulan Moron River, and 120 kilometers south of the Yellow River. This region is characterized by a temperate continental climate, marked by aridity, limited water resources, and high sensitivity to climatic fluctuations5. According to meteorological data from 1970 to 2009 in Ejin Horo Banner, the average annual precipitation was approximately 345 mm. The mean summer temperature was 20.5 °C, while the mean winter temperature was −6.3 °C. The Sharitala site is located within the loess-desert transition zone, classified as a temperate semiarid grassland characterized by steppe vegetation.

The Sharitala site covers an area of approximately 500,000 square meters. During the 2023 excavation, which covered an area of 500 square meters, archaeologists uncovered five house sites, four ash pits, six burials, three ash ditches, and one pedestal. Artifacts, including pottery shards, stone tools, bone tools, and animal remains (e.g., pigs, dogs, cattle, sheep/goats, and shellfish) were also discovered32. Based on the cultural remains excavated, archaeologists have determined the Sharitala site as belonging to the Dakou II culture, marking a critical transition phase from the late Neolithic period to the Bronze Age. The stratigraphic layers at the site are relatively thin, and all identified remains appear contemporaneous based on the artifacts recovered.

The sample about the white-plaster surface of the walls was analyzed for AMS-14C dating at the BETA laboratory in Miami, Florida, USA. The charcoal samples were analyzed for AMS-14C dating at the 14C laboratory of the Center for Scientific Archaeology, Institute of Archaeology, Chinese Academy of Social Sciences. All conventional 14C ages were calibrated to calendar years using OxCal 4.4.4 software33,34 with the IntCal20 calibration curve35. The samples from F1 and F2 date to 4220–3850 cal BP and 4070–3726 cal BP, respectively(Table 1).

Table 1 Sharitala site dating results
Full size table

Archaeobotanical analysis

A total of 24 flotation samples were collected from the Sharitala site, including 8 samples from stove pits (Z), 7 samples from ash ditches (G), and 9 samples from house sites (F, which refers to the area of the house sites except for the stove pits). The cumulative volume of these flotation samples was 244 liters, averaging about 10 L per sample. The sampling points covered almost all excavation areas, and the sampling units covered all types of remains found at the site.

The botanical remains were retrieved through flotation in small buckets using an 80-mesh sieve (0.2 mm aperture). The samples were dried and forwarded to the Department of Archaeology and Anthropology of the University of Chinese Academy of Sciences for identification and analysis. The charred plant remains were categorized into charcoal fragments and seeds. Seeds were meticulously selected and identified under a Leica S9D stereo microscope. For seed counting, only those that were intact and clearly identifiable in the embryo area were included, while other seed fragments were excluded. The precision of the balance used was 0.01 mg. Absolute counts and the frequency of occurrence statistics served as the primary research indicators. To account for variations in seed weight, a weight correction model was applied to ensure more reliable comparisons of relative seed yields36.

The formula for calculating the weight percentage of foxtail millet (Setaria italica L.) is as follows:

$${{\rm{P}}}_{({\rm{foxtail\; millet}})}=\frac{{{\rm{N}}}_{{\rm{foxtail\; millet}}}{\rm{\times }}2.6}{{{\rm{N}}}_{{\rm{foxtail\; millet}}}{\rm{\times }}2.6+{{\rm{N}}}_{{\rm{common\; millet}}}\times 7.5}$$

The formula for calculating the weight percentage of common millet (Panicum miliaceum L.) is as follows:

$${{\rm{P}}}_{{\rm{common\; millet}}}=\frac{{{\rm{N}}}_{{\rm{common\; millet}}}\times 7.5}{{{\rm{N}}}_{{\rm{foxtail\; millet}}}{\rm{\times }}2.6+{{\rm{N}}}_{{\rm{common\; millet}}}\times 7.5}$$

Stable isotope analysis

Stable carbon and nitrogen isotope analysis of plants. The quantity of common millet unearthed from the same unit was limited, so only a portion of the foxtail millet was selected for the study of plant carbon and nitrogen stable isotopes. Charred foxtail millet with clean and intact surfaces was meticulously selected and photographed under a microscope to record the state of the seeds before treatment. Both with and without husk samples were analyzed to evaluate potential differences between them. For each sample, approximately 10 mg of intact foxtail millet grains with clearly defined morphology were selected for mixing and grinding. The samples were pretreated by the acid treatment method37: the samples were placed in test tubes, 10 ml of 0.5 mol/L HCl was added, and the mixture was heated in a water bath at 80 °C for 30–60 min. The solution was poured off and washed with deionized water until neutral. The samples were lyophilized, ground, and tested. The sample information is shown in Table 3.

Stable carbon and nitrogen isotope analysis of animals and humans. Preparation and testing methods of bone collagen are as follows: The inner and outer surfaces of the bone samples were polished to remove contaminants. Then, the samples were soaked in 0.5 mol/L HCl solution at 4 °C, and the acid solution was replaced every two days until the bone samples were soft and there were no obvious air bubbles on the surface of the bone samples. The bone samples were washed with deionized water until neutral, 0.125 mol/L NaOH solution was added, and the samples were soaked at 4 °C for 20 h and then rewashed to neutral. The bone samples were immersed in 0.001 mol/L HCl solution and heated at 70 °C for 48 h. After the bone samples were completely dissolved in the solution, they were filtered into hot test tubes to obtain the collagen solution. The collagen solution was freeze-dried to obtain collagen38,39. Detailed sampling information is provided in Table 4.

Isotope samples of plants, animals, and humans were tested at the Stable Isotope Archaeology Laboratory, Department of Archaeology and Anthropology, University of Chinese Academy of Sciences. The analysis was conducted using an elemental analyzer (Vario Pyro Cube) coupled with an isotope mass spectrometer (IsoPrime-100 IRMS). Two-point calibration was used to calibrate the raw isotope values. Carbon and nitrogen isotopic compositions were measured and calibrated to VPDB and AIR with USGS40 (δ13C = −26.39‰, δ15N = −4.52‰) and USGS41 (δ13C = 37.63‰, δ15N = 47.57‰). The laboratory-prepared collagen standard sample (δ13C = −14.7 ± 0.2‰, δ15N = 7.0 ± 0.2‰) was included within the sample sequence to monitor instrumental stability throughout the analytical process. Stable isotope abundances of carbon and nitrogen were converted to δ values to reflect isotopic variation, expressed in per mille. The precision for C and N was determined to ±0.2‰ of calibration standards and check standards.

The method for establishing the isotopic baseline of plants. Plant nitrogen isotopes are influenced by various external environmental factors, including soil properties, atmospheric CO2 concentration, nitrogen deposition, and precipitation40,41. By comparing the δ15N values of charred seeds with the local δ15N isotope baseline, researchers can evaluate the degree to which crop δ15N values reflect human field management practices, independent of natural environmental influences on plant seed nitrogen isotopes42. The baseline for wild vegetation isotopes is typically established using nitrogen stable isotope measurements from wild herbivores, with the isotopic enrichment associated with trophic level elevation being subtracted. As no carbon and nitrogen stable isotope data are currently available for wildlife at the Sharitala site, this study uses nitrogen stable isotope values from wild animals at the Wuzhuangguoliang site as a proxy(Wuzhuangguoliang Site is also located in the desert loess transitional zone).

The average δ15N value of the wild grass rabbit unearthed from the Wuzhuangguoliang site in northern Shaanxi is 4.9‰(n = 3, SD = 0.1‰)43. Considering that the value of δ15N enrichment at each trophic level can be between 3 and 5‰44, the δ15N value of the local wild vegetation at the Wuzhuangguoliang site and the neighboring area would be around −0.1‰ to 1.9‰.

Data analysis methods. This study draws on established statistical methods from prior archaeological research45,46. Given that the isotope data for plants, animals, and human bones involved in this study are based on relatively small sample sizes, this study interprets the results more rigorously by not only comparing mean values but also incorporating a comprehensive analysis of the 95% confidence intervals, which are indicated in square brackets throughout the text. To prevent over interpretation of limited data, descriptive statistical methods were prioritized in the analytical approach. All statistical analyses were conducted using SPSS 27.0 software, including the calculation of mean values, estimation of confidence intervals, and performance of t-tests for significance testing.

Results

The flotation results of the Sharitala site

The total number of carbonized plant seeds obtained from flotation was 592, including four genera(see Table 2 and Supplementary Material Table 2 for details). Most of the plant remains were crops (99.5%), including foxtail millet and common millet. There were 538 foxtail millet seeds and 51 common millet seeds, and the absolute number ratio of foxtail millet to common millet was 10.5. The probability of emergence(ubiquity) was 62.5% for foxtail millet and 33.3% for common millet. The weight correction of the number of seeds by the thousand-kernel weight of the crop seed gave a weight-percentage ratio of 3.7 for foxtail and common millet (Fig. 2).

Fig. 2
figure 2

Proportions of millet at the Sharitala site.

Full size image
Table 2 The flotation results of the Sharitala site
Full size table

The plant remains were highly charred and black, and some of the charred foxtail millet had formed solid aggregates. Approximately 141 specimens of foxtail millet retained their husk and exhibited irregular granular protrusions on their surfaces. The ubiquity of foxtail millet with husk was 25%, accounting for 23.8% of the total excavated plant seeds. Ninety percent of the foxtail millet with husk was excavated in the hearth of the house site.

The types and quantities of non-agricultural crop remains were relatively small, comprising only two species of Fabaceae plants, Lespedeza bicolor L. and Melilotus officinalis L., which accounted for 0.2% and 0.3% of the total excavated plant seeds, with ubiquity of 4.2% and 8.3%, respectively. Images of these seeds are presented in Fig. 3.

Fig. 3: Charred seeds from the Sharitala site.
figure 3

1. Setaria italica; 2. Panicum miliaceum; 3. Lespedeza bicolor; 4. Melilotus officinalis; 5-6. Foxtail millet with husk.

Full size image

The charcoal recovered from the Sharitala site is entirely in fragmented form. According to the size of the charcoal fragments, they are categorized into the following size classes: >10 mm, 4–10 mm, 1–4 mm, and <1 mm. The total weight of the charcoal was 5.682 g, and the average concentration of carbonized wood charcoal was 0.0233 g/L. The content of wood charcoal varied among flotation units, and the wood charcoal in the soil sample from hearths at the site was 1.6 g.

Stable carbon and nitrogen isotopes of the plants

The relevant data are presented in Table 3. The δ13C values of charred foxtail millet ranged from −9.3‰ to −8.8‰, with an average of −9.0 ± 0.2‰[−9.2, −8.8] (n = 10). The δ15N values ranged from 4.6‰ to 5.9‰, averaging 5.1 ± 0.4‰[4.8, 5.4] (n = 10). Considering the effect of charring on seed δ15N values, 0.3‰ was subtracted to adjust the δ15N values of charred foxtail millet47. After correction for charring, the δ15N values of foxtail millet ranged from 4.3‰ to 5.6‰[4.5, 5.1](see Fig. 4).

Fig. 4
figure 4

Scatter plot of isotopic data from humans, animals, and plants at Sharitala.

Full size image
Table 3 δ13C and δ15N(uncorrected) values of archaeobotanical samples measured at the Sharitala site
Full size table

The mean δ13C and δ15N values of hulled millet were both lower than those of unhulled millet. Independent samples t-tests further revealed that there was no significant difference in the δ15N values between hulled millet (5.0 ± 0.4‰, n = 4)[4.3, 5.7] and unhulled millet (5.1 ± 0.4‰, n = 6) [4.6, 5.6](t(8) = −0.254, P = 0.806 > 0.05). In contrast, a statistically significant difference was observed in the δ13C values between hulled millet (−9.3 ± 0.1‰, n = 4)[−9.3, −9.2] and unhulled millet (−8.8 ± 0.1‰, n = 6) [−8.9, −8.8](t(8) = −11.952, P < 0.01).

Stable carbon and nitrogen isotopes of humans and animals

As shown in Table 4, collagen was successfully extracted from all samples except samples S5 and S15. The carbon content of the collagen ranged from 15.3% to 47.0%, while the nitrogen content varied between 5.5% and 17.3%. The C/N molar ratios were within the range of 2.9–3.648,49, so the samples could be regarded as uncontaminated. Although the C content of sample S7 was slightly higher than the standard value, it had a normal C/N molar ratio, so it was viewed as standard data and was not excluded. The remaining samples met the standard and could be used for subsequent stable carbon and nitrogen isotope analysis.

Table 4 δ13C and δ15N values of humans and animals measured at the Sharitala site
Full size table

Data from the long bones and pelvic bones are used in the discussion that follows. The δ13C values of humans ranged from −8.7‰ to −6.2‰ with a mean value of −7.1 ± 1.0‰[−8.4, −5.8] (n = 5). The δ15N values of humans ranged from 6.9‰ to 10.1‰ with a mean value of 8.9 ± 1.1‰ [7.3, 10.5](n = 5). The distribution of δ15N in bone collagen varied, with a range of 3.2‰ between the minimum and maximum values.

The carbon and nitrogen isotope values for pig were −8.1‰ (δ13C, n = 1) and 7.2‰ (δ15N, n = 1), respectively. The carbon and nitrogen isotope values for the dog were −8.2‰ (δ13C, n = 1) and 7.9‰(δ15N, n = 1), respectively. The carbon and nitrogen isotope values for cattle were -14.4‰ (δ13C, n = 1) and 7.0‰(δ15N, n = 1). The carbon and nitrogen isotope values for sheep/goat were −15.6‰ (δ13C, n = 1) and 6.0‰ (δ15N, n = 1).

Discussion

The stable carbon and nitrogen isotope values of human bones provide insights into the dietary patterns of occupants50. The mean value of δ13C of humans in Sharitala is −7.1 ± 1.0‰[−8.4, −5.8] (n = 5), which indicates that humans mainly consumed C4 food (including C4 plants and C4 plant-dominated animals) during their lifetime. Since the Neolithic period, foxtail millet and common millet of C4 were the main crops consumed by humans in this area31,51, and the plant seeds excavated at the Sharitala site were mainly foxtail millet and common millet, it is reasonable to conclude that humans of Sharitala should have used millet or millet-fed animals as their staple food.

In previous studies, human δ15N values have been widely used to infer the consumption levels of animal protein during an individual’s lifetime52. However, recent research suggests that the proportion of animal protein in prehistoric human diets may have been overestimated53. Since the Neolithic period, inhabitants of northern China began applying organic fertilizers to foxtail millet and common millet, a practice that has been shown to elevate the δ15N values of these staple grains54. The average δ15N value of occupants from the Sharitala site is 8.9 ± 1.1‰[7.3, 10.5], indicating that plant-based resources played a significant role in their diet. Therefore, when analyzing the nitrogen stable isotope composition in human bone collagen, it is essential to account for the isotopic contribution of foxtail millet, which has a mean δ15N value of 4.8 ± 0.4‰[4.5, 5.1]. Assuming a trophic level enrichment factor of approximately 4‰44, the expected δ15N values for occupants consuming only foxtail millet would fall within the range of 8.5‰ to 9.1‰. However, some occupants from the Sharitala site exhibit δ15N values above this interval, suggesting that their diet included millet as a staple and a notable amount of animal protein.

Considering the bones of pigs, cattle, and sheep/goats excavated from the site, it was plausible that the sources of animal protein in Sharitala occupants may have been pigs, cattle, and sheep/goats. The δ15N values of humans in this study showed no clustering of data, and the difference between the lowest and highest values exceeded 3‰, which is the minimum enrichment typically associated with a single trophic level. This suggests considerable variation in animal protein consumption within the group. The presence of different sizes of burials and jade artifacts55, implies that social stratification was already evident in the Sharitala community at that time and that the uneven distribution of meat resources may be closely associated with this hierarchical structure.

The δ13C values of pig and dog from the Sharitala site were close to humans, suggesting that C4 food constituted a significant portion of their diets. The δ15N values of dogs are similar to those of humans, indicating comparable levels of protein consumption between the two. Dogs were closely related to humans, and humans provided food scraps for dogs56; thus, dogs at the Sharitala site may have consumed human food leftovers.

The δ13C value of −18.0‰ is used as a criterion for distinguishing between C3 and C3/C4 mixed feeding57. The wild vegetation in southern-central Inner Mongolia is dominated by C3 plants58, and the dietary structure of completely free-ranging herbivores should theoretically show a bias toward C3 characteristics. Grazing on C3 plants supplemented by some foddering with C4 (millet) is indicated in the δ13C values. The dietary structure of cattle and sheep/goat suggested that occupants used grazing as the mainstay of their diet, supplemented by foddering.

Houses, burials, ash ditches, and a platform were found at the Sharitala site, indicating that occupants had a relatively stable settled life55. Agricultural tools and a large number of daily-use pottery were found, suggesting that agricultural production existed at the site around 4000BP. As illustrated in Fig. 1d, e, the terrain surrounding the Sharitala site is predominantly flat, thereby offering favorable conditions for agricultural activities.

The density of charred seeds from the Sharitala site was approximately 2 grains/L, which was relatively low compared with that of the surrounding sites59, and the diversity of plant species was limited. The low density may indicate that the area where the flotation unit is located is at an advanced stage of crop processing. The plant seeds were mostly foxtail millet and common millet, and Sharitala occupants used the C4 type of millet as their staple food. The probability, density, and absolute quantity of foxtail millet were greater than those of common millet, and occupants utilized foxtail millet more than common millet.

A large number of husked millet were recovered from the site, posing interesting questions about the use of hulled foxtail millet. This study analyzed the stable carbon and nitrogen isotopes of both foxtail millet with husk and foxtail millet without husk to compare their differences and explore the reasons for the prevalence of foxtail millet with husk at the site. The δ15N values of foxtail millet with[4.3, 5.7] and without husk[4.6, 5.6] are very close. The δ13C value difference between foxtail millet with husk and foxtail millet without husk is 0.44‰. This relatively minor difference may be attributed to the compositional distinctions between the husk and the seed tissues60. Therefore, stable carbon and nitrogen isotope analysis did not provide a satisfying answer to this question.

The content of charcoal chips and the morphology of the foxtail millet provide ideas for understanding the large amount of foxtail millet with husk excavated at the site. If the Sharitala community used wood fuel, the charcoal debris on the hearth would be abundant14. The charcoal content of the hearth was lower(1.6 g, 28.1%) than that of the other units at Sharitala. Approximately 90% of the foxtail millet with husk at the site was found mainly in the hearth. The contents of foxtail millet flow out after being subjected to rapid heat, forming tar that causes the seeds to stick and clump together61. Foxtail millet with husk exhibiting a high degree of charring and clumping was discovered at Sharitala, indicative of exposure to relatively high temperatures. At the Taosi site, pottery kilns utilized crop by-products as fuel62. Similarly, the Sharitala community likely used crop by-products such as threshing straws and husks as fuel. The charring of foxtail millet remaining on these straws has been preserved to the present day. Pollen analysis of the study area at approximately 4000 BP revealed a predominance of herbaceous pollen over woody pollen63, suggesting that tall trees were scarce in the vicinity of the Sharitala site at that time. Therefore, it was plausible that occupants relied on crop by-products for fuel.

In the northern regions of China, the fertilization behavior can be traced back to the middle to late Neolithic period54. Fertilizer application has the potential to influence both soil and crop δ15N values64. The δ15N values of the carbonized, corrected charred foxtail millet from Sharitala ranged from 4.3‰ to 5.6‰, which were higher than the δ15N baseline of the wild vegetation. The average values of millets suggest fertilizers were likely used during cultivation. It should be noted that due to the small sample size, the effectiveness of the statistical analysis is limited.

The δ15N values of the carbonized foxtail millet from Sharitala are greater than those of the unfertilized foxtail millet grains from the Lintong planting experiment(2.9‰ to 3.7‰)65. This suggests that the millet excavated from the Sharitala site may have been treated with fertilizers during the planting period. Fertilizing with sheep/goat manure for one year increased the δ15N values of foxtail millet by 0.8‰66. Thus, it is plausible that the Sharitala occupants fertilized their farmland for many years. The presence of cattle and sheep/goat at the site would have facilitated this practice.

Furthermore, the nitrogen stable isotope values of cattle and sheep/goat might have been influenced by the indirect effect of foxtail millet and common millet fertilization. Similarly, the nitrogen stable isotope values of the people who consumed these animals and millet would also be affected by the same factors.

In summary, during the 4000 BP period, the Sharitala community was engaged mainly in millet agriculture, complemented by animal husbandry involving pigs, cattle, and sheep/goats. The discovery of fishing and hunting tools at the site suggests that occupants likely participated in fishing and hunting activities to supplement their diet. Occupants had a certain level of farmland management and fertilized cultivated land. The utilization of plant resources was relatively limited, focusing on foxtail millet rather than common millet, and crop byproducts were used as fuel. Livestock management incorporated both free-range and confined breeding practices for cattle and sheep/goats.

To assess the relationship between social complexity and agricultural strategies in the surrounding loess-desert transition zone at 4000 BP, we integrated our data from Sharitala with previously published isotopic and archaeobotanical /archaeozoological data from 15 additional sites. We employed the presence or absence of stone walls as a proxy for social complexity. The stone-walled sites include Shimao, Bicun, Zhaimao, Zhaimaoliang, etc (see Supplementary Material Table 1 for details).

By 4000 BP, social complexity within this region had intensified, leading to the formation of a multi-level settlement system. Within this hierarchy, stone-walled sites functioned as central settlements relative to nearby regional sites. Archaeologically, the scale and grade of these stone-walled sites were demonstrably higher than those of non-stone-walled sites, reflecting their greater social complexity67. The substantial labor and material resources required for stone wall construction indicate both the high cost of these structures and the special status held by these walled sites.

This hierarchical differentiation between sites may have influenced occupants’ choices in agricultural strategy. For example, at 2400 BP, the diversity of botanical remains was greater at high-status sites than at non-stone-walled sites in the Ali region of Tibet. Occupants at large sites had a more flexible agricultural strategy68. At 2500 BC, higher-ranked sites in southeastern Shandong demonstrated a preference for C3 crops than non-stone-walled sites did69. Similarly, late prehistoric societies in Ohio consumed more maize and had a greater number of kernels per cob than general settlements by AD 100010.

From the point of view of the composition of crops unearthed, the stone-walled sites Shimao site unearthed foxtail millet, common millet, rice, soybeans, and other varieties of plants. The Bicun site also yielded rice remains, and the crop composition is relatively diverse, whereas the other sites contained only two millet varieties, resulting in a relatively limited diversity of crops70. Rice may have been used for brewing71, which explains its presence primarily at stone-walled sites. Given that the region’s environmental conditions are not conducive to rice cultivation, the quantity of rice remains excavated is very low. The study area is in a semiarid region72,73. Foxtail millet and common millet are suitable for growth in arid environments because of their high water-use efficiency74. Foxtail millet and common millet have a long history of cultivation in northern China75, leading to a preference for millet cultivation in this area.

In 4200BP, the occupants’ preference for millet can be observed in absolute numbers and weight percentages. Across all archaeological sites examined, the quantity of foxtail millet consistently exceeded that of common millet. Considering the different millet yields, comparisons were made after adjusting the weights of the millet seeds36. The weight percentages of foxtail millet and common millet at the Shimao, Bicun, Zhaimao, Hengshan, Sharitala, Yadi, and Shangshen sites were greater than 1, indicating that the total production of foxtail millet at these sites likely surpassed that of common millet. Overall, inhabitants of the stone-walled settlements demonstrated a greater utilization of foxtail millet compared to common millet.

Factors such as the water and temperature required for crop growth, ease of cultivation, and yield of the crop would have influenced the choice of crops by occupants. Studies indicate that foxtail millet exhibits lower drought resistance and water use efficiency compared to common millet76. Its growing period is longer than that of common millet77, and its planting conditions are more complex20. Consequently, cultivating foxtail millet requires significant labor and material resources. In contrast, inhabitants at non-stone-walled sites favored the more easily cultivated common millet.

Under optimal water and fertilizer conditions, the productivity of common millet is lower than that of foxtail millet78, and the yield of foxtail millet can be twice as high as that of common millet79. Shimao, Bicun, and Zhaimao, the stone-walled sites, featured extensive settlement areas and substantial populations. Within these settlements, there was a division of labor80, allowing a dedicated group to focus on agricultural cultivation and crop management81. Owing to their substantial population sizes, these sites required a larger food supply, leading them to prioritize higher-yielding crops such as foxtail millet. The non-stone-walled site, Sharitala, has yielded a relatively large number of jade artifacts55, suggesting that this site was relatively high status and that occupants utilized foxtail millet more than common millet did.

At the Miaoliang Site, located in the desert loess transition zone, the weight percentage of foxtail millet and common millet recovered from the site at 5000 BP (0.7%) was significantly lower than that recovered at 4500 BP (10.6%)18,82. This suggests that foxtail millet may have gradually gained importance in the daily lives of the occupants of Miaoliang Site. Additionally, the Guanzhong and Central Plains regions demonstrated a gradual upward trend in the percentage of foxtail millet weight83. Consequently, the relatively high importance of foxtail millet in the stone-walled settlements and the Sharitala site might, to some extent, have been influenced by neighboring regions.

The δ15N values of millet samples excavated from the Shimao, Sharitala, Jijiazhuang, Xinzhang, and Huoshiliang sites were greater than those of the local wild vegetation isotope baseline. These findings suggest that these crops were fertilized to some extent during cultivation.

First, we examined the soil fertility of stone-walled sites. The lowest δ15N values were observed at the Shimao site (Fig. 5c), indicating relatively poor soil fertility in the cultivation areas. Despite the substantial amount of meat resources available at the stone-walled sites of Shimao and Bicun84, the δ15N values for humans and animals at these sites were lower than those at non-stone-walled sites(Fig. 5c, d). This discrepancy may be attributed to the lower δ15N values of the millet cultivated at these sites, suggesting relatively poorer soil fertility during millet cultivation.

Fig. 5: Data graph of carbon and nitrogen stable isotopes of plants, animals, and human bones, as well as animal and plant remains from the 4200BP site in the desert loess transition zone.
figure 5

a Plant remains excavated from sites in the loess-desert transition zone: 1. Shimao 2. Bicun 3. Zhaimao 4. Zhaimaoliang 5. Hengshan 6. Sharitala 7. Dakou 8. Yadi 9. Zhukaigou 10. Shangyuan 11. Shangshen 12. Shengedaliang 13. Jijiazhuang 14. Muzhuzhuliang 15. Xinzhang 16. Huoshiliang (see Supplementary Material Table 3 for details). b Percentage of animal remains in the loess-desert transition zone area: P: Pig, C: Cattle, S: Sheep/goat, W: Wild animals(see Supplementary Material Table 5 for details). c Isotopic data from millet at sites in the loess-desert transition zone(see Supplementary Material Table 4 for details). c-1 The scattered points represent the δ15N values of the plants, and the upper and lower edges of the black box plot indicate the boundaries of the 95% confidence interval range of the data, and black points are the mean value. c-2, c-3 Error bars represent 95% confidence intervals. d Box-plot of isotopic data from animals at sites in the loess-desert transition zone(see Supplementary Material Table 4 for details). The upper and lower edges of the black box plot indicate the boundaries of the 95% confidence interval range of the data, and black points are the mean value. e Box-plot of isotopic data from humans at sites in the loess-desert transition zone: 1. Shimao 2. Bicun 3. Sharitala 4. Dakou 5. Shengedaliang 6. Muzhuzhuliang 7. Jijiazhuang 8. Xinzhang(see Supplementary Material Table 4 for details).

Full size image

Continuous cultivation of cereal crops leads to a significant decrease in the soil nitrogen content85. Long-term crop succession and increased aridity have resulted in a gradual decline in soil fertility in Mesopotamia over thousands of years86. The δ15N values of charred plants at sites surrounding the Shimao site have gradually declined21, which may also be attributed to prolonged crop succession. The manpower and material resources required for field management are extremely large. However, such stone-walled sites as Shimao and Bicun are unlikely to lack resources for fertilization and farming. Therefore, we considered the impact of continuous cropping on agricultural practices at these sites. The cultural layer thickness at Shimao ranges from 108 to 168 cm87, suggesting prolonged human occupation. The large size of the settlements and the substantial populations at Shimao and Bicun likely led to over-exploitation of the land and degradation of soil fertility.

The cultural layer at the Huoshiliang site measures approximately 0.5 m in thickness, and excavators believe that it represents a seasonal camp20. The high δ15N value of the charred common millet excavated from the Huoshiliang site may be associated with this seasonal occupation pattern. The δ15N values of millet from the Sharitala, Xinzhang, and Jijiazhuang sites were similar (Fig. 5c). Additionally, the δ15N values of millet samples from the Sharitala, Xinzhang, and Jijiazhuang sites were similar (Fig. 5c), indicating comparable levels of soil fertility and human fertilizer input across these locations.

The variability in crop δ15N values at the same site can be an approximate indicator of centralized crop management during cultivation. Under identical planting conditions, the δ15N values of foxtail millet tend to be relatively consistent, while common millet shows greater variability, with differences of less than 3‰65. Isotope samples from the Jijiazhuang, Xinzhang, and Huoshiliang sites were collected from different ash pits and ditches21,22, suggesting that these samples may originate from different household waste products. The relatively dispersed distribution of δ15N values for millet at these three sites indicates nonuniform fertilization practices, implying that individual households likely managed small plots of arable land.

The isotopic samples of charred foxtail millet from the Shimao site were collected from different units21. The relatively concentrated distribution of their δ15N values suggests that the foxtail millet was cultivated under consistent field management practices. The δ15N values of foxtail millet excavated from the Sharitala site indicate more uniform fertilizer management, possibly due to the higher status or organization level of the Sharitala site. At the Jijiazhuang site88 and the Xinzhang site89, immature foxtail millet was excavated, reflecting a lower level of farmland management. This finding suggests that, at some non-stone-walled sites, crop cultivation was not centrally managed, resulting in relatively low levels of agricultural management.

Subsequently, we discussed the different labor inputs used for millet cultivation. The weight percentages of foxtail millet and common millet excavated from Jijiazhuang (0.7) and Xinzhang (0.9) were lower than 1 (Fig. 5a), suggesting that occupants at these sites probably cultivated more common millet than foxtail millet. The δ15N values of common millet are greater than those of foxtail millet under identical fertilization conditions by over 2‰65. In terms of crop nitrogen stable isotopes, the mean δ15N values of foxtail millet excavated from the Jijiazhuang site and the Xinzhang site (4.9 ± 0.8‰, n = 10; 5.1 ± 1.0‰, n = 11) were greater than those of common millet (4.7 ± 0.8‰, n = 10; 4.1 ± 1.4‰, n = 4). At the Jijiazhuang Site, the calculated difference between common millet and foxtail millet is 95% CI [−1.0, 0.6]. At the Xinzhang Site, the difference is 95% CI [−2.4, 0.4]. This finding indicated that fertilizer management for foxtail millet at these two sites was more intensive than that for common millet. Although occupants made more use of common millet than foxtail millet, they invested more resources in foxtail millet in terms of fertilization. This may be attributed to the high-yield potential of foxtail millet and the high requirements of foxtail millet cultivation.

Baseline δ15N values can vary between regions. This study assumes a uniform isotope baseline for all sites, a simplification that may lead to some inaccuracies. However, cross-site comparisons are still essential to identify patterns in isotope data and understand the environmental and human factors behind them. Future research should focus on developing more precise, site-specific nitrogen isotope baselines to improve interpretation accuracy.

The combination of climate change and population growth significantly influences the development of agricultural strategies. At approximately 4000 BP, the loess-desert transition zone experienced a clear trend of drying and cooling, marked by a significant decline in annual precipitation63,90,91. A reduction in the average annual temperature of 1 °C resulted in a decrease in grain yield of approximately 10% per mu92 (Mu is a traditional Chinese unit of land area,1 ha = 15 mu), indicating that temperature fluctuations likely played a critical role in diminished grain production during this period.

A study of human remains from the Wuzhuangguoliang site in the desert-loess transition zone at 4500 BP revealed that the dental enamel of these individuals exhibited periodic developmental defects93. These defects are likely indicative of seasonal malnutrition caused by the one-crop-per-year farming system. Such seasonal malnutrition may reflect periodic shortages in food supply. By around 4200 BP, the region experienced an increase in sociocultural complexity, as evidenced by a rise in both the number and spatial extent of archaeological sites, suggesting a growing population94. However, agricultural technology developed at a relatively slow pace. From 4500 BP to 4200 BP, the region remained predominantly reliant on a single annual millet-based cropping system. Consequently, the increased population at 4200 BP may have exacerbated pre-existing food shortages, potentially leading to even more severe nutritional stress than in earlier periods. The large population of the region95, the high demand for millet by humans and animals (Fig. 5d, e), a drier and colder climate, and poorer soil conditions put more significant food pressure.

Agricultural intensification at different sites, including increased labor and resource inputs into the land and crop yields per unit of land area, was necessary to address the region’s relatively infertile natural environment and high food demand. Inhabitants at the stone-walled sites invested more labor in centralized crop management, likely applied fertilizers, and prioritized the cultivation of the more productive foxtail millet. In contrast, at the non-stone-walled sites, the soil fertility levels were relatively uniform. While inhabitants still applied fertilizers to enhance soil productivity, resources were allocated more heavily towards foxtail millet compared to common millet. The fertilization and centralized management of agricultural land reflected the advancement of the region’s agricultural economy, enabling societies to support larger populations and thereby contributing to societal expansion. The development of agricultural production and increased labor inputs into the land have accelerated the process of land privatization and intensified the region’s social complexity.

Nevertheless, agricultural production technology was insufficient to alleviate food pressure in the loess-desert transition zone. Over-exploitation of the land led to a decline in soil fertility at stone-walled sites, contributing to reduced millet yields. At non-stone-walled sites, the level of field fertilization for millet was inconsistent, possibly due to inadequate labor and fertilizer resources for centralized crop management. The privatization of arable land further exacerbated inconsistencies in field management. Collectively, constrained agricultural technologies, declining soil fertility, and limited crop productivity likely contributed to varying degrees of food pressure across multiple settlements.

Occupants adopted additional measures to cope with food pressure. Firstly, the utilization of Chenopodiaceae plants helped alleviate food shortages to some extent. The growing season of millet crops is short, and the time for planting is limited77; therefore, planting only millet crops is not a safe choice. As shown in Table 5, at the Zhaimaoliang14, Shimao59, and Jijiazhuang88 sites, significant quantities of concentrated Chenopodiaceae seeds were discovered, which may have been intentionally selected and cultivated. Most Chenopodiaceae species can be utilized as vegetables or crops, although they are generally less palatable. These plants can serve as an alternative to millet during shortages of agricultural produce96. The cultivation and storage of Chenopodiaceae by various social groups have mitigated the risk of crop failure.

Table 5 A table presenting Chenopodiaceae seed remains from the site in the desert-loess transition zone at 4200 BP
Full size table

The introduction of cattle and sheep/goat was also a good option for occupants to cope with pressure. Approximately 4000 BP, the loess-desert transition zone experienced a second wave of Bos Taurus introductions97. The environment in the region at approximately 4000 BP was typical grassland vegetation dominated by Artemisia/Chenopodiaceae63, which was favorable for cattle and sheep/goat husbandry. Archaeological evidence from this region includes a significant number of animal bones and bone tools, with cattle and sheep/goat remains accounting for more than 50% of all animal remains (Fig. 5b; Supplementary Material Table 5). Cattle and sheep/goat husbandry emerged as crucial pillars of the local economy27,98, leading to the establishment of a new economic model that replaced the previous reliance on pure millet agriculture.

Cattle and sheep/goats utilize mainly wild resources. The mixed C3/C4 diet of these animals reflected the combination of captive and grazing (Fig. 5d), which consumed less arable land and enhanced human utilization of wild resources. The integration of millet agriculture and animal husbandry constituted a relatively stable economic model, supporting the sustained operation of social systems.

Studies conducted in the eastern Tianshan of Xinjiang indicate that approximately 3000 years ago, there was a decline in soil fertility and reduced fertilizer application; at the same time, human communities became more dependent on nomadic practices99. This shift in subsistence strategy closely parallels the transformation observed in the loess-desert transition zone around 4200 years ago, where diminishing agricultural productivity similarly prompted early inhabitants to place greater emphasis on animal husbandry.

Overall, the relatively infertile soils of the loess-desert transition zone, the progressively cooler and drier climate at 4000 BP, and increasing population may have contributed to increased food supply pressure. In response, the inhabitants of the region adopted various strategies to mitigate these challenges.

First, specific agricultural strategies, including crop selection, labor inputs, and fertilizer inputs, were used. Enhanced field management through increased agricultural intervention aimed at increasing crop yields. These practices served as a social risk-buffering strategy, contributing to livelihood sustainability and social stability. The differences in agricultural strategies between stone-walled sites and non-stone-walled sites reflect the complexity of agricultural development during this period, which may have been influenced by environmental factors and limited social resources. The intensification of agricultural production and increased labor input likely accelerated the process of land privatization and deepened the region’s social complexity. The significance of millet increased over time, likely influenced by nearby regions.

In addition to adjusting their agricultural strategies, the inhabitants responded to food pressure by incorporating Chenopodiaceae into their diet and introducing livestock such as cattle and sheep/goats. This diversification of food sources facilitated a shift in the economic model from pure millet agriculture to one where millet agriculture remained the mainstay, supplemented by animal husbandry.