Early Pottery at 20,000 Years Ago in Xianrendong Cave, China

See allHide authors and affiliations

Science  29 Jun 2012:
Vol. 336, Issue 6089, pp. 1696-1700
DOI: 10.1126/science.1218643

Pots and Crocks

The invention of pottery allowed for more secure storage of food than was provided by baskets or hide pouches, and the vessels could also be used in cooking. The earliest pottery has been thought to have appeared in China and Japan ∼18,000 years ago, several thousands of years before the advent of agriculture. Wu et al. (p. 1696); see the Perspective by Shelach) have now dated broken pieces of pottery from a cave in China, the earliest of which date to ∼20,000 years ago, the time of the Last Glacial Maximum. Scorch marks on many pieces imply that the pottery was used in cooking.


The invention of pottery introduced fundamental shifts in human subsistence practices and sociosymbolic behaviors. Here, we describe the dating of the early pottery from Xianrendong Cave, Jiangxi Province, China, and the micromorphology of the stratigraphic contexts of the pottery sherds and radiocarbon samples. The radiocarbon ages of the archaeological contexts of the earliest sherds are 20,000 to 19,000 calendar years before the present, 2000 to 3000 years older than other pottery found in East Asia and elsewhere. The occupations in the cave demonstrate that pottery was produced by mobile foragers who hunted and gathered during the Late Glacial Maximum. These vessels may have served as cooking devices. The early date shows that pottery was first made and used 10 millennia or more before the emergence of agriculture.

Pottery making—the manufacture of fired, ceramic container forms—differs considerably from the baked clay figurines or small objects known from the Upper Paleolithic period (1) in its technological demands and in its significance both in subsistence activities, including food storage, processing, and cooking, and in social interactions (2). Pottery was until recently thought to have been developed during the so-called “Neolithic Revolution” and first made by settled, farming populations with domesticated plants and animals and ground stone tools, but recent discoveries have found earlier examples, from Late Pleistocene mobile or semimobile hunter-gatherer contexts in China, Japan, and the Russian Far East (2). One notable find, dating to ~18 to 17 thousand calendar years before the present (cal ky B.P.), is at Yuchanyan Cave (Hunan, China) (35). Here, we describe and date earlier pottery from Xianrendong Cave (Jiangxi, China).

Xianrendong Cave (28°44'10.05″N; 117°10'23.15″E) is located in Wannian County, northern Jiangxi Province, China, some 100 km south of the Yangtze River. The cave consists of a large, inner hall with a small entrance, ~2.5 m wide and 2 m high (Fig. 1). Xianrendong was excavated in 1961 and 1964 by Li (6, 7), by Sino-American joint expeditions in 1993 and 1995 (8, 9), and by Peking University and Jiangxi Provincial Institute of Cultural Relics and Archaeology in 1999 and 2000 (10). The excavations uncovered a long Late (or Upper) Paleolithic sequence, with a rich assemblage of stone, bone, and shell tools; animal bones; phytoliths; and pieces of locally made pottery vessels [(611) figs. S1 to S9].

Fig. 1

Site map of Xianrendong showing the locations of the west and east sections reopened and sampled in 2009. Modified from (10) with permission. (Inset) The location of the cave in South China.

The prehistoric deposits at Xianrendong are located in front of the cave hall entrance. For this study, in 2009 we reopened two trenches from the earlier excavations, here labeled as the “east” and “west” sections from their positions on either side of a modern path leading into the cave entrance (Fig. 1). The numbering of the layers here follows the original labeling: from top to bottom, layers 1 to 4B in the west section (Fig. 2 and fig. S10) and layers 1 to 6B in the east section (Fig. 3). There is no stratigraphic correlation between layers with the same numbering across the two trenches. Pottery sherds were found in previous excavations in layers 1A to 3C1B in the west and in layers 1A to 2B in the east, as well as in what the original excavators labeled as archaeological “features” but which actually include layers and lenses, and so profiles were redrawn in the field in 2009 (Figs. 2 and 3 and fig. S12).

Fig. 2

The stratigraphy of the Xianrendong cave west section. Modified from (10) following field observations made in 2009. Dates indicated are calibrated cal yr B.P. dates calculated by CalPal_HULU 2007. For full information, see Table 1.

Fig. 3

The stratigraphy of the Xianrendong cave east section. Modified from (10) following field observations made in 2009. Dates indicated are calibrated cal yr B.P. dates calculated by CalPal_HULU 2007. For full information, see Table 2.

Although Xianrendong pottery was known to be Late Pleistocene in age, with only a limited number of radiocarbon determinations from the original excavations and no study of the complex formation processes of the cave’s deposits, uncertainty persisted over the age of the earliest ceramics. We thus gathered systematically a new series of samples for radiocarbon determinations from the reopened and cleaned sections. We removed blocks of sediments for micromorphological analysis from the exposed sections concomitant with the collection of radiocarbon samples in order to establish the contextual integrity of both the pottery and the samples collected for dating (figs. S1 to S9) and to verify the integrity of the pottery-containing levels as recorded in earlier field observations (1214).

Some 282 pottery sherds were retrieved during the 1993 excavations at Xianrendong, from contexts below the mixed layer 1A (figs. S1 to S8). We did not recover any sherds from the reopened sections but identified one piece in micromorphological sample 6 (figs. S1 to S9). All pottery is typically tempered with crushed quartzite or feldspar. Firing of the thick, more crudely made earliest pottery was probably carried out at relatively low temperatures in open fires. The earlier pottery is plain-surfaced or cord-marked, but some, from layer 3C1B, have parallel striations on the interior and exterior surfaces, probably from smoothing with grass fibers (fig. S1). Although no vessels could be reconstructed, they had rounded bottoms with walls 0.7 to 1.2 cm thick. Two vessel-forming techniques can be identified through visual observation: sheet laminating and coiling with paddling. Many sherds bear signs of burning on their exterior surface, possibly indicating their use in cooking. From a series of in situ bone fragments that we collected from the exposed profiles in the east and west trenches, we selected fragments larger than 1 cm for dating. We also selected similar fragments that were excavated previously. Because more than 90% of the bones recovered in Xianrendong were of deer—the largest mammal in the assemblage—most probably the thick fragments we used for dating were those of this group, although we could not identify specific species. Bone was chosen because it is short-lived, and we dated fragments of this size because it is unlikely that the stratigraphic integrity of charcoal or bone samples of this size could have been disturbed after deposition (see below). Dating was done in the radiocarbon facility of Peking University (methodology is presented in the supplementary materials, section S3). In all, 45 samples have been dated. Thirteen of these were collected from the reopened sections in 2009, the dates of which are tabulated here and compared with samples collected in the 1999 and 2000 excavations and previously dated samples from the 1993 and 1995 excavations (810). Samples were measured by the laboratories of Peking University, University of California, Riverside, and the University of Arizona (Tables 1 and 2).

Table 1

Radiocarbon determinations from Xianrendong west section. Date column entries obtained by using Libby half-life. Calibration was done by using CalPal-HULU.2007 version, 1-Standard Deviation. Dates marked with an asterisk are cited by (8). Layer 3C1B contained the earliest pottery in this section. Laboratory numbers beginning with BA indicate those from Peking University radiocarbon laboratory; UCR, University of California, Riverside Radiocarbon Laboratory; AA, NSF-Arizona AMS Laboratory, Tucson. Ch indicates a charcoal sample.

View this table:
Table 2

Radiocarbon determinations from Xianrendong east section. Dates, calibrated age ranges, asterisks, and laboratory abbreviations as in Table 1. Layers with the earliest pottery are 2B1 and 2B.

View this table:

The radiocarbon dates suggest that the cave was in use with minor chronological gaps first from ~29,000 through Last Glacial Maximum (LGM) times until ~17,500 cal yr B.P. It was then abandoned and reoccupied from ~14,500 through 12,000 cal yr B.P. The earliest pottery appears in the Xianrendong sequence in layers 2B and 2B1 in the east trench and layer 3C1B in the west trench. The radiocarbon dating shows that both of these early contexts date to ~20,000 to 19,000 cal yr B.P. (Tables 1 and 2).

In order to assess the integrity and preservation of the layers and the associated samples, we studied thin sections of 24 micromorphological samples collected from the west and east trenches (1215) (supplementary materials section S2). Although there are differences between the depositional sequences of each trench, the specific sample fabrics imply that the layers in both trenches had remained stable since deposition, with only minimal cracks on a scale of millimeters or centimeters. These are not large enough to affect the large bone fragments used for radiocarbon dating. In addition, the presence of intact ice lensing in layer 4B, which probably formed during the LGM, is further proof that the sediments have not been significantly reworked.

Alluvial sediments are present in the earliest layers of the sequence in both trenches, before the appearance of pottery (levels 4A and 4B in the west and 3A in the east trench). Above these sediments, the deposits in the west trench are overall similar, with minor changes in texture, composition, and fabric. They consist of moderately to poorly sorted sandy silty clay, with mm- to cm-sized inclusions of rock fragments derived from the cave’s roof and walls. They represent a mixture of moderately sorted low-energy alluvial overbank deposits with anthropogenic contributions (charcoal, bones, sherds, and stone artifacts). The presence of bedding in layer 3C1B, where early pottery sherds, bones, and stone artifacts were found, indicates that these deposits are intact [samples 5 and 6 (16)].

The deposits in the east trench differ markedly from those of the west trench. They are generally calcareous, except for layer 3A, which is situated below the pottery bearing layers. Layer 3A [lower half of sample 19A (16)] is micaceous and strongly resembles the sediments from the west trench. A clear break is noted in the middle of sample 19A (layer 3A, Fig. 2B and fig. S13), where mica-rich sediment below changes to calcareous, ash-rich, and mica-poor sediment above.In these calcareous deposits in the east trench, the calcite is derived mostly from anthropogenic ash (accompanied by some charcoal) rather than from a geological source such as limestone: Both ash and charcoal indicate a lack of alluvial sediment, which is abundant in the western section. The lack of bedding and the virtual absence of mica in the east section in the layers from 2B2 and above suggest that most of the sediments were dumped near the cave wall by humans and were at least partially shielded from fluvial processes. It thus seems that the major occupation or activity areas at the site were located further outside the cave, beyond the excavated zone (Fig. 1), which is characterized by dumped deposits. This conclusion is supported by the lack of any intact combustion features (in spite of the large proportion of calcareous ashes), the absence of traces of bedding or any evidence of individual beds, and the mixing of a variety of materials, such as bone and charcoal: These components are chaotically arranged on a centimeter scale, whereas in occupation deposits they would normally be arranged more contiguously in a lateral direction (supplementary materials section S2).

Evidence of bioturbation by worms or similar-sized fauna, represented by centimeter-sized passage features (15), is common throughout the east trench and increases dramatically toward the top of the profile [e.g., samples 27, layer 2A1, and 28, layer 2A (16)]. Such bioturbation is responsible for the high porosity in many of the samples of the east trench [e.g., samples 27 and 28 (16)]. Bioturbation on this scale might have caused charcoal and bone with sizes of up to a centimeter to move, and thus millimeter-sized pieces of charcoal or bone would be inappropriate for dating. There is no evidence for burrowing by larger animals such as rodents. Thus, we conclude that it is unlikely that either pottery sherds or the larger bone fragments that were used in dating were displaced in either section. Individual bone-rich layers can be observed in some of the thin sections [e.g., samples 19A, layer 3A; 22B, layer 2A3; 26, layer 2A2; and 29, A and B, both feature 4).

The tabulated radiocarbon dates (Tables 1 and 2) and Figs. 1 and 2 demonstrate a high degree of consistency in dating the contexts of the earliest appearance of pottery in the west and east trenches. In the west section, layer 3C1B, where the earliest sherds were uncovered, is ~10 cm thick and has five dates ranging from 20,867 ± 318 (AA15005) to 19,283 ± 283 (BA10267) cal yr B.P. The sixth date reported from this layer (UCR3440) of charcoal, calibrated as 22,120 ± 335, is an outlier. Given the observations concerning potential mobility of charcoal flecks (supplementary text S2) and uncertainty about their original location, it was not included in Fig. 2. The five dates of layer 3C1A are younger by a few centuries than those of layer 3C2 underneath and may indicate the particular time when making pottery began. A date of 18,327 ± 264 (UCR3300) on human bone from earlier excavations in layer 3C2 is likely intrusive; the same explanation holds for UCR-3561 (14,653 ± 550 cal yr B.P.) in layer 3B1 above. Layer 3CA1, directly above layer 3C1B, is definitely younger, with dates ranging from 19,577 ± 389 (BA95143) to 17,128 ± 195 cal yr B.P. (BA09875) (Fig. 2).

Dating in the east section is consistent with that in the west. In the east, layers 2B to 2B1, with the earliest pottery, contain 10 dates ranging from 20,902 ± 358 (BA95140; from the earlier excavations) to 19,166 ± 219 (BA10263) cal yr B.P. The overall thickness of layers 2B to 2B1 is about 15 cm (Fig. 3). Above this, layers 2A3 and 2A2 produced somewhat similar dates, after which there is a depositional break. We conclude that the two trenches, 8 m apart and separated by consolidated unexcavated deposits, provide the same chronological indicators for the early pottery in Xianrendong.

Pottery making introduces a fundamental shift in human dietary history, and Xianrendong demonstrates that hunter-gatherers in East Asia used pottery for some 10,000 years before they became sedentary or began cultivating plants (1719). The age for pottery production at Xianrendong of ~20,000 years ago coincides with the peak period of the last ice age, when there was a decrease in the productivity of regional food resources (2022). When used for cooking, pottery allows energy gains from starch-rich food as well as meat (23), and scorch and soot marks on sherd exterior surfaces indicate that Xianrendong pottery likely was used for cooking. Residue, starch, or other physiochemical analyses of recovered early pottery sherds from Xianrendong and other Late Pleistocene sites in China have not been reported, so the exact function of this early pottery remains unknown. The Xianrendong assemblage contained a large number of fragmented bones, so the pottery could have been used in the extraction of marrow and grease (24, 25). Other known uses of pottery in hunter-gatherer societies include food preparation and storage, as well as brewing alcoholic beverages, and could play a vital social role in feasting (26). Thus, the early invention of pottery may have played a key role in human demographic and social adaptations to climate change in East Asia, leading to sedentism, and eventually to the emergence of wild rice cultivation during the early Holocene (17, 27).

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S24

Tables S1 to S4

References (2832)

References and Notes

  1. Detailed information for samples can be found in the supplementary materials, section S2.
  2. Acknowledgments: We thank the State Administration of Cultural Heritage of China for permission to carry out this project and the Jiangxi Provincial Institute of Cultural Relics and Archaeology for their support, in particular C. Fan and G. Zhou. W. Yan (Peking University) and B. Wang (Cultural Relics Bureau, Wannian, Jiangxi) provided assistance and information concerning the earlier excavations. We also thank S. Liu (Jiangxi Provincial Institute) for assistance. Detailed descriptions of the stratigraphy and the pottery fragments are in (17). We thank K. Liu and X. Ding for the atomic mass spectrometry (AMS) radiocarbon measurements. The American School of Prehistoric Research (Peabody Museum, Harvard University) supported the fieldwork and, in part, the preparation of the samples for microscopic analysis. P.G. is grateful to the NSF (grant no. 0917739) for partial support of this project. W.X., Z.C., and O.B.-Y. contributed to conceiving the project and organizing the fieldwork. P.G. and T.A. analyzed the micromorphology of the thin sections. W.X. and P.Y. conducted the dating of the new radiocarbon samples. W.X., O.B.-Y., P.G., Z.C., and D.C. prepared the paper. The authors declare no competing financial interests. Correspondence and requests for materials should be addressed to O.B.-Y. and X.W. (
View Abstract

Stay Connected to Science

Navigate This Article