Wednesday, 21 December 2011

Concatenating rice and language phylogenies: a recipe for single origins?

The literature has been and remains split on singular or multiple origins for rice. The discussion of whether or not all Asian rice can be traced to a single domestication event and a single cultural origin of cultivation continues, and there have been major arguments in favour of it this past year including the high profile genetic modelling paper by Molina et al published in PNAS in May, and a recent contribution on the historical linguistics front. I remain in favour of multiple origins [as per the "rice consilience" paper of last year] -- at least in terms of multiple starts of cultivation of wild rice even if not all of these starts lead on real domestication (in morpholoigcal adaptation terms) or to rice lineages still with us today. Increasingly I am coming to think that modern time-plane sampling, such as modern germplasm or modern minority language vocabularies, is biased by missing out on past diversity-- extinct landraces and wild progenitor populations, lost language diversity. Less diversity sampled will surely tend towards simpler historical reconstructions. Both disciplines also have a predilection for simplifying relationships as trees, which tend to favour single origins. Is there a sense in which genetics and historical linguistics together will tend to be biased towards identifying single points of origin? If so, then when there is evidence to contrary it should surely be given extra weight. A clearly there is an important role for archaeology to try to reclaim some of the lost diversity of the past.


Linguistic homeland for rice?


The new article on "the ethnolinguistic identity of the domesticators of Asian rice" has been published linguist George Van Driem, one of the foremost experts of Himilayan languages many highly obscure and dwindling, especially of the Tibeto-Burman language family has made numerous stabs at the contentious prize of correlating rice agricultural dispersal with historical linguistics, and its inferred history of population movements. In general it provides useful overview of the Austroastiatic and Hmong-Mien reconstruction relating rice and arguments about the homelands of these reconstructed proto-languages. Most archaeologists are probably more familiar with the hypotheses of Peter Bellwood and Charles Higham focused mainly on the Austronesian and Austroasiatic languages in Southeast Asia. (For two recent summary papers by Bellwood, from 2011, see Current Anthropology and the journal Rice). However, there has been considerable recent debate over how Sino-Tibetan (or Tibeto-Burman) languages fit in. The French Linguist Laurent Sagart, an the the instigator of a recent  Cornell conference on the rice and language, has argued in recent years for a shared ancestry of Sino-Tibetan and Austronesian, with an outward migration of rice and millet (Setaria) farmers from central-Eastern China. Van Driem tends to have more of an emphasis on Southwest China through Assam/Burma as a focal area, although he also postulates southwards migrations of  early Hmong-Mien speakers--  likely domesticators of rice (perhaps along the Middle Yangtze). Domestication of  rice by Austroasiatic speakers could have been separate although he seems to tend towards wanting a single origin. As for these Austroasiatic domesticators he places them  somewhere like Assam-- "the Northern Bay of Bengal littoral" (but I presume he does not intend of coastally adapted culture?). In this article Van Driem points to the recent Molina paper as simplifying matters of rice origins, allowing for early cultivation in India to be dismissed. In general the Molina paper and the earlier Londo et al (2006) map are used to suggest a northern Southeast Asia/NE Indian focus for rice domestication, while archaeology is cited mainly for failure to have worked in the required areas. While I would be the first to argue that we need archaeobotanical sampling across Southern China, Burma and Assam I think this paper put too much stock on modern time place samples (languages and genes) and lacks a full engagement with the material constraints that archaeology already provides-- good dates for a domestication process in the Yangtze, good dates for when cereal agriculture, fully formed, arrived in Thailand or Sichuan (e.g. recent blog). These make a dispersal into the Yangtze from a single origin elsewhere impossible, and a dispersal of early rice use into India unlikely (although later introduction of domestication genes is another matter). I do regard as plausible, even likely that there was a separate domestication pathway in the unsampled area, perhaps associated with Austroasiatic, but this must have been in addition to what was happening in the Yangtze and the Ganges. There is a likelihood to would related to the aus-rices-- a group not incorporated in the Molina et al model.


Maps based on modern/recent distributions are a recurrent weakness in the paper. Modern wild rices, like those mapped following Londo et al (or Sampled by Molina et al) are of course biased towards the more tropical South as wild rices are not extant in most of the Yangtze nor in eastern China, but this is a product of environmental change both climatic and anthropogenic-- the extirpated wild rices of the Lower Yangtze, Huai, southern Shandong are of course missing from modern sampling, but archaeobotany has the potential (and is) putting this on the map. Much of the paper  considers the linguistic paleontology for the homeland of  Austroasiatic, in the case of the latter providing a series of maps of fauna that reconstruct to Proto-Austroasiatic, which point towards a southeast Asian (or Southernmost China) distribution. While these provide an important starting point they are also biased towards modern geography rather than early/middle Holocene geography. To take two examples consider water buffalo and elephants, both mapped as going as far north as southern fringes of the Yangtze. But both are species that we know used to range into and north of the Yellow River valley at least through the Bronze Age (i.e. until perhaps 1500-1000BC). Hoffpauir's (2000) excellent chapter on the Water Buffalo in the Cambridge World History of Food provides a good map while a more schematic map occurs in my 2007 paper on non-human genetics, agricultural origins and linguistics in South Asia [pdf]. As for the elephant its former northern occurrence apparently even for Shang royal elephant hunts, provides the leitmotif and title for an excellent long-term environmental history of China,





























Mark Elvin's The Retreat of the Elephants. Compare Van Driem's map (top/left) and Elvin's retreat map (lower/right)


[note added 30 Dec. 2011: I have just come across the study by Li et al. (in press in Quaternary International) which has identified Early-Middle Holocene and Bronze Age "elephants" of northern China as the extinct Palaeoloxodon straight-tusked elephants rather than modern Asian elephants (Elephas maximus) but the point still stands since either could the be referent of an early etymon.


As for the genetic single origin (of Molina et al.)

Molina et al (2011a) carried out analysis of demographic history by attempting to model demographic parameters from several phylogenetic datasets using the new generation of collascent models (BEAST and DADI) which allow for multiple branches, bottlenecks and populations. This represents an important step away from single bottleneck models (e.g. Zhu et al 2007; Zhang et al 2009) which only attempted to estimate the correlation between early population size and length of the domestication (bottleneck period) of a single hypothetical origin. They conclude that single domestication of japonica is likely with indica derived from a subsequent bottleneck (underlying grey image below left), their discussion later hedges this with raising the importance of hybridzation/ introgression in indica, although the headline that most take away from this paper is still the idea of a single origin, single domestication event. However origin (start of cultivation) and domestication may not be the same thing.
 This study, however, does not sway me from my conclusion about the evidence for a proto-indica exploited in India before the introduction and hybridization with improved japonica, but reinforces the need for a fossil record. They are unable to sample and model the full range of diversity in all the lineages that have ever been cultivated, however briefly, over the past 8000 years. Their work focused on nuclear genetics, undoubtedly the most informative about evolutionary history in general, but they over overlooked the chloroplast. The chloroplast, which is maternally inherited, differs fundamentally between indica and japonica, nivara and rufipogon, such that there are shared characters between indica and nivara which differ with those of japonica and rufipogon, while other characters in indica are unknown in the wild. This set of chloroplast relationships is indicated in the color overlay in Figure. While they conclude that introgression, which I take to mean pollen flow, from wild rice (into the crop) in India might account for some of the genetic diversity in indica not seen in japonica, it cannot account for the chloroplast diversity, since chloroplasts are not carried in pollen. Thus pollen from wild Indian rices introduced to japonica is insufficient explanation. Instead one would have to posit pollen flow from crops into wild populations (with some nivara-related traits and some now extirpated traits) and that those wild populations retained domestication traits and were subsequently brought back into cultivation. Instead it seems still easier to posit proto-indica cultivation into which cultivated japonica , with a suite of valuable domestication traits, was brought into contact. The genetic background was largely indica into which japonica was the source of introgression, i.e. pollen flow.   I also worry the discernment between their alternative models is not very clear; comparing visually the differences in predicted outputs and actual data (Figure S4 of Molina et a 2011a) it is hard to see much of a dramatic difference between either predictions or fit. In the end I worry that they are forcing us to choose a false dichotomy between common ancestry and gene flow for explaining similarities between indica and japonica, and favouring shared ancestry on the basis an averaged phylogenetic tree across several datasets which may favour a single origin for domesticates much as is true of the neighbour-joining analysis of neutral genetic variation (Allaby et al 2008 or Allaby et al  2010 ).

Something that can certainly be remedied in the next round of models is the exclusion of temperate japonica, which was excluded from the sample set of Molina et al. As we see in the spread of rice in China, it enters the temperate zone of North China already by 3800 BC (based on a direct AMS at Nanjiaokou), where the short-grained is already evident. Based on environment and grain morphology then, we can infer the temperate japonica evolved (through its own post-domestication bottleneck) quite early. It is also from this region (the Yellow River) that diffusion westwards to central Asia and Northwest India is most likely. Therefore I would envision an early form of temperate japonica making the first hybrids with proto-indica. The sequence of bottlenecks associated with rice dispersal event was surely more than two, and some of these may have been associated with quite strong selection pressures (such as for or against photoperiodicity) as rice adapted to new ecologies: just two bottlenecks is unrealistic.  

Nevertheless Molina et al provide what probably a more realistic molecular clock estimate than has been possible before, with an initial domestication placed at 8200 BP (upto 13000), and the indica bottleneck at ca. 3900 BP. The former fits quite well with current archaeobotanical evidence for the beginnings of morphological change in Yangtze rice, and the later  date is spot on for the first appearance of the “Chinese horizon” in Pakistan and Northwest India. If this study is taken as representing the main stream of fully domesticated rice then it seems to be closing on elements of reality.

I would regard many of the problems identified above as conceptually relevant to historical linguistic hypotheses. Linguistic reconstructions inevitably work backwards from a modern time plane (although some may benefit from old texts too), towards a reconstructed common ancestor. This is like the domestication bottleneck in that while its form may be inferred from the modern data (given certain simplifying assumptions), lost side lineages, may be overlooked, and shared ancestry may therefore be easier to see than parallel developments. The dominance of the single japonica narrative found by Molina et al. (2011a) is much like the dominant Austronesian paradigm in its downplaying of substrates and language levelling processes. Indeed, recent perspectives on the later history of Southeast Asia, such as James Scott's (2009) The Art of Not Being Governed, have emphasized that ethnic affiliation has been flexible and that disgruntled overtaxed rice farmers have recurrently taken to the hills and joined the shifting-cultivator tribes, switching identity and language (at least for the past 1000 years). This is probably one factor contributing to the remarkable typological convergence across Southeast Asian language families and this may obscure early history. I find this conceptually similar to the observation that over time domesticated crop varieties come to resemble each other more than any resemble wild progenitor populations, leading to false ascertainment of monophyly (Allaby et al 2008; 2010).



. . .
My own working hypotheses on how rice phylogenetics and especially historical linguistics can be fit together within the framework of evidential constraints provided by archaeobotanical evidence, should be published soon in a special issue of the journal Rice [on-line here] arising from the Cornell conference....let the discussion continue and may it inspire new archaeobotanical sampling and genetic modelling. 

1 comment:

carro telecomandado said...

As per my own review, rice is really good for every one. I have not medical preference for rice.. But i have read tons of times about it.