Radical change and dietary conservatism: Mixing model estimates of human diets along the Inner Asia and China’s mountain corridors

Abstract

Recent research has demonstrated that a series of mountains from the eastern Iranian Plateau to eastern Kazakhstan and to western China played a significant role in trans-Eurasian exchange during the third and second millennia BC. In close association with these mountain corridors, a number of southwestern Asian cereals, notably free threshing wheat and barley, moved eastward, and broomcorn millet, among other plant foods originating in China, moved westward. In this paper, we apply Bayesian stable isotope mixing models to published and newly obtained isotopic data in order to quantitatively estimate the contribution of different food resources to human diets, and we consider the complexity of human food strategies at both ends of these mountain corridors: southern Kazakhstan and the Hexi Corridor in western China. Our results contrast the rapid adoption of wheat and/or barley in the Hexi Corridor with the gradual, incremental adoption of millet in southern Kazakhstan during the second millennium BC.

Keywords

barley isotopes millet mountain corridor mixing model wheat

Introduction

In recent years, our knowledge and understanding of the nature of the connections among groups of people in the Bronze Age Eurasia has grown. Various studies consider not only the expansion of food technologies but also the spread of the ways that food was consumed in prehistory. That is, why human diets in some areas were subject to radical changes, while in other areas, people appear to have been conservative in their dietary habits (Boivin et al., 2012; Lightfoot et al., 2013; Liu and Jones, 2014; Liu et al., 2014). In the context of this discussion, the eastern sector of Central Asia is of particular interest. Research has demonstrated that a series of foothill sites from the eastern Iranian Plateau to eastern Kazakhstan form an ‘Inner Asian Mountain Corridor’ which played a significant role in trans-Eurasian exchange during the third and second millennia BC (Frachetti, 2012; Frachetti et al., 2010; Motuzaite Matuzeviciute et al., 2015b; Spengler, 2015; Spengler and Willcox, 2013; Spengler et al., 2014a, 2014b). Other scholars have extended this mountain corridor concept to consider broadly the foothills of the Tianshan Mountains, as well as the northern edges of the Tibetan Plateau, namely, the Hexi corridor, and the western loess plateau in China (Chen et al., 2014; Flad et al., 2010; Liu et al., 2014, in press; Zhang et al., 2015).

There is now much evidence that a series of connections between human populations were established through these mountain corridors since the Bronze Age. From the evidence of early horse management, metallurgy, and a range of associated artifacts, we can trace the origins of these connections to the latter part of the second millennium BC (Frachetti, 2008; Levine, 1999; Lin, 2015; Linduff and Mei, 2009; Mei, 2003; Rawson, 2013; Sherratt, 2005). Archaeobotanical evidence from cereal crops may be able to push the date of these connections back even further into at least the third millennium BC or earlier. This process has been referred as ‘food globalisation in prehistory’ (e.g. Jones et al., 2011, in press). In close association with these mountain corridors, a number of southwestern Asian cereals, notably free threshing wheat and barley, moved eastward, and broomcorn millet, among other plant foods originating in China, moved westward (see Barton and An, 2014; Liu et al., 2016; Motuzaite Matuzeviciute et al., 2013; Spengler et al., 2014b, for reviews). Both Chinese and southwest Asian cereals were documented in the late third millennium BC in Begash in east Kazakhstan (directly dated, 2461–2154 cal. BC), while a number of sites from west China provide similar evidence dated to the late third and early second millennia BC (2100–1800 cal. BC) (Frachetti et al., 2010; Liu et al., 2016).

While the archaeobotanical evidence provides locations, dates, and site contexts, it stops short of elucidating the significance of the trans-located crops in contemporary human food chains. However, stable carbon and nitrogen isotopic evidence allows us to quantify the dietary proportions of foodstuffs with different isotopic ratios. It so happens that two of the crops moving westward from China (broomcorn and foxtail millet) are C₄ plants that leave an isotopic signature throughout the food chain which is distinctive from that left by southwest Asian crops moving eastward (which are all C₃ plants). A number of recent isotope studies follow the movements in both directions by investigating possible human diets. These studies have demonstrated that the westward expansion of Chinese millets and the eastward movement of the Fertile Crescent cereals follow different isotopically detectable patterns (Lightfoot et al., 2014; Motuzaite Matuzeviciute et al., 2015a). Other studies have addressed questions such as why a new type of cereal could be adopted in a region with an already existing indigenous cereal cultivation tradition; and why a new type of cereal was welcomed in one region but rejected initially in another (Lightfoot et al., 2013; Liu et al., 2014). In this paper, we apply Bayesian stable isotope mixing models to previously published and newly obtained isotopic data in order to quantitatively estimate the contribution of different food resources to human diets. In the light of recent isotope studies, we consider the driving force(s) behind the expansions of these cereal crops, and we consider the complexity of human food strategies at both ends of the mountain corridors: southern Kazakhstan and the Hexi Corridor in northwest China.

Expansions of the Fertile Crescent crops and the Chinese millets

By c. 1500 BC, the geographical range of two of the southwest Asian crops, free threshing wheat and barley, and of one of the north China’s domesticates, broomcorn millet, extended from the Pacific to the Atlantic Ocean. The chronology, pathways, and driving forces of these expansions have been subjected to scholarly investigations (e.g. Jones et al., 2011).

The pattern of the eastern movement of wheat and barley from southwest Asia has recently become clear. Evidence for the cultivation and domestication of various wheat and barley species appears in southwest Asia from at least 10,000 years ago (e.g. Weiss and Zohary, 2011). They are recorded in western Central Asia by around 6000–5000 years ago (Harris, 2010; Miller, 2003) and in Pakistan by around 7000 years ago (Costantini, 1984; Meadow, 1996; Petrie, 2015). After these initial spreads, the subsequent, more extensive movement of wheat and barley appears to be restricted to the free threshing forms. Research in eastern Central Asia and western China shows that the mountain corridors played an important role in the expansion of free threshing wheat during the third millennium BC (see Liu et al., 2016, for a review).

Beyond Central Asia, we can distinguish two separate expansion sequences of free threshing wheat to the north and south of the Tibetan Plateau in the third and second millennia BC (Liu et al., 2016). In the north, the sequence runs from the eastern range of the Inner Asian Mountain Corridor through the Tianshan Mountains and the Hexi corridor. It then extends to the middle and lower reaches of the Yellow River. In the south, a second sequence may be inferred from the Indus and northwest India through South India and the Ganges region. The source of free threshing wheat in Tibet could be from either the northern or the southern corridor. The pathway by which barley spread east is less constrained. It is likely that the spread of barley was associated with the movement of free threshing wheat but perhaps barley spread at different rates along the northern and southern routes.

The earliest sites with broomcorn and foxtail millet in northern China before 5000 BC are patchily distributed along foothills at the margins of the loess plateau (e.g. Liu et al., 2009, 2015; Yang et al., 2012). The subsequent two millennia (5000–3000 BC) see a high concentration of millet sites on the loess plateau, particularly along the valleys of the Wei River and its tributaries (Wagner et al., 2012). Broomcorn millet expanded westward to east Kazakhstan (Frachetti et al., 2010) in the late third millennium BC. It remains unclear, however, how broomcorn millet got to Europe from Central Asia. Foxtail millet, however, expanded to southwest and southeast China during the third millennium BC. Evidence of foxtail millet is documented in Taiwan, Thailand, and India in the second millennium BC (Castillo, 2011; Webber, 1991, 2001). An enigmatic feature of the records for broomcorn millet is its earliest apparent occurrence on both sides of the Old World at the same time. During the sixth and fifth millennia BC, some 20 sites from Europe and the Caucasus report broomcorn millet identified to species level (Hunt et al., 2008). However, attempts to directly date some of these fifth or sixth millennium BC millets have shown that they date to the Late Bronze Age and later periods; therefore, the date at which Asian millets reached Europe remains a matter for enquiry and ratification (Motuzaite Matuzeviciute et al., 2013).

Dietary shifts in Kazakhstan and China

Recent works have revealed dietary patterns for the broader eastern Central Asia and western China during the Bronze and Iron Ages (Guan et al., 2007; Lightfoot et al., 2013, 2014; Liu et al., 2014; Ma et al., 2013a, 2013b; Motuzaite Matuzeviciute et al., 2015a; Murphy et al., 2013; Shishlina et al., 2012; Svyatko et al., 2013; Ventresca Miller et al., 2014; Zhang and Zhu, 2011; Zhang et al., 2010; Zhou and Garvie-Lok, 2015). Of particular interest to this paper, dietary changes are documented on both sides of the mountain corridors around 4000 years ago in southern Kazakhstan and in the Hexi Corridor of western China (Lightfoot et al., 2014; Liu et al., 2014; Motuzaite Matuzeviciute et al., 2015a). In seven cemetery sites studied by Liu et al. (2014), carbon isotope data cluster individual skeletons into two distinct groups. The individuals in one group correspond to a diet dominated by C₄ foods (possibly millet), while those in the other group correspond to a mixed diet containing C₃ and C₄ resources, most likely wheat and/or barley and millet and possibly animals fed on these crops. The first of these groups comprises all of the individuals from two sites dating to before c. 1900 BC. The second group comprises virtually all of the individuals from five sites dating to after 1900 BC. It would appear that all of the individuals sampled at each site shared a common dietary pattern. There are no evident ‘trail-blazers’ in the novel post-1900 BC pattern. Liu et al. (2014) inferred that this considerable agrarian shift was a ‘bottom-up’ change, rather than an elite-led process. The ‘tidal wave’ of western crops after 1900 BC clearly corresponded to an agricultural revolution in the Hexi corridor. It should be noted that these burials from the Hexi Corridor were from seemingly public cemeteries, presumably representing a cross section of society. This is different from the kurgans in Central Asia which may represent elite burials.

A similar dietary shift has been observed at the other end of the mountain corridors, that is, the southern part of Kazakhstan. Work performed by Motuzaite Matuzeviciute et al. (2015a) showed that C₄ plants became established food sources in southern Kazakhstan from the Middle–Late Bronze Age onward (c. 1800 BC) and were likely eaten directly by humans, while there is no such evidence in northern Kazakhstan. Given the archaeobotanical evidence from eastern Central Asia, it is reasonable to attribute the C₄ signal to the consumption of millet(s). Such evidence constitutes the earliest-to-date directly dated isotopic signals of millet consumption outside of China (Motuzaite Matuzeviciute et al., 2015a). The isotopic data from southern Kazakhstan indicate that while millet(s) was a significant dietary component for a number of individuals, and it is possible that they were not staple foods in the Bronze Age as they were in the Iron Age.

Carbon and nitrogen isotope values measured in bone collagen are primarily derived from dietary protein (Ambrose and Norr, 1993; Hare et al., 1991; Jim et al., 2006). When human consumption of animal protein is relatively high, collagen isotopic values likely fail to capture isotopic signatures from relatively low-protein, carbohydrate-rich cereal crops, such as millet, wheat and barley. However, in the context of naturally isotopically labeled food resources – for example, C₄-based carbohydrate source (millet) and C₃-based protein source (meat fed on C₃ plants) – consideration of the diversity of food inputs (both plant and animal) of a given community becomes plausible. In the following section, we apply the Bayesian stable isotope mixing model Stable Isotope Analysis in R (SIAR-R package ‘simmr’ version 0.2) to published and newly obtained δ¹³C and δ¹⁵N data from human and animal bone collagen and carbonized plant remains in order to quantify the contribution of different food protein resources to human diets at six Bronze Age and Iron Age sites in southern Kazakhstan and the Hexi Corridor in China.

Materials and methods

We sourced carbon and nitrogen isotope data for humans, animals, and plants from the literature. All data from Kazakhstan are previously published (Motuzaite Matuzeviciute et al., 2015a). Isotopic results from human specimens from China have also been published by Liu et al. (2014). Results of isotope analyses for Chinese animal and plant remains are published here for the first time. These samples are from three sites in Gansu province in west China: Xihetan, Huoshaogou, and Sanbadongzi. Procedures for sampling and analysis follow those described in Liu et al. (2014). Site information and isotope results are summarized in Tables 1 and 2. Site locations are shown in Figure 1.

Table 1.

Site information derived from Liu et al. (2014) and Motuzaite Matuzeviciute et al. (2015a).

Site	Region	Period	Cultural group	Age (years BC or AD)
Oi-Dzailau VII	Southern Kazakhstan	Late Bronze Age	Andronovo	1600–1400 BC^a (n = 2)
Kyzyl-Bulak	Southern Kazakhstan	Late Bronze Age	Andronovo	1750–1500 BC^a (n = 2)
Kainar-Bulak I	Southern Kazakhstan	Early Iron Age	Wusun or Hunic	200 BC–AD 300^a (n = 2)
Karatuma	Southern Kazakhstan	Early Iron Age	Wusun	400–100 BC^a (n = 3)
Huoshaogou	Western China	Bronze Age	Shanma or Siba	1900–1300^a (n = 8)
Ganguai	Western China	Bronze Age	Siba	1350–950^a (n = 1)
Sanbadongzi	Western China	Bronze Age	Siba	1500–1000^b
Wuba	Western China	Bronze Age	Banshan or Machang or Qijia or transitional type	2450–1950^a (n = 4)
Xihetan	Western China	Bronze Age	Qijia or Siba	2300–1000^b

n: number of radiocarbon dates.

Age intervals based on the radiocarbon determinations from human bones.

Ages based on cultural context.

Table 2.

Isotope and diet composition data used in modeling.

Site	Source	δ¹³C ‰	SD δ¹³C ‰	[C]	δ¹⁵N ‰	SD δ¹⁵N ‰	[N]	n
Oi-Dzailau VII
Oi-Dzailau VII	Humans	−15.6	1.8	N/A	13.8	0.6	N/A	11
Huoshaogou	Hordeum vulgare	−24.2	1.0	0.46	4.4	1.0	0.02	1
Huoshaogou	Panicum miliaceum	−9.9	1.0	0.47	9.4	1.0	0.02	1
Oi-Dzailau, Shimkent, Kyzyl-Bulak I	Sheep/goat	−22.9	0.5	0.61	7.9	1.5	0.09	3; 1 from each site
Kainar-Bulak I
Kainar-Bulak I	Humans	−12.6	1.4	N/A	12.1	1.4	N/A	17
Huoshaogou	Hordeum vulgare	−24.2	1.0	0.46	4.4	1.0	0.02	1
Huoshaogou	Panicum miliaceum	−9.9	1.0	0.47	9.4	1.0	0.02	1
Kainar-Bulak I	Dog	−20.3	2.5	0.60	12.3	1.5	0.11	2
Kainar-Bulak I	Domestic animals	−23.1	0.8	0.60	7.6	1.3	0.09	Sheep/goat = 11; Cattle Bos = 4; Horse = 4
Kainar-Bulak I	Elk	−23.1	1.0	0.54	5.9	1.0	0.14	1
Karatuma
Karatuma	Humans	−15.4	1.0	N/A	12.3	1.6	N/A	39
Huoshaogou	Hordeum vulgare	−24.2	1.0	0.46	4.4	1.0	0.02	Hordeum valgare = 1
Huoshaogou	Panicum miliaceum	−9.9	1.0	0.47	9.4	1.0	0.02	1
Kainar-Bulak I	Domestic animals	−22.78	0.97	0.60	8.12	1.42	0.09	Sheep/goat = 13; Cattle = 4
Wuba
Wuba	Humans	−7.4	0.9	N/A	9.2	1.1	N/A	56
Donghuishan	Triticum eastivum	−24.6	1.0	0.46	5.8	1.0	0.02	1
Donghuishan and Xihetan	Setaria italic or Panicum miliaceum	−10.1	0.4	0.47	7.9	1.2	0.02	Setaria italic = 2; Panicum miliaceum = 1
Xihetan	Dog	−15.7	2.8	0.60	10.0	1.2	0.11	7
Xihetan	Cattle	−22.2	1.1	0.61	5.7	0.8	0.09	7
Xihetan	Sheep/goat	−20.1	1.5	0.61	7.1	1.3	0.09	14
Xihetan	Elk	−22.5	0.3	0.54	3.3	1.0	0.14	5
Huoshaogou
Huoshaogou	Humans	−12.0	1.9	N/A	12.0	1.3	N/A	30
Huoshaogou	Triticum eastivum.	−20.3	1.0	0.46	8.0	1.0	0.02	1
Huoshaogou	Hordeum vulgare	−24.2	1.0	0.46	4.4	1.0	0.02	1
Huoshaogou	Panicum miliaceum	−9.9	1.0	0.47	9.4	1.0	0.02	1
Huoshaogou	Elk	−22.6	0.8	0.60	10.7	1.1	0.11	10
Huoshaogou	Domestic animals	−22.1	1.1	0.61	8.4	2.1	0.09	Sheep/goat = 41; Cattle = 28; Pig = 7; Horse = 4
Ganguai
Ganguai	Humans	−15.3	1.5	N/A	11.6	0.9	N/A	30
Huoshaogou	Triticum eastivu	−20.3	1.0	0.46	8.0	1.0	0.02	1
Huoshaogou	Hordeum vulgare	−24.2	1.0	0.46	4.4	1.0	0.02	1
Huoshaogou	Panicum miliaceum	−9.9	1.0	0.47	9.4	1.0	0.02	1
Sanbadongzi	Sheep, goat and elk	−21.7	0.9	0.59	6.6	1.6	0.10	Sheep/goat = 16; Elk = 8
Sanbadongzi	Pig and cattle	−21.1	1.5	0.63	9.2	1.1	0.07	Pig = 10; Cattle = 10

SD: standard deviation; n: number of samples.

We made the following assumptions about the isotope values of food sources and humans: (1) δ¹³C_muscle–δ¹³C_collagen=4‰, (2) δ¹³C_diet – δ¹³C_collagen=5‰, and (3) δ¹⁵N_diet – δ¹⁵N_collagen=5‰ (Oi-Dzailau) or 4‰ (all other sites). Digestible [C] and [N] were calculated using the US Department of Agriculture (USDA) Nutrient Database following Koch and Phillips (2002).

Figure 1.

Maps showing location and topography of the study region and the sites from Kazakhstan and China discussed in this paper (see Table 1 for detailed information).

To quantify the proportional contributions of various plant and animal food sources assimilated dietary protein of archaeological humans, we used SIAR-R package (‘simmr’ version 0.2; Parnell et al., 2010). We followed the best practices for stable isotope mixing models as delineated by Phillips and et al. (2014). SIAR is a Bayesian stable isotope mixing model that is capable of accounting for error in estimates of discrimination factors (i.e. the difference in isotopic composition between an individual’s tissue and diet: Δ¹³C_{diet-collagen} and Δ¹⁵N_{diet-collagen}), variation in source isotopic values, and different food source elemental concentrations of C and N.

When diets include both plant and animal resources, which have very different C and N concentrations, failure to employ a concentration-weighted model can significantly affect the results (Phillips and Koch, 2002). We calculated digestible [C] and [N] of human foods from the US Department of Agriculture (USDA) Nutrient Database following the calculations presented in Koch and Phillips (2002). To avoid a severely underdetermined model, we grouped logically related resources (e.g. domestic animals) a priori when their average isotope values were similar; for example, at Kainar-Bulak I, we combined sheep and/or goat (n = 11; δ¹³C: −23.2 ± 0.6‰, δ¹⁵N: 7.7 ± 1.2‰), cattle (n = 4; δ¹³C: −22.7 ± 0.8‰, δ¹⁵N: 8.1 ± 0.8‰), and horse (n = 4; δ¹³C: −23.4 ± 1.2‰, δ¹⁵N: 6.7 ± 1.8‰) into one domestic animal food source (δ¹³C: −23.1 ± 0.8‰, δ¹⁵N: 7.6 ± 1.3‰). The isotopic values of the food sources considered at each site were determined from animal and plant remains found in the same or adjacent contemporaneous sites wherever possible. To account for systematic diet-to-collagen isotopic difference, we used a Δ¹³C_{diet-collagen} value of 5 ± 1‰ and a Δ¹⁵N_{diet-collagen} of 4–5 ± 1‰. These values are within the average range used in previous paleodiet studies (Ambrose and Norr, 1993; Hedges and Reynard, 2007; Lee-Thorp et al., 1989; Newsome et al., 2004; Schoeninger, 1985). Discrimination factors are also influenced by the isotopic compositions of dietary macronutrients (Froehle et al., 2010; Jim et al., 2006). Given the variation in C₃ and C₄ resource use across our study area, variable discrimination factors may ultimately be warranted, but we chose to use static values for the sake of simplicity; in the absence of experimental estimates for the localities in our study, this is also the most parsimonious choice. Finally, we did not explicitly consider the isotopic routing of dietary protein to bone collagen in the models. Instead, we assume that bone collagen δ¹³C and δ¹⁵N values are derived from dietary protein. Our mixing model results, therefore, are quantitative estimates of the proportional contribution of various plant and animal resources to human dietary protein. Because dietary routing may vary with the protein content of the diet (e.g. Martínez del Rio et al., 2009; Newsome et al., 2011; Schwarcz, 2002), it will be important to consider this issue in further investigations.

It is important to note that we have analyzed representative samples of all potential food sources found during excavation; however, we cannot rule out the possibility that other food sources were consumed but were not present in the archaeological record. As δ¹³C and δ¹⁵N values are not currently available from archaeological plant material in Kazakhstan, so for those sites, we used values from western China; this decision involves significant assumptions, which we address in the ‘Results’ section.

Results

Figure 2 and Table 3 summarize the results of the mixing models performed on six sites in southern Kazakhstan and western China. Model performance varied from site to site. In western China, the model shows the most certainty in Wuba, with all six possible dietary sources exhibiting constrained uncertainties (see Figure 2, the boxplots show Bayesian 95% credible intervals). Wheat and barley were less well constrained for Huoshaogou, in part because they were negatively correlated (−0.62), meaning that the model could not easily discern between the sources; model solutions that favored wheat excluded barley and vice versa. This was also true for wheat at Ganguai, which was negatively correlated with all of the other sources but particularly strongly with pig or cattle (−0.85) and millet (−0.76). The best-constrained results from Kazakhstan were for Karatuma (see Figure 2). Nonetheless, domestic animals from Karatuma were negatively correlated with several other sources at Kainar-Bulak I (−0.55 with Cervus, −0.52 with dog, and −0.43 with the Fertile Crescent cereals), making it difficult for the model to distinguish the difference among them in dietary contributions. The Fertile Crescent crops and sheep/goat were strongly negatively correlated (−0.87) at Oi-Dzailau and Karatuma (domestic animals and C₃ crops: −0.98), although millet was very well constrained at that site.

Figure 2.

Summary of stable isotope data input into the mixing models and the mixing model results for sites in Kazakhstan (upper panel, a–c) and China (lower panel, d–f). Within each panel, sites are ordered from oldest (left) to youngest (right). The results are labeled as where human bones are recovered. Diet composition data are drawn from the nearest possible locations (see Tables 1 and 2 for details).

Table 3.

Estimated proportional dietary contributions of various animal and plant food resources to archaeological humans with Bayesian credible intervals.

Site	Source	Mean	SD	95% credible interval		Range
Site	Source	Mean	SD	2.5%	97.5%	Range
Oi-Dzailau VII	Wheat or barley	0.17	0.10	0.02	0.39	0.37
	Millet	0.24	0.06	0.13	0.36	0.23
	Sheep or goat	0.59	0.12	0.35	0.79	0.44
	SD δ¹³C	1.97	0.58	1.15	3.38
	SD δ¹⁵N	0.60	0.48	0.03	1.76
Kainar-Bulak I	Wheat or barley	0.14	0.10	0.02	0.37	0.35
	Millet	0.44	0.03	0.37	0.50	0.13
	Dog	0.11	0.05	0.03	0.22	0.19
	Domestic animals (sheep or goat and cattle and horse)	0.17	0.10	0.02	0.39	0.37
	Deer	0.14	0.08	0.02	0.32	0.30
	SD δ¹³C	1.33	0.33	0.80	2.09
	SD δ¹⁵N	1.25	0.33	0.70	1.98
Karatuma	Wheat or barley	0.12	0.08	0.02	0.31	0.29
	Millet	0.23	0.02	0.20	0.26	0.06
	Domestic animals (sheep or goat and cattle)	0.65	0.08	0.45	0.76	0.32
	SD δ¹³C	0.61	0.24	0.09	1.06
	SD δ¹⁵N	0.75	0.38	0.06	1.50
Wuba	Wheat	0.01	0.01	0.00	0.03	0.03
	Millet	0.82	0.01	0.80	0.84	0.04
	Dog	0.01	0.01	0.00	0.02	0.02
	Cattle	0.01	0.01	0.00	0.03	0.03
	Sheep or goat	0.01	0.01	0.00	0.03	0.02
	Deer	0.14	0.01	0.12	0.17	0.05
	SD δ¹³C	0.80	0.12	0.54	1.04
	SD δ¹⁵N	0.88	0.16	0.58	1.21
Huoshaogou	Wheat	0.14	0.09	0.02	0.43	0.41
	Barley	0.26	0.09	0.06	0.41	0.35
	Millet	0.45	0.04	0.35	0.52	0.17
	Dog	0.03	0.02	0.01	0.07	0.07
	Domestic animals (sheep or cattle and pig and horse)	0.11	0.08	0.01	0.31	0.30
	SD δ¹³C	1.85	0.29	1.38	2.51
	SD δ¹⁵N	1.05	0.36	0.17	1.67
Ganguai	Wheat	0.34	0.23	0.03	0.82	0.79
	Barley	0.17	0.11	0.02	0.42	0.40
	Millet	0.11	0.05	0.02	0.20	0.18
	Sheep or goat and deer	0.18	0.10	0.03	0.38	0.35
	Pig or cattle	0.21	0.10	0.02	0.40	0.38
	SD δ¹³C	1.43	0.25	1.01	1.97
	SD δ¹⁵N	0.46	0.26	0.03	0.96

SD: standard deviation.

In terms of the proportion (Bayesian 95% credible intervals), C₄ protein (likely millet) in human diets became significant in the Bronze and Iron Ages in southern Kazakhstan. At a population level, millet(s) comprises approximately 10–35% of dietary protein at Oi-Dzailau VII (1600–1400 BC), about 20–25% in Karatuma (400–200 BC), and 35–50% at Kainar-Bulak I (200 BC–AD 300). In China, the millet contribution to dietary protein falls during the Bronze Age from approximately 80–85% at Wuba (2450–1950 BC) to 35–55% at Huoshaogou (1900–1300 BC) to just 2–20% at Ganguai (1350–950 BC). In addition to the uncertainties outlined above, these estimated proportions also reflect the individual differences within a community. For example, at Oi-Dzailau VII, two individuals that are mostly enriched and depleted in ¹³C were selected for further analyses (see Table S1 for individual proportional dietary contributions, available online; individuals selected for analyses are indicated in Figure 2). In individual X (δ¹³C: 8.8‰, δ¹⁵N: 22.5‰), millet contributed to approximately 0–10% of dietary protein, and in individual Y (δ¹³C: 8.8‰, δ¹⁵N: 22.5‰), the figure is about 30–50%.

The people of southern Kazakhstan and western China also diverge in their reliance on animal versus plant protein. In southern Kazakhstan, animal protein makes up the largest proportion of the protein component of human diets in the second millennium BC, comprising between approximately 35% and 80% at Oi-Dzailau VII. This resource probably remained important in the first millennium BC at Karatuma (~60–70%) and at Kainar-Bulak I. When all animal resources are considered together, animal protein is as important as millet in Kainar-Bulak I (approximately 40% of the dietary protein input). In contrast, millet is the principal food resource for the inhabitants of Wuba, comprising >80% of human diets there. In later periods, wheat and/or barley gain significance – at Huoshaogou (averaging at about 15% and 25%, respectively), and at Ganguai, wheat becomes the most important food resource (>30% on average) followed most closely by pig and cattle group (about 20% on average). In all three cases, cereals contribute the most protein to human diets in China.

One problem with the Karatuma model is that the human data do not plot within the mixing space created by the possible food sources. This suggests that an important food resource was not available for isotopic analysis (e.g. freshwater fish and/or wild plants; Ventresca Miller et al., 2014) and/or the discrimination factors used were incorrect. Nonetheless, the result that terrestrial animal protein comprised the primary dietary protein intake for individuals from this site is consistent with the previous work (Ventresca Miller et al., 2014).

Two issues are worth consideration. The sample size of the plant data is very small, so variations within the same crop are not fully captured. Our results, therefore, offer quantitative predictions as to the importance of these food sources to human diets that we hope will be tested as more data become available. More importantly, these results are dependent on isotope values for millet and barley from China. This is because, to the best of our knowledge, there are hitherto no published isotope data for either modern or archaeological millet and barley from Kazakhstan. Plant δ¹⁵N values are related to multiple factors, such as the availability of nutrients and water, and can, therefore, vary significantly geographically, so values from China may be poor approximations for Kazakhstan. For example, possible explanations for the high δ¹⁵N values of the Chinese millets could be manuring or aridity. Another study looking at stable isotope values in archaeological plant remains of Bronze Age samples from the Eurasian Caspian Steppe, to the east of Kazakhstan, found average δ¹³C values for C₃ and C₄ plants of −26‰ and −12‰, respectively (Shishlina et al., 2012). δ¹⁵N values of Lithospermum officinale seeds ranged from 9.6‰ to 13.0‰ and amaranth seeds had a value of 6.4‰. To investigate how the mixing model results might change if millet values from Kazakhstan have much lower δ¹⁵N values than those found in China, we ran the models for Oi-Dzailau, Karatuma, and Kainar-Bulak with an arbitrary value of 3‰ instead of 9.4‰. At Oi-Dzailau, this lower nitrogen isotope value model predicts that millet still contributes a mean value of 19% compared with 24% in the original model, but the confidence intervals are larger (8–30%). For Karatuma, there is no change in the results; millet still makes up 19–25% of the diet. And finally, at Kainar-Bulak, millet contribution also stays approximately the same (decreases slightly in importance from a mean value of 44% to 42%). These results suggest that the carbon isotope value is the primary driver, which makes sense, given that this is a concentration-weighted model and digestible [N] from plants is much smaller than from animal food sources. Regardless, the pattern of increasing reliance on millet over time holds.

Discussion

Rapid adoption of wheat and/or barley and delayed millet consumption

The results of the mixing models confirm the dietary shifts previously documented in both regions (Liu et al., 2014; Motuzaite Matuzeviciute et al., 2015a). A few aspects are worth further consideration. There was a considerable delay between the earliest evidence for millet grains and the adoption of millet as a substantial food resource in southeast Kazakhstan. The earliest archaeobotanical evidence of millet (from Begash) precedes the first carbon isotopic evidence of substantial millet consumption (from Oi-Dzailau VII) by a few centuries. It is worthy of note, however, that all of the isotopic evidence to date post-dates the archaeobotanical evidence from Begash. Stable isotope analysis is relatively insensitive to minor dietary components; thus, the isotopic indication of millet consumption at Oi-Dzailau suggests that some individuals consumed significant quantities of millet and others did not. Our model estimates that on average, millet contributes approximately 10–35% of the dietary protein. Nonetheless, this eating habit varies according to individuals; in Oi-Dzailau, some consumed millet as 30–50% dietary protein source, and others consumed almost none.

This pattern is in contrast to that observed in western China, where the earliest archaeobotanical evidence for the presence of C₃ cereals (primarily wheat and/or barley) in Gansu is archaeologically approximately contemporaneous with isotopic evidence for the large-scale consumption of C₃ plant resources in the early second millennium BC. Our data show that wheat and/or barley contribute about 40% (mean value) of the dietary protein in Huoshaogou and Ganguai, indicating that Fertile Crescent crops were consumed in significant quantities concurrent with or soon after their first appearance in this region. Although the proportion varies, all individuals from Huoshaogou and Ganguai had consumed C₃ grains directly. This would support the argument that the Fertile Crescent crops were moved and/or adopted in order to provide a new staple foodstuff to the people.

The delay of isotopically detectable millet consumption in Kazakhstan, however, suggests its initial introduction as a minor component of human diets. One interpretation is that millet was not first introduced to Central Asia as a staple food, but for other reasons, perhaps ritual use or as an indictor of wealth or status. Indeed, some of the earliest evidence of millet was from a burial context rather than from a settlement, indicating that it may well have had a special significance beyond subsistence. Frachetti (2015), for example, argued that the crop remains from Begash were the ‘seeds for the soul’. It could be the case that some foodstuffs would be specifically associated with special ritual events, such as the key life history rites de passage. Nonetheless, these events typically emphasize timelessness and ancestry and do not offer an immediate conceptual link with either novel or normal cereal grains. It is also entirely possible that millet was exploited as a risk reduction strategy or in the context of a seasonal mobility model. Thus, there are a variety of reasons why an introduced crop may remain a minor component of the diet, and we currently lack the full range of contextual evidence necessary to infer the most plausible reason in this case.

Our data also reveal different strategies of consumption of dietary animal protein. Contemporary people in Kazakhstan and western China diverge in their reliance on animal versus plant protein. Animal protein derived from domestic animals or game consistently made a substantial contribution to human diets in Kazakhstan. In western China, however, while meat consumption increased relatively in the period after 1900 BC, various crops (either C₃ or C₄) were consistently the most important food resource for the population in this region. The dietary shift documented in Gansu was probably connected with the direct consumption of wheat and/or barley.

Our data allow us to contrast the dietary situations at each end of the ‘Mountain Corridors’, that is, the Hexi Corridor and southern Kazakhstan. In between the two regions lies Xinjiang (Xinjiang Uyghur Autonomous Region). Human isotopic values from several sites in this region have been published (Zhang and Li, 2006; Zhang and Zhu, 2011; Zhang et al., 2009, 2010). The current evidence suggests that the dietary choices in Xinjiang vary from community to community during the second millennium BC. It is seemingly the millet consumers lived side by side with communities consumed more meat and C₃ foodstuff and less millet. However, published isotopic values derived from animal remains are limited at the moment. Future isotopic research to combine human and animal data may be anticipated.

Agencies of translocations of wheat and/or barley and millet

The gradual adoption of millet in southern Kazakhstan serves to highlight the rapidity of the adoption of wheat and barley in west China. During the second millennium BC, wheat and barley were both growing in the fields and being widely consumed in significant quantities in Gansu, constituting a revolution in eating habits. By contrast, the first appearance of millet currently predates the isotopic evidence of consumption on a significant scale in Kazakhstan.

Liu et al. (2014) argued that the rapid adoption of wheat and/or barley in Gansu could be understood in the context of social and ecological challenges in relation to the expansion of human settlement to arid northwest China. In Gansu, the global drought event of the time was complicated by a widespread weakening of the summer monsoon between 2500 and 2000 BC (An et al., 2005a, 2005b). A plausible explanation of the rapid dietary change in this region is that the early colonizers faced a series of ecological and social challenges: the threat of challenging environments, the threat of famine, and population pressure in spatially limited arable lands. In the face of such challenges, conservative food choices may simply not have been sustainable. In addition, the data lend support to the argument that wheat and/or barley were initially introduced to northwest China as a new staple food to serve the needs of the poor majority, rather than to the rich minority.

By contrast, dietary conservatism may be more in evidence elsewhere. In Kazakhstan, although the archaeobotanical evidence indicates that millet was present from the late third millennium BC, the current isotopic evidence suggests that millet was not consumed on a significant scale until almost a few hundred years later. In the later case, the newly developed millet-eating habit varies according to individuals. The individual variation and the time lag between the initial introduction of millet and its subsequent consumption on a significant scale could be explained by the initial use being connected with a minority, for example, elites, ritual specialists, or seasonal mobile groups.

Conclusion

In this paper, we apply Bayesian stable isotope mixing models to published and newly obtained isotopic data in order to quantitatively estimate the contribution of different food resources to human diets, and we consider the complexity of human food strategies at both ends of the mountain corridors: southern Kazakhstan and the Hexi Corridor in western China. Our results contrast the rapid adoption of wheat and/or barley in the Hexi Corridor with the gradual, incremental adoption of millet in southern Kazakhstan during the second millennium BC. The people who lived in the Hexi Corridor consumed wheat and/or barley on a considerable scale shortly after the crops were introduced into the region. In Kazakhstan, millet consumption is not isotopically detectable until around a few hundred years after the first archaeobotanical evidence from the region. In addition, the people in southern Kazakhstan and western China diverge in their reliance on animal, as opposed to plant, protein. These patterns are best explained as different driving forces facilitating the eastward expansion of the Fertile Crescent crops and westward expansion of millet.

Footnotes

Acknowledgements

The authors are thankful to Catherine Kneale and James Rolfe from Cambridge for assistance with isotopic analysis. We are also grateful to Pavel Tarasov for helping with the manuscript and to Professors Mayke Wagner and Pavel Tarasov and Dr Robert Spengler for organizing the workshop, entitled ‘The Introduction and Intensification of Agriculture in Central Eurasia’, Berlin in 2015.

Funding

This work was supported by the European Research Council, under grant 24964 (FOGLIP); Washington University Deanery Office Grant; American Association of University Women (AAUW); and International Center for Advanced Renewable Energy and Sustainability (I-CARES).

References

Ambrose

Norr

(1993) Experimental evidence for the relationship of the carbon isotope ratios of whole diet and dietary protein to those of bone collagen and carbonate. In: Lambert

Grupe

(eds) Prehistoric Human Bone: Archaeology at the Molecular Level. Berlin: Springer-Verlag, pp. 1–37.

Feng

Z-D

Barton

(2005a) Dry or humid? Mid-Holocene humidity changes in arid and semi arid China. Quaternary Science Reviews 25: 351–361.

Tang

Barton

. (2005b) Climate change and cultural response around 4000 cal yr B.P. in the western part of Chinese Loess Plateau. Quaternary Research 63.

Barton

(2014) An evaluation of competing hypotheses for the early adoption of wheat in East Asia. World Archaeology 46: 775–798.

Boivin

Fuller

Crowther

(2012) Old World globalization and the Columbian exchange: Comparison and contrast. World Archaeology 44: 452–469.

Castillo

(2011) Rice in Thailand: The archaeobotanical contribution. Rice 4: 114–120.

Chen

F-H

Dong

F-H

Zhang

D-J

. (2014) Agriculture facilitated permanent human occupation of the Tibetan Plateau after 3600BP. Science 347: 248–250.

Costantini

(1984) The beginning of agriculture in the Kachi Plain: The evidence of Mehrgarh. In: Allchin

(ed.) South Asian Archaeology. Cambridge: Cambridge University Press, pp. 29–33.

Flad

. (2010) Early wheat in China: Results from new studies at Donhuishan in the Hexi Corridor. The Holocene 17: 555–560.

10.

Frachetti

(2008) Pastoralist Landscape and Social Interaction in Bronze Age Eurasia. Berkeley, CA: University of California Press.

11.

Frachetti

(2012) Multiregional emergence of mobile pastoralism and nonuniform institutional complexity across Eurasia. Current Anthropology 53: 2–38.

12.

Frachetti

(2014) Seeds for the soul: ideology and diffusion of domesticated grains across Inner Asia. In: Mair

Hickman

(eds) Reconfiguring the Silk Road. Philadelphia, PA: University of Pennsylvania Museum Press, pp. 41–54.

13.

Frachetti

Spengler

Fritz

. (2010) Earliest direct evidence for broomcorn millet and wheat in the central Eurasia steppe region. Antiquity 84: 993–1010.

14.

Froehle

Kellner

Schoeninger

(2010) FOCUS: Effect of diet and protein source on carbon stable isotope ratios in collagen: Follow up to Warinner and Tuross (2009). Journal of Archaeological Science 37: 2662–2670.

15.

Guan

Tang

. (2007) Tonghua Wanbazi yizhi zhugu de C, N wending tongweisu fenxi [Stable isotopic analysis on sus bones from the Wanfabazi site, Tonghua, Jilin]. Kexue Tongbao (Chinese Science Bulletin) 52: 3393–3396.

16.

Hare

Fogel

Stafford

. (1991) The isotopic composition of carbon and nitrogen in individual amino-acids isolated from modern and fossil proteins. Journal of Archaeological Science 18: 277–292.

17.

Harris

(2010) Origins of Agriculture in Western Central Asia. Philadelphia, PA: University of Pennsylvania Museum.

18.

Hedges

REM

Reynard

(2007) Nitrogen isotopes and the trophic level of humans in archaeology. Journal of Archaeological Science 34: 1240–1251.

19.

Hunt

Linden

Liu

. (2008) Millets across Eurasia: Chronology and context of early records of the genera Panicum and Setaria from archaeological sites in the Old World. Vegetation History and Archaeobotany 17: 5–18.

20.

Jim

Jones

Ambrose

. (2006) Quantifying dietary macronutrient sources of carbon for bone collagen biosynthesis using natural abundance stable carbon isotope analysis. British Journal of Nutrition 95: 1055–1062.

21.

Jones

Harriet

Kneale

. (in press) Food globalisation in prehistory: The agrarian foundation of an interconnected continent. Journal of the British Academy.

22.

Jones

Hunt

Lightfoot

. (2011) Food globalization in prehistory. World Archaeology 43: 665–675.

23.

Koch

Phillips

(2002) Incorporating concentration dependence in stable isotope mixing models: A reply to Robbins, Hilderbrand and Farley (2002). Oecologia 133: 14–18.

24.

Lee-Thorp

Sealy

Van der Merwe

(1989) Stable carbon isotope ratio differences between cone collagen and bone apatite and their relationship to diet. Journal of Archaeological Science 18: 343–354.

25.

Levine

(1999) Late Prehistoric Exploitation of the Eurasian Steppe. Cambridge: McDonald Institute for Archaeological Research.

26.

Lightfoot

Liu

Jones

(2013) Why move starchy cereals? A review of the isotopic evidence for prehistoric millet consumption across Eurasia. World Archaeology 45: 574–623.

27.

Lightfoot

Motuzaite-Matuzeviciute

O’Connell

. (2014) How ‘pastoral’ is pastoralism? Dietary diversity in Bronze Age communities in the central Kazakhstan steppes. Archaeometry 57: 232–249.

28.

Lin

M-C

(2015) Saiyima-Tubinnuo wenhua yu shiqian sichou zhilu [Seima-Turbino and prehistoric silk road]. Wenwu (Cultral Relics) 10: 49–63.

29.

Linduff

Mei

(2009) Metallurgy in ancient Eastern Asia: Retrospect and prospects. Journal of World Prehistory 22: 265–281.

30.

Liu

Fuller

Jones

(2015) Early agriculture in China. In: Barker

Goucher

(eds) The Cambridge World History: A World with Agriculture, 12,000 BCE-500CE, vol. II. Cambridge: Cambridge University Press, pp. 310–334.

31.

Liu

Hunt

Jones

(2009) River valleys and foothills: Changing archaeological perceptions of north China’s earliest farms. Antiquity 83: 82–95.

32.

Liu

Jones

(2014) Food globalisation in prehistory: Top down or bottom up? Antiquity 88: 956–963.

33.

Liu

Lightfoot

O’Connell

. (2014) From necessity to choice: Dietary revolutions in west China in the second millennium BC. World Archaeology 46: 661–680.

34.

Liu

Lister

Zhao

Z-Z

. (2016) The virtues of small grain size: Potential pathways to a distinguishing feature of Asian wheats. Quaternary International. DOI: 10.1016/j.quaint.2016.1002.1059.

35.

Dong

Lightfoot

. (2013a) Stable isotope analysis of human and faunal remains in the western loess plateau, approximately 2000 cal. BC. Archaeometry 65(S1): 237–255.

36.

Dong

Liu

. (2013b) Stable isotope analysis of human and animal remains at the Qijiaping site in middle Gansu, China. International Journal of Osteoarchaeology. Epub ahead of print 13 December. DOI: 10.1002/oa.2379.

37.

Martínez del Rio

Wolf

Carleton

. (2009) Isotopic ecology ten years after a call for more laboratory experiments. Biological Reviews 84: 91–111.

38.

Meadow

(1996) The origins and spread of agriculture and pastoralism in north western South Asia. In: Harris

(ed.) The Origins and Spread of Agriculture and Pastoralism in Eurasi. London: UCL Press, pp. 390–412.

39.

Mei

(2003) Qijia and seima-turbino: The question of early contacts between Northwest China and Eurasian setppe. Bulletin of the Museum of Far Eastern Antiquities 75: 31–54.

40.

Miller

(2003) The use of plants at Anau North. In: Hiebert

Kurdansakhatov

(eds) A Central Asian Village at the Dawn of Civilization: Excavations at Anau, Turkmenistan. Philadelphia, PA: University of Pennsylvania Museum, pp. 127–138.

41.

Motuzaite Matuzeviciute

Lightfoot

O’Connell

. (2015a) The extent of agriculture among pastoralist societies in Kazakhstan using stable isotope analysis of animal bone. Journal of Archaeological Science 59: 23–24.

42.

Motuzaite Matuzeviciute

Preece

Wang

. (2015b) Ecology and subsistence at the Mesolithic and Bronze Age site of Aigyrzhal-2, Naryn valley, Kyrgystan. Quaternary International. Epub ahead of print 18 August 2015. DOI: 2010.1016/j.quaint.2015.2006.2065.

43.

Motuzaite Matuzeviciute

Staff

Hunt

. (2013) The early chronology of broomcorn millet (Panicum miliaceum) in Europe. Antiquity 87: 1073–1085.

44.

Murphy

Schulting

Beer

. (2013) Iron Age pastoral nomadism and agriculture in the Eastern Eurasian Steppe: Implications from dental palaeopathology and stable carbon and nitrogen isotopes. Journal of Archaeological Science 40: 2547–2560.

45.

Newsome

Fogel

Kelly

. (2011) Contributions of direct incorporation from diet and microbial amino acids to protein synthesis in Nile tilapia. Functional Ecology 25: 1051–1062.

46.

Newsome

Phillips

Culleton

. (2004) Dietary reconstruction of an early to middle Holocene human population from the central California coast: Insights from advanced stable isotope-mixing models. Journal of Archaeological Science 25: 1051–1062.

47.

Parnell

Inger

Bearhop

. (2010) Source partitioning using stable isotopes: Coping with too much variation. PLoS ONE 5: e9672.

48.

Petrie

(2015) Mehrgarh, Pakistan. In: Barker

Goucher

(eds) The Cambridge World History: A World with Agriculture, 12000 BCE-500 CE, vol. II. Cambridge: Cambridge University Press, pp. 289–309.

49.

Phillips

Koch

(2002) Incorporating concentration dependence in stable isotope mixing models. Oecologia 130: 114–125.

50.

Phillips

Inger

Bearhop

. (2014) Best practices for use in stable isotope mixing models in food-web studies. Canadian Journal of Zoology 92: 823–835.

51.

Rawson

(2013) Ordering the exotic: Ritual practices in the late Western and early Eastern Zhou. Artibus Asiae 73: 5–24.

52.

Schoeninger

(1985) Trophic level effects on 15N/14N and 13C/12C ratios in bone collagen and strontium levels in bone mineral. Journal of Human Evolution 14: 515–525.

53.

Schwarcz

(2002) Some biochemical aspects of carbon isotopic paleodiet studies. In: Ambrose

Katzenberg

(eds) Biogeochemical Approaches to Paleodietary Analysis. New York: Kluwer Academic/Plenum Publishers.

54.

Sherratt

(2005) The Trans-Eurasian exchange: The prehistory of Chinese relations with the West. In: Mair

(ed.) Contact and Exchange in the Ancient World. Honolulu: Hawaii University Press, pp. 30–61.

55.

Shishlina

Sevastyanov

Hedges

REM

(2012) Isotope ratio study of Bronze Age samples from the Eurasian Caspian Steppes. In: Kaiser

Burger

Schier

(eds) Population Dynamics in Prehistory and Early History. New Approaches by Using Stable Isotopes and Genetics. Berlin; Boston, MA: De Gruyter, pp. 177–197.

56.

Spengler

(2015) Agriculture in the Central Asian Bronze Age. Journal of World Prehistory 28: 215–253.

57.

Spengler

Willcox

(2013) Archaeobotanical results from Sarazm, Tajikistan, an Early Bronze Age settlement on the edge: Agriculture and exchange. Journal of Environmental Archaeology 18: 211–221.

58.

Spengler

Frachetti

Doumani

(2014a) Late Bronze Age agriculture at Rasbas in the Dzhungar Mountains of eastern Kazakhstan. Quaternary International 348: 147–157.

59.

Spengler

Frachetti

Doumani

. (2014b) Early agriculture and crop transmission among Bronze Age mobile pastoralists of Central Eurasia. Proceedings of the Royal Society B: Biological Sciences 281: 1–7.

60.

Svyatko

Schulting

Mallory

. (2013) Stable isotope sietary analysis of prehistoric populations from the Minusinsk Basin, Southern Siberia, Russia: A new chronological framework for the introduction of millet to the eastern Eurasian steppe. Journal of Archaeological Science 40: 3936–3945.

61.

Ventresca Miller

Usmanova

Logvin

. (2014) Subsistence and social change in Central Eurasia: Stable isotope analysis of populations spanning the Bronze Age transition. Journal of Archaeological Science 42: 525–538.

62.

Wagner

Tarasov

Hosner

. (2012) Mapping of the spatial and temporal distribution of archaeological sites of northern China during the Neolithic and Bronze Age. Quaternary International 290–291: 344–357.

63.

Webber

(1991) Plants and Harappan Subsistence: An Example of Stability and Change from Rojdi. Boulder, CO: Westview Press.

64.

Webber

(2001) Ancient seeds: Their role in understanding South Asia and its past. In: Ford

(ed.) Ethnobiology at the Millennium, Past Promise and Future Prospects. Ann Arbor, MI: Museum of Anthropology, University of Michigan.

65.

Weiss

Zohary

(2011) The Neollithic Southwest Asian Founder Crops: Their biology and archaeobotany. Current Anthropology 52: S237–S254.

66.

Yang

Wan

Perry

. (2012) Early millet use in northern China. Proceedings of the National Academy of Sciences of the United States of America 109: 3726–3730.

67.

Zhang

Wang

Ferguson

. (2015) Ancient plant use and palaeoenvironmental analysis at the Gumugou Cemetery, Xinjiang, China: Implication from desiccated plant remains. Archaeological and Anthropological Sciences. Epub ahead of print 21 May. DOI 10.1007/s12520-015-0246-3.

68.

Zhang

(2006) Xinjiang Nileke Xian Kongkeke yihao mudi gudai jumin de shiwu jiegou fenxi [Analysis on paleodiet of ancient inhabitants in no. 1 cemetery of Qiongkeke at Nika County, Xinjiang]. Xiyu Yanjiu (The Western Regions Studies) 4: 78–82.

69.

Zhang

Zhu

(2011) Xinjiang Gumugou mudi rengu de wending tongweisu fenxi: Zaoqi Luobubo xianmin yinshi jiegou chutan (Stable isotopic analysis on human bones from Gumugou cemetery, Xinjiang: Paleodietary reconstruction on Luobubo population). Xiyu Yanjiu (The Western Regions Studies) 3: 91–96.

70.

Zhang

Chang

Liu

(2009) Xinjiang Balikun xian Heigouliang mudi chutu rengu de shixing fenxi [Stable isotopic analysis on human bones from Heigouliang cemetery in Barkol, Xinjiang]. Xiyu Yanjiu (The Western Regions Studies) 3: 45–49.

71.

Zhang

Chang

Liu

(2010) Xinjiang Hami Tianshanbeilu mudi chutu rengu de wending tongweisu fenxi [Stable isotopic analysis of humanbones unearthed from Tianshan Beilu cemetery in Hami, Xinjiang]. Xiyu Yanjiu (The Western Regions Studies) 2: 38–43.

72.

Zhou

Garvie-Lok

(2015) Isotopic evidence for the expansion of wheat consumption in northern China. Archaeological Research in Asia 4: 25–35.