Following articles by Jiang & Liu on new evidence for the origins of sedentism and rice domestication in the Lower Yangzi (or Yangtze) region in China in 2006 (Antiquity 80: 355-61) and by Fuller, Harvey & Qin placing rice domestication in the fifth millennium BC in 2007 (Antiquity 81: 316-31), the debate has moved on to the Project Gallery, with a response by Liu et al. in 2007 (http://antiquity.ac.uk/ProjGall/liu1/index.html). Here Dorian Fuller and colleagues continue the discussion.

Introduction

We are heartened to see some constructive debate on the study of early rice in China. Liu et al. (2007 http://antiquity.ac.uk/ProjGall/liu1/index.html) have replied to our recent reconsideration of Lower Yangtze rice domestication, but they seem to have conceded our essential points, which were that wild rice was used in prehistoric China, domestication was a long-drawn out process which may have happened in parallel in more than one region, and available evidence from Chinese archaeobotany remains ambiguous, at best, for documenting the rice domestication process. Nevertheless, some of their counterarguments require a few clarifications. A further response would appear to be in order.

The evidence from Shangshan itself remains inadequate. A few impressions are available in a burnt mud clod (ceramic fragment), from which a single spikelet has been measured in full, and from which two spikelet bases have been isolated for examination. Statistically, a sample size of one or two tells us little nevertheless it is worth noting that Zheng & Jiang (2007) have recently suggested that a mixture of wild and domestic type rice might have been present here (although concerns over inadequate sample size apply, and we must await new data).

Mis-readings

Liu et al. have misquoted us as believing Longqiuzhuang Level 4 rice is wild, when we specifically argued that this rice is likely to be domesticated because it shows a marked shift in grain size by comparison to earlier levels at that same site (Fuller et al. 2007: Figure 6). This shift in grain size is consistent both with morphological change under domestication and/or a shift to harvesting of fully mature (and predominantly/wholly non-shattering/domesticated) panicles, or a combination of these processes. By contrast we hypothesise that the earlier levels at the same site have some major input from smaller and/or immature-harvested grains, consistent with wild populations and populations that are not fully domesticated and under pre-domestication cultivation. Liu et al. offer some comparisons with rice grains from other regions, such as Longshan era Huizui. One difficulty with comparing grains across regions is that much variation in rice grain length can be linked to climatic variation: more northerly temperate japonica landraces are short grains, while tropical varieties (the javanica race rices) are massively long. In order to avoid this complication we have tried to focus on changes within a single site (Longqiuzhuang) and local region, Zhejiang (Kuahuqiao and Chuodun). Nevertheless, one can note that the Huizui rices are markedly wider than those of Kuahuqiao, falling in the expected mature/domesticated range.

In our paper we demonstrated from modern data that length:width ratios of rice spikelets (or grains) are insufficient for identifying domesticates (see Fuller et al. 2007: Figure 2). Liu et al. assert that the Shangshan spikelet 'length/width ratio falls into the range of later domesticated rice at Hemudu', but this ratio (~2.7) is also the same as the modal ratio in multiple wild rice species, including Oryza nivara and Oryza rufipogon. We have looked at grain (caryopsis) measurements rather than spikelets (lemma and palea) as these appear to show more distinction (despite overlaps) between species, are more widely observable archaeologically (including on charred grains), and changes can be seen through archaeological time. In addition grains, unlike spikelets, show marked changes during maturation which may also be relevant for tracking domestication processes as we discussed.

Jiahu

Rice was an important element of diet at Jiahu, as the archaeobotanical record indicates, and we did not deny this. At issue is whether rice was cultivated, whether domesticated, and whether the rice here continued into modern landraces. Liu et al. in asserting the importance of rice consumption at this site may have mislead their readers. They cite carbon isotope data (from Hu et al. 2006) as indicating rice consumption: this is a red herring. What these data indicate is a predominance of consumption of foods from plants with C₃ photosynthetic pathway, which is the most common pathway in all plants (Reference SageSage 2004). As such, thousands of plants in Henan province are C₃, not only rice, and amongst C₃ plants are other food plants known from Jiahu, incuding acorns (Quercus spp. sensu lato), Trapa waterchestnuts, wild soybeans (Glycine spp.) and wild grapes (also detected in ceramic residue analysis by McGovern et al. 2004). Possible tuber foods, such as wild yams or lotus roots are also C₃. The only sensible conclusion from the Jiahu isotope data is that these people had little/no C₄ plant intake and were therefore consuming little in the way of millets. This is an interesting conclusion, but irrelevant to the rice issues.

What is important to reiterate, however, is that the people at Jiahu, as also at Hemudu or Kuahuqiao, did not live on rice alone. They used significant quantities of other potential staples from wild sources, such as water chestnuts (Trapa bispinosa) and acorns. This pattern is in stark contrast to later Neolithic periods, such as systematically sampled Late Yangshao or Longshan period sites where samples are heavily dominated by rice (e.g. Crawford et al. 2005) or domesticated millets (e.g. Reference ZhaoZhao 2006; Lee et al. 2007; Fuller & Zhang 2007) and wild fruits or nuts are absent or exceedingly rare. Crops also dominate the record of the Daxi period site of Chengtoushan in the Middle Yangzte, dating to the later fourth and third Millennium BC (Nasu et al. 2007). We believe the heavy use of nuts is consistent with economies that are in the earlier stages of developing agriculture (i.e. low-level food producers, sensu Reference SmithSmith 2001) and are not yet locked into dependence on cultivars as later societies were. One contributing factor in this is that rice may not have been fully domesticated yet and as such provided lower yields; it was one resource amongst many, and further consideration of those many others has been hampered by focusing on rice and presuming it to be domesticated.

Another issue from Jiahu, is the report of new metrical data from the site. Of course new data may always force us to change our interpretations, but what is troubling here is more of an ethical issue. We considered Jiahu on the basis of a final published monograph of which one of Liu's co-authors, Zhang, was chief editor. In excavation reports one expects to see full publication of excavated samples and expert analysis, as is the standard procedure with ceramics and stone tools. That there is additional material, so different from what was published, that was not included and discussed, then pulled out as though a parlour trick, highlights a problem with the archaeobotanical data. It raises questions: what contexts and periods produced such divergent rice (and why); what other plants, such as wild foods, or possible weeds, might be associated? In other words it begs for archaeobotanical analysis. We may note more optimistically, however, that in the past few years there has been a growing attention to the systematic collection and study of archaeobotanical material in China (Reference ZhaoZhao 2005; 2006; Crawford et al. 2005; Lee et al. 2007).

Spikelet bases and studying domestication

An important recent development has been the study of archaeological spikelet base remains from some sites in southern China. Contrary to the statement of Liu et al. (2007) that spikelet bases are 'often used for distinguising domesticated from wild rice', this has almost never been the case, until recently. Reference ThompsonThompson (1996) first outlined the potential importance of this line of evidence in general, while Sato (2002) provided some observations on a small (and unsystematically collected) sample from Hemudu. The main problem is that spikelet bases have rarely been recognised archaeologically, although this is now changing as archaeobotanists tune their eyes (and sieve sizes) to finding them. Charred spikelet bases were reported from Mali (Fuller 2000) and were central to discussions of rice-crop-processing in Yayoi Japan (Reference HosoyaHosoya 2002), but this crucial plant part has been reported elsewhere rarely. Recent flotation samples from later Neolithic sites in Henan suggest that charred spikelet bases can be well-preserved in quantity where rice is prominent (cf. Fuller & Zhang 2007), while waterlogged sites in the Lower Yangzte region, from the Hemudu and Majiabang cultures, offer large assemblages. Recently, Zheng et al. (2007) have reported their first quantified results from a re-examination of Kuahuqiao, Majiabang period Lujiajiao spikelet bases, and a small sample from Hemudu period Tian Luo Shan. In all these assemblages it is clear that there is a mixture of wild and domesticated types, with nearly half or less being domesticated. This implies that these sites are on the trajectory towards the fixation of this key domestication trait, i.e. domestication as plant evolution is in process. This implies cultivation (human behaviour), but as domestication is not yet predominant nor fixed (in population genetic terms); this represents pre-domestication cultivation as per our model (Fuller et al. 2007: Figure 8; see also Fuller 2007). It must be noted, however, that methodological challenges remain, and distinguishing true domesticated spikelet bases from immature harvested wild types is a challenge yet to be overcome.

In studying domestication, we need to have multiple lines of evidence to consider different aspects of the process by which cultivated plants became adapted to human cultivators. Some important lessons can be drawn from other crops, and from archaeobotany in other regions. In the Near East, for example, it is now clear that fixation of tough rachis (domesticated) wheat and barley was an extended and slow process counted in millennia (Tanno & Willcox 2006; Fuller 2007). Although flotation began in the Near East in the 1960s, a sizable body of studies with quantified rachis data (equivalent to rice spikelet bases) has only become available decades later, causing some major new thinking about domestication processes and their slow pace (Tanno & Willcox 2006; Weiss et al. 2006; Willcox et al. 2007; Fuller 2007). It is also now clear that changes in grain size and shape did not evolve simultaneously and at the same rate as non-shattering domesticated ears. A further lesson from the Near East is that there were dead-end cultivars and early domesticates, such as the upper Euphrates rye (cf. Hillman et al. 2001; Willcox et al. 2007), two-grained einkorn (see Van Zeist 1999), an extinct striate emmeroid wheat (Reference Kohler-schneiderKohler-Schneider 2003) or Gilgal oats (Weiss et al. 2006). From this comparative perspective we might expect some lost early lines of rice, whether wild gathered, cultivated or even domesticated. Candidates for lost crops in China could be the rice from Jiahu and the recent finds further north at Yuezhang (Crawford et al. 2006), in a region where subsequent millennia (Yangshao and Dawenkou cultures) are dominated by millet agriculture to which rice appears to be added as a freshly adopted crop after 3000 BC (Late Yangshao and Longshan periods).

The increasing efforts by archaeologists in China to collect archaeobotanical evidence, and the growing number of specialists analysing these samples, represents the surest path to improving our understanding of early agriculture in this vast region. But it is also important that we reconsider existing evidence and debate our models. This exchange is testimony to that process.