Agricultural origins in North China pushed back to the Pleistocene–Holocene boundary

Two grains, common (proso or broomcorn) millet (Panicum miliaceum) (Fig. 1) and foxtail millet (Setaria italica), were fundamental to the development of agricultural societies that eventually evolved into the first urban societies of China between 4500 and 3800 calibrated years (cal.) B.P. (1). Today, these grains are important mainly in parts of Russia, South Asia, and East Asia. How, when, and in what settings these millets initially evolved is not well known (2). One hypothesis holds that common millet was domesticated rapidly in the central Wei river basin shortly after ca. 8000 cal. B.P. (3). Another hypothesis proposes that common millet was domesticated in the Northeast China Liao river basin around the same time (4). In reality, archaeological data have simply not been adequate to resolve the issues surrounding the domestication of millet and the development of the first agricultural communities in North China. Complicating the problem, common millet is also present in Europe ca. 8000–7500 cal. B.P. (2), so this timing opens the possibility that the crop was domesticated more than once. Otherwise, its origins must predate 8000 cal. B.P. The Early Holocene Cishan site in North China is one of several sites considered key to understanding millet domestication and the origin of dry-land agriculture in China, yet the dating and identity of the crops recovered there have never been adequately documented. The study published in this issue of PNAS (5) revisits Cishan, located on a terrace on the western edge of the North China Plain ≈9 km from where the Nanming river emerges from the Taihang mountains. Two outstanding issues regarding the early archaeological record of millet at Cishan first reported nearly 30 years ago (6, 7), their dating and identification, are resolved in the new study.

Fig. 1.

Modern broomcorn millet (P. miliaceum).

Cishan, a Neolithic village at least 400 m square, has an exceptionally-rich pottery, stone, plant, and animal bone assemblage. The pottery consists of plates, bowls, and pots decorated by cord-wrapped paddle and incising/trailing; none of it is painted. No specific precursors are known. The houses and pits are divided into 2 occupations, periods I and II. Excavations revealed 476 pits, among which 376 are rectangular in plan view. Some of the large, oval pits are houses. Most of the pits belong to period II (7). Many of these characteristics are still unique to the Cishan culture but it was the large quantity of silicified millet remnants in 88 of the pits, nearly 80% from period I, that was so surprising. When the millet was first exposed, it decomposed rapidly, leaving some grains that were too fragile to be recovered and the tough chaff whose cells are composed mainly of silica. At first, the grain was thought to be foxtail millet (S. italica) but details of the morphology were never reported and the identification has not been universally accepted. The age of the site was also surprising considering its complexity: ≈8000 cal. B.P. Archaeologists were uncertain how long the site was occupied, if the millet was as old as the charcoal that provided the radiocarbon dates, or if, indeed, the site was as old as they surmised. Lu et al. (5) report that some of the pits are at least 2,000 years older than previously thought and the plant remains are composed primarily of common millet rather than foxtail millet.

To make an incontrovertible case for the age and association of crops, archaeologists require secure dating, well-documented identifications, and proper documentation of the context from which the dates and the crops were obtained. Lu et al. (5) obtained 46 new samples of plant remains and 9 accelerator mass spectrometry (AMS) radiocarbon dates directly on the millet. Two of the AMS dates are younger and 7 are older than the non-AMS radiocarbon dates on charcoal from the original excavations. The structure and contents of the site appear to be consistent with the extended chronology. Archaeologists suspected that the site has a longer history, and in 2008 Lu et al.'s team reviewed the stratigraphy at Cishan and learned that the opening of pit CS-I was just at the bottom of the Holocene soil layer, while the mouth of the younger CS-II pit was excavated into slightly shallower Holocene soil. The stratigraphy structures of the site appear to be consistent with the extended chronology (Lu et al., personal communication). Three of the older dates extend the occupation to 10,400–10,100 cal. B.P. The dates also show for the first time that the community spanned a period of ≈3,000 years (ca. 10,400 to 7500 cal. B.P.).

How were the crops identified? The remains are extremely fragmentary, being the remnants of chaff. Identification is relatively straightforward when the actual grains are present but in this case Lu et al. (5) needed some ingenuity to identify the specimens. They used 3 techniques to clarify that broomcorn millet was the only crop in period I and the primary crop in period II. A small quantity of foxtail millet is present in period II, indicating that it was a later addition to the crop assemblage and was not a significant crop there. The first identification method, an examination of the epidermal cells of the husks, involved measuring the amplitude of the undulations of the long cells and statistically testing for significant differences. The measurements have a bimodal distribution, indicating that, indeed, the epidermal cell measurements form 2 clusters, 1 for common millet and 1 for foxtail millet. Furthermore, the morphology of silica bodies (phytoliths) in chaff epidermal cells have been used to distinguish plant taxa and evaluate their domesticated status. In a related study Lu et al. (8) demonstrate that phytoliths can, indeed, distinguish foxtail and broomcorn millet. The phytoliths are consistent with the identifications based on epidermal cell morphology. Furthermore, the analysis adds to the understanding of the sophisticated storage at Cishan. Leaves of a wild grass, Phragmites australis, along with millet chaff, were layered on the bottom of the pits. Finally, to alleviate any potential doubt that the majority of the Cishan millet is broomcorn millet, Lu et al. sought biomarkers that could distinguish the millets. They discovered that 5 biomarkers distinguish broomcorn and foxtail millet. The biomarker analysis confirms the common millet identification.

Why are the new Cishan site data significant? Understanding developments in North China has been frustrated by the seemingly sudden appearance of millet production ca. 8000–7500 cal. B.P. in an area ≈1,000 km east–west (Yuezhang to Dadiwan) and the same distance north–south (Xinglonggou to Peiligang) (3, 4, 9, 10). Cishan now has the oldest clear evidence for a significant degree of food production in China instead of being one of several sites in North China where millet suddenly appeared over an exceptionally large area. Furthermore, the 3,000-year occupation by a food-producing community is a longevity rivaled elsewhere at the time only by sites such as Abu Hureyra in Syria (11). We still do not know how continuous the occupation was, but even if Cishan was occupied intermittently, it was still clearly an important location for early hunter–gatherer–fisher–farmers. Cishan represents an extremely successful adaptation to an area bounded on the west by mountains and the loess plateau and situated in a riverine habitat that provided a rich array of resources that included at least 23 species of mammals, tortoise, birds, fish, and shellfish (6). The location likely mitigated some of the effects of the early occupation period that was cooler and drier than it was after 8700 cal. B.P. However, rhesus monkey and masked civet were hunted and are evidence that the climate was warmer than it was in the late 1900s, but it is possible that these animals are associated with period II. Plant exploitation and anthropogenic habitats must have been extremely important at Cishan but neither has been thoroughly investigated.

Cishan now has the oldest clear evidence for a significant degree of food production in China.

The events at Cishan for the first time place the timing of agricultural origins in North China at the Pleistocene–Holocene transition, similar to its timing in Southwest Asia, Mexico, and South America. How the circumstances of agricultural developments compare are still open to question. Despite the relatively continuous Upper Palaeolithic through Neolithic record in Southwest Asia (12) the earliest stages of agriculture there are not yet clear. What we do know is that the shift to agriculture is evident only by the Pre-Pottery Neolithic B (10,500–8200 cal. B.P.) (13, 14), the earliest dates of which correspond to the oldest dates at Cishan. However, in contrast to Southwest Asia, pottery had developed in China long before obvious traces of agriculture, villages, and storage pits. In South America some of the oldest traces of agriculture date between 11,000 and 6000 cal. B.P (15). Mexican agricultural origins appear not to be associated with villages or pottery (16, 17). A key exception to agriculture appearing during the Pleistocene–Holocene transition is eastern North America where native plants were being domesticated between 5000 and 3800 cal. B.P (18). However, like many other regions, the setting was relatively resource rich; populations were not struggling for food. Cishan adds another variant to the comparative mix of developmental puzzles that are the origins of agriculture (19).

The new study (5) indicates that we still do not have quality data pertaining to the onset of millet domestication and agriculture in North China because food production was already established by 10,400–10,100 cal. B.P. at Cishan. If our understanding of Cishan is being significantly revised, what about the other Early Neolithic sites in North China reported in the 1970s and 1980s? An issue that is still unresolved is foxtail millet domestication. The new data are consistent with other reports that it developed later than broomcorn millet, but we still need to be careful about Cishan because so few of the plant remains have been examined. We may yet learn that other pits contained more foxtail millet or other crops. Furthermore, so few plant remains assemblages have been recovered from the period between 10,000 and 7500 B.P. in North China that we may still learn that the 2 millets were domesticated independently. Crop complexes may form well after individual taxa were separately domesticated (18). We still await a comprehensive study of subsistence and human–environment interaction at Cishan but with this new study (5) the development of agriculture in North China is becoming clearer.