The origins of specialized pottery and diverse alcohol fermentation techniques in Early Neolithic China

Significance

China is well-known for its distinctive techniques in alcohol fermentation. Here we present archaeological evidence of alcohol making based on analyses of starch granules, phytoliths, and fungi in food residues adhering to 8,000- to 7,000-y-old Neolithic pottery vessels. We demonstrate the earliest association between the wide occurrences of globular jars as liquid storage vessels and the development of two methods of alcohol making: use of cereal malts and use of moldy grain and herbs as starters. The latter method was arguably a unique invention initiated in China. Neolithic people made low-alcohol beverages with broomcorn millet, Triticeae grasses, Job’s tears, rice, beans, snake gourd root, ginger, yam, lily, and so forth. Such fermented beverages may have served social, spiritual, and medicinal functions.

Abstract

In China, pottery containers first appeared about 20000 cal. BP, and became diverse in form during the Early Neolithic (9000–7000 cal. BP), signaling the emergence of functionally specialized vessels. China is also well-known for its early development of alcohol production. However, few studies have focused on the connections between the two technologies. Based on the analysis of residues (starch, phytolith, and fungus) adhering to pottery from two Early Neolithic sites in north China, here we demonstrate that three material changes occurring in the Early Neolithic signal innovation of specialized alcoholic making known in north China: (i) the spread of cereal domestication (millet and rice), (ii) the emergence of dedicated pottery types, particularly globular jars as liquid storage vessels, and (iii) the development of cereal-based alcohol production with at least two fermentation methods: the use of cereal malts and the use of moldy grain and herbs (qu and caoqu) as starters. The latter method was arguably a unique invention initiated in China, and our findings account for the earliest known examples of this technique. The major ingredients include broomcorn millet, Triticeae grasses, Job’s tears, rice, beans, snake gourd root, ginger, possible yam and lily, and other plants, some probably with medicinal properties (e.g., ginger). Alcoholic beverages made with these methods were named li, jiu, and chang in ancient texts, first recorded in the Shang oracle-bone inscriptions (ca. 3200 cal. BP); our findings have revealed a much deeper history of these diverse fermentation technologies in China.

China holds the earliest record for pottery production in the world, represented by about 16 sites dating to ca. 20000–10000 cal. BP (1). These incipient vessels are mainly open-mouthed jars or basins with round or flat bases, some of which may have been used for boiling or processing various plant foods (2). By the Early Neolithic Period (ca. 9000–7000 cal. BP), which is defined in this study as the presence of clear evidence for domesticated millet and rice, ceramic vessels became increasingly diverse in type, forming several regional traditions (3). The new vessel types included tall cylindrical jars with or without legs, globular jars, bowls, and cups, of which globular jars are among the most widespread type, distributed over a broad region from the Yellow River to the Yangzi River valleys (Fig. 1) (4). A globular jar is normally constructed with a restricted mouth, a short neck, and a large spherical body. Such a vessel form is suitable for storing liquid and is commonly used for fermenting alcoholic beverages, as its narrow neck can be effectively sealed, to exclude as much air as possible and encourage anaerobic conditions (5). Supporting this hypothesis, chemical analyses of sherds of globular jars from Jiahu in Henan (ca. 9000–7500 cal. BP) have revealed the earliest fermented beverages made of rice, honey, and fruits in China (6). However, we know little about the fermentation methods that were used at this time.

Fig. 1.

Globular jars unearthed from major sites in early Neolithic China (9000–7000 cal. BP). 1, Dadiwan; 2, Guantaoyuan; 3, Baijia-Lingkou; 4, Jiahu; 5, Shuiquan; 6, Pengtoushan; 7, Kuahuqiao; 8, Xiaohuangshan; 9, Shangshan; 10, Houli.

The production of an alcoholic drink from cereals involves two separate biochemical steps: (i) saccharification, hydrolysis of the starch in the cereal to fermentable sugars with enzymes called amylases, and (ii) fermentation, conversion of the sugars by yeasts to ethyl alcohol and carbon dioxide. Since yeast is ubiquitous in the environments and thus readily introduced incidentally, the key step is saccharification. In ancient China, saccharification may have been achieved by utilizing the amylases in three sources: (i) human saliva (mastication; by chewing and spitting out cereals), (ii) sprouted grains, and (iii) jiuqu 酒麴 (or qu 麴), a fermentation starter made of moldy cereals, in some cases including herbs (herbaceous plants), which are rich with microorganisms (filamentous fungi, yeasts, and bacteria) (7–9). Qu is functional as an agent for simultaneous saccharification and fermentation; during this process, filamentous fungi secrete various enzymes to degrade starch material into fermentable sugars, while yeasts convert sugars to carbon dioxide and alcohol. There are several variants of qu, such as daqu 大麴, made of wheat (Triticum sp.), barley (Hordeum sp.), and/or pea (Pisum sativum); xiaoqu 小麴, made of rice (Oryza sativa); and fuqu 麩麴, made of bran (the hard outer layers of wheat grain) (10–12). Herbs used as part of qu are often referred to as caoqu 草麴 (7, 9, 13). Alcohol types first appeared in oracle-bone inscriptions of the Late Shang (ca. 1250–1046 BC), recorded as li 醴, jiu 酒, and chang 鬯 (14). In later texts, such as Shangshu (compiled by the fourth century BC), the fermentation methods of these alcohol types were further described as li made of nie 糵 (sprouted cereals), jiu made of qu, and chang made of millet and herbs. Use of the qu method for alcohol production is generally thought to be unique to East Asia, and was initiated in China (7, 9).

Beer made of sprouted millet has been identified in vessels of three Yangshao culture sites in north China (ca. 5700–4700 cal. BP) (15–17), but whether or not qu was used in the Neolithic is unknown. To investigate the origins of fermentation technologies in ancient China, we analyzed residues adhering to ceramics from two Early Neolithic sites, Lingkou in Lintong and Guantaoyuan in Baoji, both situated in the Wei River valley in Shaanxi Province (SI Appendix, Fig. S1). The analyses revealed abundant starch granules, phytoliths, and fungi. We have conducted a series of experimental studies on starch morphological changes caused by fermentation, using different cereals and tubers (18). We also studied traditional millet beer brewing in north Shaanxi (SI Appendix, Fig. S2). The fermented products from these projects were used as comparative references. Here we present the results to address the following issues: (i) the relationship between the occurrence of globular jars and the development of alcohol production in Neolithic north China; (ii) the regional variations in terms of ingredients in alcohols and brewing methods; (iii) the medicinal functions of early alcohol; and (iv) the connections between the spread of cereal domestication and early alcohol production.

Lingkou (LK hereafter), excavated in the 1990s, is located on the banks of the Ling River, a tributary of the Wei River. This area is likely to have had abundant water resources, as indicated by the presence of numerous water buffalo remains at the contemporaneous Baijia site in the vicinity (19). The material deposits at LK indicate a long sequence of human occupations from the Early to the Middle Neolithic Period. The pottery types from the earliest deposits include bowls, tripod vessels, flat-based jars, and globular jars, all belonging to the Baijia culture (ca. 7900–7000 cal. BP) (20). We analyzed six vessel sherds, including three globular jars and three tripods; the former is hypothetically related to alcohol making, whereas the latter may have been used for more general culinary purposes (SI Appendix, Fig. S1).

Guantaoyuan (GTY hereafter), excavated in the 2000s, is ∼300 km west of LK. It is situated on a high platform in a small basin near the Wei River. The earliest material deposits date to the Baijia Period (7800–6900 cal. BP) (21), from which we sampled 14 pottery and stone objects, which can be classified into three functional groups. Group I (n = 6) consists of vessels hypothetically associated with alcohol making, including five globular jars (POT1–5) and one perforated basin (POT9) likely used as a funnel and steamer (see discussion below). Group II (n = 4) consists of vessels for general food storage and serving, including one tripod (POT6), one jar (POT7), and two bowls (POT11 and 12). Group III (n = 4) consists of four grinding stones (GS1–4) for food processing (SI Appendix, Fig. S1). Neither macrobotanical nor pollen analysis has been carried out at the sites, and thus our study of these artifacts provides comparative datasets for general food preparation and consumption. Together, we analyzed residue samples extracted from 20 artifacts, of which the pottery vessels are among the earliest known ceramics in the Wei River Valley. We also analyzed eight control samples.

Results

Starch Remains.

A total of 1,128 starch granules were recovered from the artifacts, and 744 of them (66%) can be identified as one of eight types, corresponding to certain plant taxa (Fig. 2, Table 1, and SI Appendix, Fig. S3).

• Type I: Panicoideae (n = 394; 34.9% of the total; 100% ubiquity), which may include broomcorn millet (Panicum miliaceum), foxtail millet (Setaria italica), and Job’s tears (Coix lacryma-jobi).

• Type II: Job’s tears (n = 27; 2.4% of the total; 40% ubiquity), which show diagnostic characteristics specifically of Job’s tears and can be clearly separated from millets (22).

• Type III: Triticeae (n = 234; 20.7% of the total; 90% ubiquity), likely wild grasses, such as the genera of Agropyron, Elymus, and Leymus, which are native to north China (23).

• Type IV: rice (Oryza sp.; n = 25; 2.2% of the total; 5% ubiquity), almost certainly of domesticated form, since no evidence suggests that the Wei River region was within wild rice’s (Oryza rufipogon) endemic range (24).

• Type V: snake gourd (Trichosanthes kirilowii; n = 26; 2.3% of the total; 25% ubiquity), commonly found in north China (25); its roots were used as food and medicine in ancient China (26).

• Type VI: ginger (Zingiber sp.; n = 5; 0.4% of the total; 20% ubiquity), whose roots have been used as spice and medicine since ancient times in China (see discussion below).

• Type VII: underground storage organs (USOs; n = 27; 2.4% of the total; 40% ubiquity), which probably include yam (Dioscorea polystachya) and lily (Lilium sp.), widely distributed in China (27). Most type VII starches are badly damaged, and are not diagnostic to specific taxa.

• Type VIII: bean (Fabaceae; n = 6; 0.5% of the total; 25% ubiquity), perhaps wild species of Vicia pea. There are 17 species of Vicia peas growing in Shaanxi (28). Notably, domesticated pea (P. sativum) is one of the ingredients often used for making daqu starter today (10) (see SI Appendix, Fig. S3 for the description and modern reference images of diagnostic features of each starch type).

Fig. 2.

LK and GTY starch types. 1, type I, Panicoideae, likely broomcorn millet; 2, type II, Job’s tears; 3, type III, Triticeae; 4, type IV, rice; 5 and 6, type V, snake gourd root; 7 and 8, type VI, ginger (LK and GTY); 9, ginger starch from Mijiaya for comparison; 10, type VII, USO, possibly yam; 11, type VII, USO, possibly lily; 12, type VIII, bean (1–4 and 7, LK; 5, 6, 8, and 10–12, GTY) [each starch is shown in DIC/bright-field (Left) and polarized (Right) views].

Table 1.

Lingkou and Guantaoyuan starch record

	Panicoideae	Job’s tears	Triticeae	Rice	Snake gourd	Ginger	USO	Bean
Taxa	I	II	III	IV	V	VI	VII	VIII	UNID	Total	Damaged	Gelatinized 1	Gelatinized 2
LK globular (n = 3)	27	4	9	25	3	1			30	99	19	34
LK tripods (n = 3)	19	10	20					2	54	105	61	35	1
LK total, N	46	14	29	25	3	1		2	84	204	80	69	1
LK total, %	22.5	6.9	14.2	12.3	1.5	0.5		1.0	41.2	100	39.2	33.8	0.5
LK ubiquity, %	100	50	100	16.7	16.7	16.7		16.7	100		100	100	16.7
GTY globular-funnel (n = 6)	216	13	148		23	4	14	4	250	672	234	202
GTY general pots (n = 4)	55		3				6		7	71	39	9
GTY grinding stones (n = 4)	77		54				7		43	181	118	11
GTY total, N	348	13	205		23	4	27	4	300	924	391	218	4
GTY total, %	37.7	1.4	22.2		2.5	0.4	2.9	0.4	32.5	100	42.3	23.6	0.4
GTY ubiquity, %	100	42.9	71.4		28.6	21.4	57.1	28.6	85.7		100	78.6	7.1
LK&GTY, total	394	27	234	25	26	5	27	6	384	1,128	471	292
LK&GTY, %	34.9	2.4	20.7	2.2	2.3	0.4	2.4	0.5	34.0	100	41.8	25.9
LK&GTY ubiquity, %	100	40	90	5	25	20	40	25	90	100	100	95

Among these plants, starches of millet, Job’s tears, Triticeae grasses, snake gourd root, yam, lily, and beans have all been found on grinding stones from Upper Paleolithic and Neolithic sites in north China (29, 30).

A large majority of the starches (n = 763; 67.6%) show signs of morphological alterations, which can be classified into two types, each probably representing different food-processing techniques. The first is damaged starches without gelatinization (n = 471; 41.8% of the total; 100% ubiquity). They show random pitting, deep channels, broken edges, missing or pronounced lamellae, central depression, and/or disappearing extinction crosses under polarized light. These features are consistent with the morphological changes caused by enzymatic attack, which are commonly found on malted cereal starches (Fig. 3, 1–3). Some of these kinds of damage also occur due to grinding, such as broken edges (Fig. 3, 4). The second is damaged starches with gelatinization (n = 292; 25.9% of the total; 95% ubiquity). The gelatinized starches can be further divided into two groups: (i) Most granules exhibit moderate swelling with a hollowed center, resembling a diagnostic damage type caused by low-temperature mashing and fermentation, representing a majority of the granules (Fig. 3, 5 and 6 and SI Appendix, Fig. S3, 9–13); this type of modification is different from other food-processing techniques (e.g., steaming and boiling), based on our experimental study (18) and published data (31); and (ii) few granules (n = 5; from LK-POT4 and GTY-POT9) show expansion without evidence of enzymatic attack, similar to those from cooked (boiling and steaming) as well as mashed cereals (Fig. 3, 7 and SI Appendix, Fig. S3, 14 and 15). GTY-POT9, a perforated vessel, revealed both types of gelatinized starches. It may have been used as a funnel for processing fermented beverages, but could also be used as a steamer.

Fig. 3.

Damaged and gelatinized starches from LK and GTY pottery. 1, millet starch with central depression and broken edge; 2, millet starch with central depression and deep channels; 3, Triticeae starch with deep channels, pronounced lamellae, and pitting; 4, Triticeae starch with missing part (perhaps caused by grinding), central depression, and pitting; 5, UNID starch with central depression, pitting, and partial gelatinization; 6, fermented starch expanded with hollowed center and birefringent edge; 7, a cluster of Triticeae starch showing different stages of gelatinization, some with central depression (pointed with stemmed arrow), some flattened and expanded (pointed with nonstemmed arrow), and some still intact; 8, gelatinized starch stained with Congo red, showing red in bright-field light and golden glow in polarized light (1 and 2 from LK; others from GTY; each starch is shown in DIC and polarized views).

Some damaged features of starch could be the result of microbe activity in the postdepositional environment (32), but the damaged starch granules from seven control samples were much lower in frequency (n = 1–5) than residue samples (n = 1–92), and no gelatinized starch was present in five of the control samples (SI Appendix, Table S1). To confirm the presence of gelatinized starch, we applied the Congo red staining method (33) on four samples (GTY-POT2, 3, 5, and 9), which all revealed gelatinized starches, characterized by their red color under the bright field, and orange-red or gold-green glow under polarized light (Fig. 3, 8; see SI Appendix for the staining method). More than one-third of the starches (n = 384; 34%) are not identifiable (UNID) due to their lack of diagnostic features, or due to being severely damaged.

Of the eight plant taxa identified, six are present in both the LK and GTY assemblages, including Panicoideae, Job’s tears, Triticeae, snake gourd root, ginger, and bean. Panicoideae and Triticeae are the most numerous in count and in ubiquity. Rice was only found in LK, and yam and lily only in GTY (Table 1 and SI Appendix, Table S1).

Phytolith Remains.

The LK samples yielded a large quantity of phytoliths (n = 389; SI Appendix, Table S2), with the majority derived from millet and rice husks. Phytoliths from Paniceae husks predominate the assemblage (n = 217; ubiquity 100%), including 28 η-type phytoliths (Fig. 4, 7) from broomcorn millet, one Ω-type phytolith (Fig. 4, 6) from foxtail millet (34), and 188 from undetermined millets (Fig. 4, 14) (35, 36). Such a pattern suggests that broomcorn, rather than foxtail, was likely the main type of millet being processed in the LK vessels. Double-peak phytoliths (Fig. 4, 5) (n = 13) from rice husks also occurred in high ubiquity (66.7%, POT2–5), suggesting that the crop might have been commonly consumed at the site. These observations are consistent with the results of staple isotope analysis of human bones from the nearby Baijia site, indicating that human diets were composed of both C3 and C4 plants (37), such as rice and millet. Other Poaceae family morphotypes mainly include cross, bilobate, rondel, and common bulliform (Fig. 4, 1–4 and 8). Cross phytoliths show a considerable variation in form and size, some larger than 18 µm in width (Fig. 4, 1), which are most comparable to the large variant 1 cross type produced by Job’s tears (38). One raphide is present (39), likely derived from a USO or external contamination. The profile of LK phytoliths largely corroborates the starch granule assemblage, indicating the presence of millets, rice, Job’s tears, and tubers.

Fig. 4.

Phytoliths from LK and GTY. 1, cross (Job’s tears); 2, cross; 3, bilobate; 4, rondel; 5, rice double peak from husk; 6, foxtail millet husk; 7, broomcorn millet husk; 8, bulliform; 9, Asteraceae opaque perforated platelets; 10 and 11, hair cells; 12 and 13, raphides; 14, Paniceae husk; 15, skeleton of long cells from grass leaf (1–8, 13, and 14 from LK; others from GTY).

The phytolith assemblage from GTY residue samples shows a clearly different profile. Among the 387 recovered phytoliths, none were identified as millet husk, whereas the majority were elongate cell phytoliths from grass stems/leaves (n > 240) and hair cells (n = 59). Raphides (n = 4) were also found, consistent with the presence of USO starches, possibly from yam. One opaque perforated platelet was recovered (POT2), consistent with those from Asteraceae inflorescences (40) (Fig. 4 and SI Appendix, Table S3). The absence of husk phytoliths indicates that cereals were dehusked before processing in the pots.

Fungi: Molds and Possible Yeasts.

Numerous fungus particles were present in the GTY assemblage, including spores, sporangia/vesicles, hyphae, and possible yeasts. We examined eight residue samples (two globular jars, a tripod, a jar, the funnel-steamer, a bowl, and two slabs) and five control samples, and recorded 1,001 individuals and particle clusters of fungi. Fungi were most abundant in globular jars, the funnel-steamer, and the bowl (range 114–296), and were less common in other vessels and grinding stones (range 9–36), with the lowest frequencies from the control samples (range 2–10). When comparing fungus counts in the residues with those in the controls from the same vessels (two globular jars and one bowl), the ratios are 55:1, 30:1, and 44:1 (SI Appendix, Fig. S4A). This pattern suggests very high concentrations of fungi associated with the functional areas of these vessels.

Some fungi were consistent morphologically with those from Aspergillus and Rhizopus, which are among the most commonly used species in modern qu starter (10, 11). For example, several fungi show a round head with a long stem, resembling a vesicle with conidiophore from Aspergillus (Fig. 5, 1 and 2 compared with 9 and 10 and SI Appendix, Fig. S4B). Some spores (ca. 10 µm in diameter) are round and still interconnected, likely from a chain of conidia, which is characteristic of Aspergillus; another abundant spore form is very small (ca. 3–5 µm), ovoid in shape, consistent with sporangiospores in Rhizopus (Fig. 5, 7 and 8 compared with 12–15 and SI Appendix, Fig. S4C). Some spores appear in germination (Fig. 5, 5). Hypha fragments were numerous, many entangled, similar to mycelium (Fig. 5, 3, 4, and 6 compared with 10 and 11), but they cannot be identified to the genus level (see SI Appendix for further explanation).

Fig. 5.

Molds and possible yeast cells from GTY compared with modern references. GTY samples: 1 and 2, vesicle/sporangia with stipe, without phialides/spores attached (POT2), compared with 9 and 14; 3, cf. sporangiophores growing out of rhizoids (POT1), compared with Rhizopus in SI Appendix, Fig. S4C; 4, hyphae (POT2), compared with 11; 5, cf. spores in germination (POT9); 6, a cluster of hyphae (POT9), compared with 10 and 11; 7, spores in chain (POT2), compared with chained conidia from A. oryzae (12); 8, cf. sporangiospores released from sporangia (POT9), compared with Rhizopus (15); 17 and 18, possible yeast cells in the initial budding process (POT2 and 9). Modern samples: 9, A. oryzae conidial head; 10, conidial head with a cluster of hyphae; 11, a group of conidiophores with vesicles; 12, conidia in chain; 13, Rhizopus sp. sporangia showing sporangiospores; 14, a group of sporangia with and without sporangiospores entangled with hyphae; 15, sporangia releasing small, ovoid sporangiospores; 16, sporangia without sporangiospores; 19, wild S. cerevisiae yeast in budding, Shimao millet beer; 20, cultured, domesticated S. cerevisiae yeast in various budding forms.

Some particles identified as possible yeast cells (n = 21) were present in the globular jars, the funnel-steamer, and the bowl. They are subround or oval in shape, 4.9–11.75 µm in size. Several show a small protuberance on the parent cell. These protuberances are a frequent feature of both budding and mating processes of Saccharomyces (41). They are morphologically comparable to wild Saccharomyces cerevisiae yeast in the modern millet beer from Shimao, north Shaanxi (3.47–12.16 µm in size), but larger than a cultured S. cerevisiae strain (2.64–8.83 µm) in our database (Fig. 5, 17–20 and SI Appendix). Further testing with DNA and other diagnostic biomolecules will help verify this tentative identification, as S. cerevisiae is the most commonly occurring yeast species in modern Chinese alcohol production (10, 11), and China is likely the center of origin of its domesticated populations (42).

Other types of fungus spores were also found in the GTY samples, some consistent with Trichothecium and Alternaria (41). However, these are less abundant and do not represent known components of modern qu, making interpretation more difficult.

Summary.

Millets (mainly broomcorn millet) are present in all artifact groups at both sites with the highest percentage and ubiquity, followed by Triticeae and Job’s tears. These plants are likely to have been the main sources of staple food. Snake gourd root, bean, and ginger are also present at both sites, but represented at much lower levels. Rice starch and phytoliths are present only at LK, while yam and lily starches are only found at GTY.

Damaged and gelatinized starch granules are present in all vessels, suggesting that they were in contact with cooked and/or fermented foods. The gelatinized starches are particularly high in percentage among the fermentation vessels (13–70%). On the other hand, grinding stones revealed a very high proportion of damaged starch (44–74%) but a very low percentage of gelatinized ones (0–17%), a composition consistent with their function for grinding (SI Appendix, Fig. S5C and Table S1). When comparing the starch counts between residue and control samples on the same vessels, the former yielded much greater numbers, about 12–22 times of the latter (SI Appendix, Fig. S5 A and B), suggesting that most starch granules in the residue samples are related to the original function of the artifacts.

While cereal husks are absent, fungi are present in abundance at GTY. They are especially numerous in the globular jars, the funnel-steamer, and the bowl, which are likely to have been relevant to alcohol fermentation and consumption (SI Appendix, Fig. S4A). Fungi naturally exist on leaves, and thus the association of fungi with leaf phytoliths explains why grasses were used as caoqu.

Discussion

Some marked differences are notable between the fermentation-related vessel group (globular jars and funnel-steamer) and other artifact groups. First, the starches with fermentation-related gelatinization were found in highest proportions in the fermentation group (SI Appendix, Fig. S5C), suggesting that the former group is preferentially associated with fermentation-based food processing. Second, some plant taxa are present only in the former group, including snake gourd root and ginger, indicating that they may have had a special function related to alcohol making. Finally, fungi (including possible yeasts) are present in greater abundance in GTY globular jars, the funnel-steamer, and the bowl, indicating that they were likely in more frequent contact with fermented foods.

Globular jars (for effective fermentation and storage) and funnel-steamers (for transferring and filtering liquid, as well as steaming grains) appear to have been developed as a set of equipment used for producing fermented beverages, among other culinary functions. The presence of these new pottery types represents a point of departure from making ordinary porridges with open-mouthed pots before 9000 cal. BP. This change parallels the transition in ceramic forms from open-mouthed bowls to globular jars in sub-Saharan Africa, which has been seen as the development of a millet/sorghum beer-brewing tradition (43, 44).

The combination of a globular jar with a single-perforation funnel-steamer predates the more specialized alcohol-making toolkit consisting of jiandiping amphora, funnel, and multiperforation steamer, which was developed during the Yangshao culture (ca. 7000–5000 cal. BP) (7, 15).

Regional Variations in Alcohol Production.

There were apparently regional differences in plant ingredients used for alcohol making. The presence of rice only at LK is likely attributable to the different ecological conditions of the sites. LK is situated in the alluvial plain with abundant water resources, an environment more favorable for rice growing than GTY, which is located on the highland plateau region with a cooler and dryer climate (SI Appendix, Fig. S1). Samples from additional sites would be beneficial to better establish this contrast.

There were also regional variations in brewing methods, particularly relating to the methods for achieving saccharification. The starch remains from both sites exhibit evidence of likely cereal-based alcohol fermentation, but the phytolith assemblages differ in composition between the two sites. The LK phytolith assemblage consists of a significant proportion of husk phytoliths from Paniceae, broomcorn millet, and rice, suggesting that the fermented beverages were made with unhulled, sprouted broomcorn millet, and probably rice, mixed with additional ingredients (Triticeae, beans, tubers, and probably other materials). The use of sprouted broomcorn millet for beer brewing has indeed been testified to by the residue analyses of pottery vessels from Neolithic Yangshao culture sites at Yangguanzhai, Xinjie, and Mijiaya, all located within an 80-km radius of LK (SI Appendix, Fig. S1) (15–17). An ethnographic comparison of millet-based beer making can be found in Yulin, north Shaanxi, where the local residents use sprouted wheat or maize as the saccharifying agent to brew unfiltered, low-alcohol broomcorn millet beer, called hunjiu 浑酒 (turbid alcohol) (SI Appendix, Fig. S2).

In contrast, the GTY assemblage contains no husk phytoliths, but many from leaves and stems. In addition, GTY samples revealed ubiquitous fungi, some of which are consistent with Aspergillus sp. and Rhizopus sp., commonly found in qu today. During the fermentation process, fungal growth and reproduction (indicated by the development of hyphae and sporangia) require certain conditions, such as the presence of sources of nutrients, a moist environment, and a certain temperature range (10, 12, 45, 46). Such conditions do not naturally occur in normal soil matrix, but may be achieved by using particular types of containers. Therefore, much higher abundancies of fungal particles associated with ubiquitous starches from residues on vessel interiors than in the sediments on exteriors indicate the presence of cereal-based fermentation using qu as a starter (SI Appendix, Figs. S4A and S5 A and B).

According to ancient Chinese texts and ethnohistoric data, diverse grains and plants were used for making qu and caoqu. Taiwanese aboriginals, for example, made qu by leaving cooked grain in the open for days until it became moldy, and then used it as the starter for making millet or rice alcoholic beverages. They also made caoqu with plants from such families as Rutaceae, Fabaceae, Asteraceae, and Chenopodiaceae (7–9, 47). In south China, Polygonum spp. are used as caoqu for making rice wine (13). Many herbs have also been added as ingredients during the fermentation process, including ginseng (Panax ginseng), hemlock parsley (Lingusticum wallichii), and ginger (Zingiber sp.), among others (7). If qu, caoqu, and herbs were used for cereal-based fermentation, we would expect to find damaged and gelatinized starches, phytoliths from leaves, filamentous fungi, and yeasts on the vessel surface. All these are indeed present in the GTY residues, including some herbs (e.g., ginger) traditionally used for fermentation.

Evidently, Early Neolithic people in north China experimented with diverse fermentation methods, which were later elaborated in dynastic times, described in ancient texts as using sprouted cereals (nie) to make li, moldy grains (qu) to make jiu, and herbs (caoqu) to make chang (7, 9). We do not have clear evidence for mastication based on the current data, as this method may not be easily detected archaeologically, though this does not rule out the possibility that it was also used in ancient times. We need to develop methods to test for mastication in archaeological contexts.

Ginger and Medicinal Function of Early Alcoholic Beverages.

The ginger starch granules from LK and GTY are nearly identical to the modern cultivated ginger (Zingiber officinale) from north China (Fig. 2, 7 and 8 compared with SI Appendix, Fig. S3, 6). The wild progenitor of ginger is unknown, and the origin of its domestication is unclear. However, ginger was under cultivation as a spice and a medicinal plant from ancient times in India and China (48). In China, ginger starch granules have been identified at the Neolithic Dadiwan site (ca. 7800–7300 cal. BP), about 200 km west of GTY (SI Appendix, Fig. S1) (49). A ginger starch granule was found in the beer residues on a funnel from Mijiaya (ca. 5000 cal. BP) (Fig. 2, 9), although it was misidentified as yam in the original publication (figure 3E in ref. 15). The earliest ginger root remains have been unearthed in water-logged Wangshan Tomb 2 in Hubei (50), dating to the end of the Eastern Zhou (770–256 BC) (51). Ginger has been traditionally used as an ingredient for making medicinal alcohol, named jiangjiu 薑酒 (ginger alcohol), as recorded in Bencao Gangmu in the 16th century (52). It is conceivable that some of the earliest fermented beverages were made for medicinal purposes, among other functions, beginning a long tradition of alcohol use for health benefits in China.

The Spread of Domesticated Cereals and the Emergence of Alcohol-Making Pottery.

The domestication processes of millet and rice probably initiated in the Yellow River and Yangzi River valleys, respectively, around 10000 cal. BP (53, 54), predating the first appearance of globular jars. In north China, domesticated millets, both broomcorn and foxtail, became widespread by 8000 cal. BP, while rice was less commonly cultivated. In the region north of the Huai River, the arguably earliest domesticated rice remains have been identified at Jiahu (ca. 9000 cal. BP) (55), where the earliest rice-based beer was also recovered (6). The rice starch and phytoliths on the alcohol-making pottery from LK account for the first appearance of this crop in northwest China. Given its association with fermentation-related vessels, rice’s northwestward expansion nearly 8,000 y ago may have been partially related to its importance in alcohol production.

Based on the current data, three material changes occurring in 9000–7000 cal. BP signal the innovation of specialized alcoholic making in some regions in China: (i) the spread of cereal domestication (millet and rice), (ii) the emergence of dedicated pottery types (globular jar and funnel-steamer, among others), and (iii) the development of cereal-based alcohol production with at least two fermentation methods: use of malts and qu starters.

There was a period of 11,000 y from the first simple pottery vessels (20000 cal. BP) to the first alcohol-making globular jars, raising the question of whether people had already experimented with alcohol fermentation before 9000 cal. BP. A pre-Neolithic alcohol is not impossible, as exemplified by the discovery of 13,000-y-old wheat/barley-based beer brewing in stone mortars at Raqefet Cave, Israel (56). It is also unclear what other functions the early globular jars may have had, in addition to alcohol fermentation. More functional studies on Paleolithic and Early Neolithic pottery are needed.

Conclusions

The appearance of new pottery types (globular jars and funnel-steamers) around 9000–7000 cal. BP marks the earliest development of specialized alcohol production. The major ingredients of cereal-based alcoholic beverages include broomcorn millet, Triticeae, Job’s tears, and rice, which were mixed with snake gourd root, ginger, yam, lily, and other plant additives, some likely used for medicinal properties. Different regional traditions in alcohol production may have already been developed at this time. People utilized diverse local plants and experimented with different brewing methods, such as malting millet, making qu with moldy grain, and using herbs as caoqu. These ancient beverages were likely very low in alcohol content, similar to the millet beer hunjiu made by people in north Shaanxi today, but they were the prototypes of ancient alcohol, recorded in the earliest Chinese writing system as li, jiu, and chang. These beverages may have served social, economic, spiritual, and medicinal functions, helping the development of sophisticated cultural and ritual traditions in ancient China.

Methods

Residue samples were processed with protocols established in the Stanford Archaeology Center (see SI Appendix for details). Starch and phytolith identifications are based on our modern reference collection of over 1,100 specimens and published information (34, 40, 57). Fungi in modern qu and koji reference and ancient samples were identified based on descriptions in published sourcebooks (41, 58, 59). Yeast cells were compared with modern samples identified by DNA testing. In brief, modern reference samples had DNA isolated and the resulting material was high throughput-sequenced to identify species, and these were compared with existing reference samples to determine likely origin (see Dataset S1 for the raw data and SI Appendix for more detailed descriptions of methods) (60).

Our laboratory in the Stanford Archaeology Center has been regularly cleaned and checked to prevent starch contamination. We collected eight control samples from artifacts and tested them for starch, phytolith, and fungus. The results show considerably higher counts of microfossils in the residues than the controls, and compositions between the two are often very different, supporting the authenticities of the microfossils in the residue samples (SI Appendix, Figs. S4 and S5 and Table S4).