Indian Archaeobotany watch: Lahuradewa 2008

Critical comments on the archaeobotany of Lahuradewa. The Pragdhara volume 18 (for 2008) arrived in London in the first week of June. It constitutes a special issue on the Neolithic and agricultural origins, with various reviews on other parts of the world (some I was involved in writing), and for various parts of South Asia. Perhaps the centrepiece of the issue is the latest report on Lahuradewa, excavated by the Uttar Pradesh State Department of Archaeology, directed by Rakesh Tewari. This one article from the latest Pragdhara has been made available on-line here. I have been a disbeliever in print in the past on the rice from this site (see JWP 2006, and others). Not about whether it is an important site, nor the fact that it has the earliest finds of rice from South Asia (>6000 BC), nor the earliest pottery in South Asia-- demonstrably earlier than Mehrgarh Period 2, which starts from ca. 6000 BC. But I do not think, and I now doubt even more, that the rice was domesticated. It is not even clear that it was cultivated, and is plausibly (perhaps safest interepreted as) wild gathered.

They report three new radiocarbon dates on bulk charcoal samples, which calibrate to between 8000 and 9000 BC. This means that the 50cm or so of cultural stratigraphy now has to account for 5000 years, or more, of human occupation. One has to conclude that this occupation was unlikely to have been permanent and sedentary. Importantly, they also recovered more plant remains, including more rice from the lower levels (Period 1A). Details of numbers, densities and samples from flotation are not reported. New finds also include a large ceramic fragment tempered with rice husk, and apparently some rice grains, as well as carbonized grains and spikelets. They suggest that these are domesticated on the basis of three criteria, grain size and grain ratios (using what might be termed the ‘Vishnu-Mittre index’), husk patterns, and the alleged presence of non-shattering rachises (i.e. spikelet bases).

Spikelet bases. Lets start with the last observation first.

Clear criteria for distinguishing three categories of spikelet bases, one of which is definitely of domesticated type, have been recently published (Fuller et al. Science 2009; Fuller & Qin 2008), although these publications probably post-date when this report went to press. Nevertheless, earlier work by Gill Thompson (1996; 1997) provided clear illustration of the differences between typical wild and typical domesticated spikelet bases. There are four spikelet bases shown in their Figure 16, one which is shown in close-up (Fig 16.3: above) as an example of the non-shattering type. Its long rachilla is still attached, which is a trait occassionally (but rarely!) encountered in domesticated rice, and when it does occur it usual in East Asia rices that possess multiple non-shattering alleles and it seems most common in modern varieties adapted to machine harvesting. Rather the attached rachilla is typical of rice harvested immature and green. As noted in the Lower Yangzte and China generally, spikelet bases with protruding rachillae are common in the earlier Neolithic (e.g. at Kuahuqiao and Tianluoshan) but these forms decrease over time (see Fuller et al Science 2009 [follow the link from here]), until by ca. 2000 BC in Chinese sites they are very rare (<10%).

It should be noted that both of these represent spikelets that do not appear to have broken during dehusking, and that appear thin and deformed, and are likely immature (green spikelets), which did not contain fully-formed grains. These therefore look more like green-harvested, wild rice spikelets than the threshed remains of a domestic rice harvest! But these are illustrated as the best candidates of Lahuadewa "domesticates". What is more they both have preserved awn bases. While the loss of awns is not a definitive trait of domesticated rice (many varieties, especially of tropical japonica) are awned, the presence of awns is typical of wild rices. The pictures therefore do not agree with what is stated in the text, but quite the opposite.

What about husk patterns? The basis of using husk patterns to distinguish definitively between O. nivara, O. rufipogon and O. sativa has never been clearly demonstrated or published. Quite the contrary this seems to be a non-replicable, subjective judgement. The idea is that domesticated rice is nicely ordered with square cells, and wild rice is wild and disordered. There is perhaps more of the magic of metaphors than a real method here I suspect—in any case I have never been able to see this, and one can find exceptions to this in evefy box of wild or domesticated reference material. The original inspiration of this came from the work of T. T. Chang (and was then developed by Vishnu-Mittre and his students in Lucknow), and although Chang often assigned archaeological material on the basis of husk patterns, this relied on a kind of personal magic and authority and had always had to be taken on faith. Chang borrowed this method (which was termed the SUMP method-- Susuki’s Universal Micro-Printing) from the earlier work of Katayama (1969). In the original study it is determined that there are no significant differences among the species of the sativa-complex (including rufipogon) but only between these and other wild rices (non-AA genomes rices). [Katayama, T, 1969. Botanical studies in the genus Oryza I. Morphological and anatomical investigation of glume- and leaf surface with the SUMP and histological method. Memoirs of the Faculty of Agriculture Kagoshima University 7/1, 89-117.]

I suspect that there may be some tendencies of difference between wild and domesticated spikelets husks on a popualtional level, akin to the weak tendencies in husk phytolith form, all of which are probably linked to selection for larger, fatter grains. The husk patterns therefore should show trends of gradual change overtime as grains do, but until methods of measuring and quantifying this over time are developed, this is a non-method, and seems a leap of faith too far.

Grains. This report provides a table of grain measurements, on 26 grains (although judging by the photos I wonder if some of this included attached husk, which would elevate some measures and create greater variance). It should be noted that these are all Period 1A grains with no comparison provided to later periods. Thus there is no possibility of looking for the temporal trends that one expects with domestication. In any case it is clear from examining these measurements that they break into two size groups, one is small and the other larger. This is easily illustrated in the following chart.

The smaller-grained group is comparable to non-sativa small-grained rices (e.g. O. officinalis), while the other falls into a size range that could be domesticated rice. However, when length and width measurements are taken as a scatter plot, all of these grains fall within the range defined by modern O. rufipogon and (especially) O. nivara. None of them fall into the range of domesticated rice. In order words none of them is bigger than a baseline that might be defined on the basis of modern measurements. Both the large and small groups contain ‘Vishnu-Mittre indices’ that are >2 and ~1.7, which are alleged to distinguish domesticated and wild rices. Internally this data deconstructs the usefulness of this index as a marker of domestication. Modern measurements on populations of wild domesticated rice grains certainly do not bear these indices out!

The two populations are illustrated also by scatter plot, below, where the Lahuradeva specimens (light green) are plotted over the scatter of modern populations that were plotted in Fuller et al (2007, Antiqiuty; measurements by Emma Harvey). To compare the modern and ancient grains I have added a +10% increase to the archaeological specimens as a reasonable standard correction for charring. It can be seen that the Lahuradewa grains plots nicely with Oryza nivara, while the shorter grains plot with O. granulata and O. officinalis.

Because comparison with modern rice grains may be complicated by the charring factor, I have taken two archaeological populations from India for comparison as well. Both are Early Historic (perhaps ca. 200 BC, or so) and both come from regions beyond the range of wild rice. One is from a lense of charred grains at Terr, Maharashtra, and the other is from Balathal, Rajasthan. Both were measured in London, 300 grains each. As can be seen there is considerable variation (part of that is probably due to the fact that the Balathal grains were charred as grains, whereas the Terr grains were charred as spikelets and many retain husk fragments), but these later Indian archaeological domesticated rices are clearly shifted away from the range of reference nivara and from Lahuradewa. Thus, using later archaeological rices as a baseline the larger Lahuradewa grains do not look domesticated. There is nothing beyond reasonable doubt to accuse these grains of being domesticated.

Interestingly, if these grains are compared to those from the later Neolithic in the Ganges, e.g. measured grains from Mahagara and Koldihwa, they fit nicely together. Based on recent spikelet base finds from Mahagara (1800-1600 BC), we know this rice to be domesticated. This then implies that there has been no appreciable change in grain size in rice the middle Ganges between ca. 6000 BC and ca. 1800 BC, despste being domesticated by the end of this period. This implies a very different domestication process than that in Chinese rice, in which grains get progressively larger over time as spikelet bases become non-shattering, or from wheat and barley in the Near East (see my Annals of Botany 2007 paper). Indeed, it suggests that there was hardly any selection for increasing grain size in the proto-indica cultivated in the Ganges plains (whenever cultivation began, which remains unclear). This makes sense in terms of hypothesized patterns of wild rice exploitation and minimalist wild plant food production (pre-domestication) cultivation that can be postulated for the Ganges (Fuller & Qin 2009).

This evidence is probably to be expected, given that genetic evidence indicates that several key mutations had to be introduced to proto-indica via hybridization from domesticated japonica, including sh4, for non-shattering, prog1 for erect growth habit, as well as rc for white pericarp. The real leap forward for indica rice was perhaps closer to 2000-1800 BC. Nevertheless the roots of rice cultivation were laid down earlier, but it remains unclear if this was as early of the eariest dates at Lahuradewa or whether these were periodic seasonal rice gatherers.

Diatoms. It is also suggested that the diatom assemblage from the lake sediments indicates rice growing fields. Are they suggesting, implausibly, paddy fields at this date? There is simply too little background work on the ecology of diatoms in natural wetlands where Oryza nivara, O. officinalis, etc, grow to be able to justify this statement. The diatom species that now inhabit rice fields existed before there were rice fields, and they had to come from somewhere. The habitat of wild rices seems the obvious place.

(Appendix) Some general notes on the plant assemblage. Plant taxa reported from Period 1A are: rice (reported as wild and domesticated, but see below), wild Setaria (referred to yellow foxtail millet, S. pumila), Chenopodium (referred to C. album), Coix lachryma-jobi, Artemisia sp., Silene conoidea. The Silene appear to have intact light-coloured hila (Fig. 6.8), which makes one a little concerned that they may be uncharred and intrusive, but maybe not. The rice grains as illustrated are for the most part plump and appear mature, but they are relatively short (more a feature of O. nivara than typical modrn indica), except for a few elongate, thinner grains (Fig. 6.5), at least one of which is poorly formed, which are referred to O. rufipogon; indeed they are quite plausibly rufipogon, but may also include immature grains.

Period 1B (probably 2500-2000 BC, although one wood charcoal date goes back to ca. 2800/2900 BC): apart from rice, finds include free-threshing wheat, barley, lentil, Cyperus, Coix lachryma-jobi, Artemisia, Setaria cf. pumila [Saraswat persists in the use of S. glauca, a taxonomically illegitimate name—Linnaeus’ type specimen was pearl millet not yellow foxtail!-- but lets not squabble], kodo millet (Paspalum scrobiculatum)—these are in the husk and look more likely to be wild/weedy specimens rather then the crop. The rice includes many grains referred to Oryza sativa (reasonable), some O. rufipogon (which again look like they may include immature grains: Fig. 8.8), and some O. officinalis (very short and wide), with length of ~3mm or less (Fig. 8.9). It’s a pity that these and the sativa type grains were not measured for comparison to the Period 1A material. Impressively there is some husk material of O. officinalis. This adds another site to evidence for the exploitation (or at least harvesting) of more than one rice species in the Ganges plain. Emma Harvey (in her PhD, UCL, 2006), at Mahagara and Koldihwa, Neolithic sites of the Second Millennium BC, measured husk phytoliths (the double-peaked morphotypes) and found a small proportion of husk phytoliths that fell well outside of the range of the sativa-rufipogon complex, some of which overlapped reference measurements for O. officinalis. This suggests that we should expect multiple rice species to occur on many early sites in this region.