Pottery Impressions Reveal Earlier Westward Dispersal of Foxtail Millet in Inner Asian Mountain Corridor

Abstract

The Inner Asian Mountain Corridor (IAMC) has been identified as a major pathway for the westward dispersal of millet from Northern China, where it was initially cultivated. Cross-disciplinary investigations are necessary to distinguish cultivated millet taxa from their wild relatives and to clarify the social context underlying millet adoption in novel environments. Despite the ambiguity in distinguishing Setaria italica from Panicum miliaceum or other Setaria species using conventional analysis of charred macro remains, recent attention has focused on the time gap between the introduction of S. italica to IAMC following P. miliaceum. Here, we employed a pottery impression casting method on materials from four Bronze Age sites in eastern/southeastern Kazakhstan to investigate the surface textures of grain impressions on the surface of pottery containers. We successfully identified both millets (Setaeria and Panicum) from three of the sites, Begash, Tasbas, and Dali in the IAMC. Based on our findings, two species of millet were introduced to the region within a much shorter range of time than previously estimated. In addition, the current evidence supports the premise that these cereals were likely utilized for human consumption.

1. Introduction

Central Asia holds the key to understanding the early spread of cereal crops across Eurasia, including the eastward dispersal of wheat and barley and the westward movement of millet during the third and second millennium BC. The latter issue has drawn significant scholarly attention in recent years and inspired cross-disciplinary contributions. The dynamism and vibrant connectivity among prehistoric communities in the region have been well-documented through archaeological research using material culture, zooarchaeological evidence, archaeobotany, and stable isotope evidence [1,2,3,4,5,6,7,8,9,10,11,12,13]. Nevertheless, our understanding of Central Asian subsistence and dietary patterns is constrained by broad-sweeping categorizations of regional differences (e.g., mountainous terrain vs. steppes), despite the substantial ecological and cultural diversities within each region. In this study, we focus on localized millet use at key sites in Central Asia, dating back to the third and second millennium BC. The application of scanning electron microscopy (SEM) and silicon replicas allows for identifications of pottery impressions at the species level, which has been previously undocumented. These results can enrich our understanding of the timing and function of millet adaptation in the region.

This paper presents previously unobtained high-resolution evidence from ceramic impressions that inform past agricultural practices in Central Asia. We obtained our research results by employing a novel analytical technique known as pottery impression silicon casting. While the general technique has been developed for several years, recent refinements in using SEM to examine silicon replicas have greatly enhanced the level of detail that this seemingly simple and “traditional” analytic approach can document. The improved technique enables identification of plant species with much greater reliability compared to similar studies conducted during the past century. An additional advantage of this method is that it enables investigations of the rich ceramic assemblages previously excavated, eliminating the need for new excavations for data collection. This method has been previously applied in Japan [14,15], South Korea [16], eastern Russia [17], Ukraine [18], Belarus [19], and Western Kazakhstan [20], and has been proven productive in providing insights into agricultural dispersals, their chronology, and changes in subsistence across time and space. The specific procedure and methodology have been presented in several papers [21,22,23].

Our study area within Kazakhstan has previously yielded relatively rich archaeobotanical data concerning the spread of cultivated cereals between east and west through the Inner Asian Mountain Corridor (IAMC) [24] from the mid-third millennium BC onwards. Through the investigation of ceramic assemblages from the sites of Begash, Dali, and Tasbas (in southeastern Kazakhstan) and the site of Koken (in eastern Kazakhstan), this paper offers new and comparative insights into millet dispersal and its presence/absence in two contrasting environments of Kazakhstan. Furthermore, we address unanswered questions regarding the time gap between foxtail and broomcorn millet dispersals through the mountain corridor of Central Asia, as well as the role of millet in human diets and/or livestock feed from the early to final Bronze Age.

2. Materials and Methods

Regarding the formation process of pottery impressions, J. M. Renfrew [25] suggested that “Stray grains spilled during meal preparation, entered the moist clay, and thus became embedded in the walls of the vessel”. Although there may be other pathways, including deliberately adding grains to the clay for aesthetic or functional purposes, we consider accidental spilling as the main type. We made this observation simply because impressions were not frequently encountered in our investigated samples. Despite ambiguities in impression formation and its connection with millet processing, pottery impressions provide complementary information to the macrofossil analysis, enabling morphological details that cannot be preserved through charring.

The samples consisted of pottery sherds unearthed from four Bronze Age sites in southeastern and eastern Kazakhstan. Three of these sites, Begash, Tasbas, and Dali, are located along the IAMC, a well-established pathway for East and West exchange [26]. Wheat originating from West Asia and broomcorn millet from East Asia have been reported from Begash, dating back to the later third millennium BC, constituting one of the oldest of such records across Eurasia. To establish a comparative scope, we also examined the pottery from Koken, a site located in a more northerly environment beyond the IAMC with a different ecological setting. Figure 1 shows the locations of the four sites. In total, we examined 31.4 kg of pottery sherds excavated from the four sites (Begash: 3 kg, 169 pieces; Tasbas: 5.2 kg, 203 pieces; Dali: 17.8 kg, 1903 pieces; Koken: 5.4 kg, 750 pieces).

Regarding the chronologies of the observed plant impressions, the examined pottery was linked to the stratigraphic units of each individual site and associated radiocarbon results (Table 1). These AMS dates were obtained from measurements on charcoal, carbonized grains, and/or animal and human bones. Furthermore, regional ceramic typologies provide an additional useful reference point for aligning the relevant potsherds with their associated Bronze Age phase (early, middle, late, etc.).

We employed a pottery impression casting method that combines two techniques: a newly developed approach using silicon as a molding material and SEM. In the past, gypsum and clay were used as materials for making molds, but silicon is a more suitable material for ceramic conservation and provides better visual representations. Furthermore, the flexibility of silicon makes it easier to capture shapes embedded deeply within the ceramic fabric. This approach enables the documentation of macro-morphological evidence with higher resolution compared to previous studies. Traditionally, the identification of plant impressions in pottery was achieved with the naked eye or with a magnifying glass, often resulting in low-quality results [27,28,29]. However, with the aid of SEM, researchers can now observe the surface texture of grain and plant fragment impressions at very high magnification (up to ×500, depending on the condition of the replica). Therefore, high-precision analysis can be expected.

The sampling procedure has been previously described [21,22,23] and is as follows: (1) begin by searching for holes of impressions on the surface and cross-section of pottery fragments using the naked eye and magnifying glass. (2) Pay attention to impressions that are not fully exposed, as their shape is mostly located inside the pottery fabric and is only accessible through a small hole on the surface. (3) Carefully remove any sediments in the holes using a soft brush. (4) Coat the impression and the surrounding area of the pottery surface with acetone containing 5% Paraloid B72 to protect the pottery surface. (5) Fill the voids with silicone (Tokuyama Fit Tester, Tokuyama, Japan). (6) Remove the cast from the hole after it has hardened. (7) Clean the surface by removing the coating liquid using 100% acetone.

Once the silicon has hardened, it is removed and then observed and identified using SEM (in our case, a KEYENCE VE-8800, Tokyo Japan) with a low accelerating voltage observation (0.5~20 kV). By targeting small objectives, vacuuming of the material storage can be completed within a short time (approximately five minutes). By using low voltage, there is no need for vapor deposition as a preparation step before observation.

SEM also enables measurements of the width and length of the grains. The identified millets were compared with the existing datasets from Japan (P. miliaceum: n = 70, from five sites, and S. italica: n = 78, from five sites, spanning from the end of the Final Jomon period to the Early Yayoi period), Korea (P. miliaceum: n = 15, from four sites, and S. italica: n = 21, from three sites belonging to Neolithic and Bronze Age), and Ukraine (P. miliaceum: n = 36, from two sites belonging to the Late Bronze Age).

Table 1. Summary of the archaeological sites surveyed in this study.

Site Name	Height Above Sea Level	Average Annual Precipitation	Ecological Environment	Anthropogenic Context		References
Begash	900 m.a.s.l.	400–600 mm	highland steppe ecotone, mountain foothill	1a (2450–1950BC)	burial cist, ritual fire-pit	[1,2]
				1b (1950–1700BC)	domestic hearth
				2 (1625–1000BC)	domestic hearth
Tasbas	1500 m.a.s.l.	400–600 mm	highland steppe ecotone	1 (2832–2492BC)	ritual fire pit	[30,31]
				2a (1416–1287BC)	domestic hearth
				2b (1208–1054BC)	settlement
				3 (930–806BC)	settlement
Dali	1500 m.a.s.l.	400–600 mm	highland steppe ecotone	1 (2850–2500BC)	settlement	[3,32]
				2 (1850–1500BC)	settlement
				3 (1300–1100BC)	settlement
Koken	360 m.a.s.l.	275 mm	semiarid steppe grasslands	1 (2800–2200BC)	settlement	[33]
				Burial (1750–1400BC)	burial
				2 (1400–900BC)	settlement

Table 1. Summary of the archaeological sites surveyed in this study.

Site Name	Height Above Sea Level	Average Annual Precipitation	Ecological Environment	Anthropogenic Context		References
Begash	900 m.a.s.l.	400–600 mm	highland steppe ecotone, mountain foothill	1a (2450–1950BC)	burial cist, ritual fire-pit	[1,2]
				1b (1950–1700BC)	domestic hearth
				2 (1625–1000BC)	domestic hearth
Tasbas	1500 m.a.s.l.	400–600 mm	highland steppe ecotone	1 (2832–2492BC)	ritual fire pit	[30,31]
				2a (1416–1287BC)	domestic hearth
				2b (1208–1054BC)	settlement
				3 (930–806BC)	settlement
Dali	1500 m.a.s.l.	400–600 mm	highland steppe ecotone	1 (2850–2500BC)	settlement	[3,32]
				2 (1850–1500BC)	settlement
				3 (1300–1100BC)	settlement
Koken	360 m.a.s.l.	275 mm	semiarid steppe grasslands	1 (2800–2200BC)	settlement	[33]
				Burial (1750–1400BC)	burial
				2 (1400–900BC)	settlement

3. Results

In Total, 58 casts were taken, and fourteen millet impressions were successfully identified, while 44 impressions could not by attributed to any specific species. Among the identified impressions, there was one grain of foxtail millet and two grains of broomcorn millet from Begash, one grain of foxtail millet and four grains of broomcorn millet from Tasbas, and four grains of foxtail millet and two grains of broomcorn millet from Dali. All of these impressions describe grains that were still wrapped in palea (inner) and lemma (outer) glumes (Figure 2).

Table 2. Summary of the results of seed impression identification.

Site Name	No. of Replica	Site Occupation Phase	Parts in Pottery	Location on/in the Vessel Wall	Botanical Identification	Plant Part	Length (mm)	Width (mm)	Figure Number
Begash	0001	phase 2	rim	outer	Panicum miliaceum	grain with palea and lemma	2.21	1.84	Figure 1A
Begash	0003	phase 1a	body	wall cross-section	Panicum miliaceum	grain with palea and lemma	x	2.37	Figure 1B
Begash	0004	phase 2	body	outer	Setaria italica	grain with palea and lemma	1.91	1.59	Figure 1C
Tasbas	0002	phase 2b	body	outer	Panicum miliaceum	grain with palea and lemma	2.66	2.05	Figure 1D
Tasbas	0005	phase 2b	body	body	Panicum miliaceum	grain with palea and lemma	2.18	1.76	Figure 1E
Tasbas	0006	phase 3	body	wall cross-section	Panicum miliaceum	grain with palea and lemma	2.23	x	Figure 1F
Tasbas	0008	phase 3	base	outer	Setaria italica	grain with palea and lemma	2.11	1.78	Figure 1G
Tasbas	0009	phase 2a	body	body	Panicum miliaceum	grain with palea and lemma	2.93	1.70	Figure 1H
Dali	0002	phase 2	body	wall cross-section	Setaria italica	grain with palea and lemma	1.88	1.69	Figure 1I
Dali	0005	phase 1 or 2	body	outer	Setaria italica	grain with palea and lemma	1.98	1.62	Figure 1J
Dali	0006	phase 2	rim	outer	Setaria italica	grain with palea and lemma	1.81	1.79	Figure 1K
Dali	0011-1	phase 2	rim	inner	Panicum miliaceum	grain with palea and lemma	2.66	2.16	Figure 1L
Dali	0011-3	phase 2	rim	inner	Panicum miliaceum	grain with palea and lemma	2.59	2.23	Figure 1M
Dali	0017	phase 1	rim	lip	Setaria italica	grain with palea and lemma	1.97	1.53	Figure 1N

provides a breakdown of identified millet from each site and their corresponding site occupation phase.

3.1. Millet Identification Criteria

Distinguishing between small seeds of different species, particularly those that are morphologically similar, such as millet, is sometimes challenging for microscopic work on macro-botanical remains. The silicone casting SEM method offers an advantage for fine-grained studies of plant species in archaeological contexts [23]. While charred macro remains extracted using the flotation method are often deformed due to carbonization, seeds kneaded into clay provide undistorted prints. This is especially true for small, seeds such as millet that often lose their palea and lemma due to charring or dehusking. In most cases, the palea and lemma are present at the time when impressions are formed, and the differences in palea and lemma surface texture provide a critical criterion for distinguishing between millet species.

The most distinctive criteria are papillae that are observed on the palea and lemma of foxtail millet (Figure 2C,G,I–K,N). In contrast, the surface of broomcorn millet is smooth without protrusions (Figure 2A,B,D–F,H,L,M). In addition, a crescent-shaped smooth feature sits at the boundary between the palea and lemma of foxtail millet, where the papillae are absent (Figure 2C,G,I,J,N). This foxtail-specific papilla is observed not only in cultivated foxtail millet (Setaria italica), but also in foxtail grass (Setaria viridis). Another criterion of importance is that foxtail millet tends to swell on the lemma side and the palea side tends to be flat (Figure 2C,G,I–K,N), while most broomcorn millet swells both on the lemma and palea sides (Figure 2B,D,F,H,M).

The criteria discussed in this paper are useful for distinguishing between broomcorn millet, foxtail millet, and their closely related relatives at the species level, which is the central focus of this study. Future research to establish identification criteria for other Panicum and Setaria species, as well as other millet-related taxa such as Brachiaria, Echinochloa, Paspalum, and Digitaria, will be plausible to extend the geographical range of this application.

3.2. Appearance Rate of Cereal Impressions

As carbonized material detected by the flotation method has been quantitatively analyzed using the volume of the sample soil as the denominator, it is a common practice to quantify the ubiquity of impression encounters based on the total weight of the studied ceramics [34]. The ubiquity rate of cereal impressions per 10 kg of the three sites was 6.7% at Begash, 9.6% at Tasbas, and 3.4% at Dali. We did not identify any cereal impressions from Koken. It is worth noting that these ubiquity rates are not overly informative due to the small sample size. It nonetheless provides a useful reference dataset for future studies in the region to compare with.

For reference, in the survey of the Japanese archipelago from the Final Jomon period to the Yayoi period (transitional period to farming) which has been well-investigated, the ubiquity rate varies from 0.6% to 26.9%. It is interesting to note that the ubiquity does not necessarily increase over time, as different trends are observed in different typological groups.

3.3. Size of Identified Millets

The measurements of breadth and length of the identified broomcorn millets are listed in Table 2. As previously clarified from the morphological examinations, both the long and short axes of broomcorn millet are longer than those of foxtail millet, even in impressions.

4. Discussion

4.1. Effectiveness of Pottery Impression Analysis

Although this research focused on limited numbers of sites, it successfully identified broomcorn and foxtail millet impressions to the species level from three sites (Begash, Dali, and Tasbas) dating back to the third to second millennium B.C. This study also confirmed the feasibility of the casting method for Central Asian materials. As shown in Figure 2, the species of grains can be clearly identified based on the SEM images of the casts taken from impressions, and reveal the species of grains, making it particularly effective for differentiating small grains such as broomcorn and foxtail millets. However, the chronology of the identified impressions depends on the stratigraphy of the site. While the possibility of intrusion from a younger layer cannot be excluded in theory, our results present the earliest evidence of foxtail millet outside of China, a point we will elaborate on in the following section. Our results align with a previous impression study in Ukraine, informing the introduction of broomcorn millet in the Pontic region during the second millennium B.C. [18]. These results, together with direct AMS dating of charred millet [35,36,37], clarify the chronology of millet dispersal.

4.2. The Morphological Comparison of the Identified Millet Grains

The size of broomcorn millet identified in this study falls within a similar range as that found in the Japanese archipelago, Korea, Kazakhstan, and Ukraine. Similarly, the size of foxtail millet is also common among the Japanese archipelago, Korea, and Kazakhstan (Figure 3 and Figure 4). In particular, the long/short axis ratio or the length of the short axis of foxtail millet is used as criterion for distinguishing between wild and cultivated species. The size of the foxtail millet impressions identified in this study is comparable to or larger than the impressions of cultivated foxtail identified in East Asia.

On the other hand, conducting a precise morphological comparison between seed impressions and charred seeds is difficult. This is because seeds expand by absorbing moisture from the pottery clay, and during the firing of pottery, the surface develops bubbles and then shrinks. Additionally, while carbonized materials often appear in caryopsis form only, impressions record the caryopsis covered with palea and lemma. Therefore, a simple size comparison between the carbonized material and the impressions is not suitable.

4.3. Chronologies of Western Dispersals of Asian Millets

The foxtail millet identified from the impression on an incised rim of a jar from Dali constitutes the oldest record of this cultivated species outside of China, provided that the chronology is confirmed. The archaeological context suggests it belongs to the Phase 1 occupation period, dating to 2850–2500 cal. B.C. based on eight AMS determinations derived from wood charcoal, human bones, and sheep bones as shown in Table 1 in [32]. The potsherd was recovered from sediments associated with a pit-house from the third millennium B.C. Three additional fragments from the same parent vessel were recovered from this context, and all three fragments were located near a small hearth feature inside the pit-house. Although the context is sealed by a thick sediment layer as part of a later Phase 2 house, it is necessary to consider the possibility of intrusions from the upper strata. The same ornament was found on a potsherd from a different vessel in the upper strata near the modern surface. Currently there is insufficient stylistic information to assign this vessel type to Dali Phase 1 or 2 with certainty. If the rim sherd containing the millet impression indeed belongs to the Phase 1 occupation from which it was recovered, it would constitute the earliest foxtail millet outside of China. Otherwise, the current oldest record of charred foxtail millet in the IAMC is from Tasbas Phase 2, dating to the middle of the second millennium B.C. Without a direct date for the pottery sherd with the impression, the appearance of foxtail millet in Central Asia during the third millennium B.C. remains an open question, and further investigation with more extensive sampling will be needed.

As for broomcorn millet, charred grains have already been reported at Begash and Tasbas [1,2], dating as early as the third millennium, which is consistent with the macrofossil evidence. Broomcorn millet was identified from the stratigraphic layer of the third millennium B.C. to the second millennium B.C. at three of the four sites (Table 2). The only third millennium B.C. finding includes the foxtail millet impression from Dali Phase 1, as discussed earlier. Four additional foxtail millet impressions dated to the second millennium B.C. were identified in this study, allowing us to infer that foxtail millet had already reached southeastern Kazakhstan by at least the early second millennium. In particular, three grains of foxtail millet were identified in the Dali Phase 2 (1850–1500 BC) pottery. This is consistent with the macrofossil evidence.

Regarding the spread of millet beyond southeastern Kazakhstan, it is already clear that the arrival of broomcorn millet in Europe dates back to the second millennium B.C. based on direct AMS dating of charred grains [37]. However, the timing and pathway of the western dispersal of foxtail millet across Eurasia remain unclear. In Ukraine, no impressions of foxtail millet were detected in Sabatynivka culture pottery at the Novokyivka site of the Late Bronze Age, although 44 grains of broomcorn millet were identified [18].

Therefore, the arrival of foxtail millet in Europe is thought to postdate broomcorn millet by centuries. This contrasts with the pattern observed in East Asia, where broomcorn and foxtail millets tend to be introduced to a region simultaneously [15,16,17,38,39,40]. However, in Southeast Asia, current evidence suggests that foxtail millet predates broomcorn [41]. For future research, it will be helpful to determine the dates of impressions not only by relying on stratigraphic and typological information but also by employ newly available methods such as direct radiocarbon measurements of lipid residue collected from ceramics [42] or charred remains embedded within pottery fabric [43].

4.4. Mechanism Underlying the Introduction of Millet to Central Asia

Regarding the mechanism underlying the introduction of millets from East to Central Asia, it is reasonable to predict that millets from ancient China were well-integrated into the mobile pastoralist economies during the Bronze Age. A recent ethnographic report of the northwestern side of the Tianshan Mountains shows that modern herders settle every year in spring camps from March to May, summer camps from May to September, spring camps again from September to December, and winter camps from December to March [44]. They also note that “the spring/autumn camp is also near arable land suitable for growing crops or hay and most herders have a block of land which is used primarily for the cultivation of fodder crops to supplement winter grazing”. Such multi-resource integrations with vertical transhumance, which is also proposed for Begash and Dali as analyzed here [3,45], would be advantageous for millet cultivation adapted to the short cultivation period, i.e., sowing in spring camps and harvesting in autumn.

In contrast to the West Asian wheat, which underwent compaction on its journey to the East [11,46], the East Asian millet arrived in southeastern Kazakhstan with similar morphology and size (although the mean grain length is slightly higher than those from Korea and Japan), and the broomcorn millet that arrived in Ukraine did not differ from millet in East Asia (Figure 5). However, this observation will need to be further investigated with more systematic sampling across a broad geographical range. A recent study based on grain metrics measured from a few thousand charred millets across Asia suggests an East-to-West increase in grain size for both broomcorn and foxtail millet (Sun Y. et al., forthcoming). This finding aligns with the previously documented West-to-East grain reductions in wheat and barley, indicating active culinary selections [47].

4.5. The Role of Millet in the Human Diet

Whether millet was initially introduced to the IAMC as a cereal for human consumption or as a commodity with other functions remains an open question that our results may contribute to. The mortuary context of charred broomcorn millet from Begash suggests a non-food status, and it has been hypothesized that the connection with human diets was ambiguous during the third millennium B.C. [1,26]. Faunal isotopic evidence from the early phases of Begash and Dali indicates seasonal consumption of C₄ plants during winter [3]. This implies millet cultivation in the area, possibly for fodder use [3]. On the other hand, recent isotopic work in Kyrgyzstan suggests that some individuals consumed millet in significant quantities, while others ate C₃ resources in the third millennium B.C. [48]. However, in the study area of Eastern Kazakhstan, human isotopic evidence is currently lacking, and evidence of crop processing is unclear, making it challenging to evaluate the role of millet in human diets at the initial stage of introduction.

Our results shed additional light on the matter. The grain impressions we observed clearly depict the inner husks (palea and lemma of the second floret), but with the absence of the outer chaff (glumes and lemma of the first floret). This indicates that the grains incorporated in pottery were products of a late stage of crop processing, e.g., post-winnowing (to remove lighter chaff) and before de-husking (to separate the second floret from the caryopsis). This observation suggests the possibility that pottery making was carried out in the immediate vicinity of crop processing activities. As investigated elsewhere, millet crop processing [47,49] and pottery making [30] were key components of the everyday economy, closely associated with other household activities. This aligns with J. Renfrew’s observation that impressions resulted from stilled grains in food preparation inadvertently entering ceramic making. This scenario implies an unambiguous connection with human consumption. Other possibilities should be acknowledged, including grains being added to moist clay deliberately or being traded as post-winnowing products. However, it is difficult to envision such labor intensive-operations without a human diet aspect. Furthermore, miliacin, a biomarker associated with Panicum miliaceum, was detected in pottery in Eastern Kazakhstan dating to the Early Iron Age [50]. This implies that broomcorn millet was involved in culinary preparation involving boiling during the 1st millennium B.C. at the latest. Further research will be valuable in clarifying the role of millet in culinary practices during the late third millennium B.C., when they were first introduced to the region.

5. Conclusions

We conducted a pottery impression analysis on ceramics sherds dating to the third to second millennium B.C. from four sites in Kazakhstan. Our results provide two inferences concerning the chronology of millet dispersal and its role in the human diet, both contributing to ongoing debates.

Broomcorn and foxtail millet grains from three sites along the IAMC, including Begash, Tasbas, and Dali, were identified. Pottery from Koken, located in the steppes of northeast Kazakhstan, was devoid of millet grain impressions. This absence affirms the current understanding that millet cultivation was not adopted in the steppe zone until a later date. Turning to the three sites with positive evidence in the IAMC, broomcorn millet can be dated to as early as the third millennium BC, which is consistent with previously published macrofossil results. A foxtail millet impression was identified in a Phase 1 context in Dali, dating to the third millennium B.C. Although further scrutiny of stratigraphic information and ceramic designs is needed, this constitutes the earliest evidence of foxtail millet in Central Asia. It also infers synchronous movements of the two Asian millet species. Future research in this domain with more extensive sampling and additional radiocarbon measurements will be helpful to confirm these findings.

Our results underscore the importance of the IAMC as a key pathway for millet dispersal. It is now well-established that broomcorn millet was introduced to the Pontic regions and the Eastern Mediterranean during the second millennium B.C. However, the timing, routes, and crop status of the western dispersal of foxtail millet remain less clear, and our study contributes new insights into this discussion.

The results also inform a possible connection between millet and human consumption. Grain impressions represent a late stage of crop processing, suggesting pottery making in the immediate vicinity of food preparation activities. This resonates with the recent isotopic evidence of human consumption of millet in Kyrgyzstan during the third millennium B.C. Whether millet was introduced to the IAMC initially as human food or as a commodity with other functions remains an open question. Our results shed new light on these connections with human consumption.