Sunday 6 February 2022

The drug before the calorie? Some hazy thinking on Cannabis domestication

Many have been excited this week about headlines claiming marijuana (Cannabis) was domestication in China 12,000 years (making it the first crop in East Asia). As the reputable journal Nature put it "pot farming first blossomed" in China 12,000 years ago. But was it so? How clear or otherwise in the evidence? Is it really farming? The study by Ren et al in Science Advances paper is important-- it represents the largest collection of Cannabis genomes sequences, it provides some important information on subpopulations and genes that have been selected for differently in fibre hemp from drug strains. However, I finds is dicussion and conclusions riddles with both imprecision (about chronology, geography and cultures) and inaccuracies. So what should be questioned?.

A major source of inaccuracy lurks in sampling and geographical representation. This is compounded by the fact that in many countries growing or collection or transporting of Cannabis is illegal (although legalization is on the rise). Thus traditional drug varieties have not been sampled across most of Central Asian countries, Afghanistan, Russia, Iran. It should go without saying that sampling in the modern time plane will miss past diversity that has been lost to ha bitat destruction and environmental change; such a problem, for example, has plagued some genetic studies of rice-- as wild population no longer exist in the regions were it was first cultivated (such that a modern genome map is not a map of origins). But even so the limited sampling across free-growing (feral/ wild) outside of South Asia and China) is notable. Compare the map of Ren et al (above) with that from the Cannabis book by Clarke and Merlin (2013), which highlight how much more wild/feral diversity there is out there across Asia and eastern most Europe, including 'wild' populations south of ot eh Caspian sea in Iran and along the Volga River in Russia. Even the map from Clarke and Merlin is incomplete with regards to Afghanistan where Vavilov collected apparently wild drug types of Cannabis (C. indiva var. afghanica) in the 1920s.

Given how few wild/feral samples that they have can they really rule out multiple domestications. Genetic analyses often err on the side of single origins. Simulation work (Allaby et al 2008 PNAS) have shown that this will be true even for crops with multiple origins, because gene flow among crops of different origins and pruning of lost branches (not sampled or not surviving to present). Their analytical methods are more modern and more sophisticated but I am not sure they can rule out multiple origins, and they certainly can’t rule out origins from regions not sampled or where wild populations are extirpated (e.g. Japan). Also wild populations (lets assume there are some in central Asia) can be heavily inundated with gene flow from crops over time make their original dinstictiveness hard to find in modern genomes.

A major issue is imprecision in dating. Genomic dervied estimates will never rival radiocarbon dating on archaeobotanical remains. Their proposed date of origins (is this domeatication?) has 6000 year error margins. And it not clear what they date! Was a divergence  between two wild population separated due to climatic vicariance at the end of the Pleistiocene or start of the Holocene, or is it meant to be the domestication bottleneck? The equivalent genomic dates for Asian rice domestication are ca. 18,000 BP and something close to 10,000 for African rice. Both of which are way off. African rice is domestication at more like 3000 BP and Asian rice at more like 8000 BP (of course it matters whether one is talking about the beginning or end of a process, as domestication takes 3000-4000 years in terms of morphological evolution/ genetic fixation.

By contrast archaeological dates are much more precise, at worst with 100-200 year error margins, but Ren at all quote these very imprecisely. It is as if they wish archaeology was less precise, but even then it would not approach the dating imprecission attached the the genomic dates. The say "~3000 BP" for the appearance of Cannabis in India, but if we are rounding off it is closer to 4000, as it occurs in the Late Harappan horizen (3900-3500 BC). Although Indian epics are not well dated some parts of them are from oral traditions that probably also date around then and make reference to Cannabis- I think Ren et al refer to this as ~2000 BP. Cannabis comes to India in my view as part of “Chinese horizon”, which is really just piece meal adoption of various things coming in via central Asia including crops and technologoies (harvest knives) from China, peaches, apricots, millets, japoonica rice. For discussions of this see, for example Fuller & Boivin (2009); Stevens et al (2016). Whether or not it arrived in South Asia earlier, or was already utilized from wild populations in the Himalayas, does not really make much differece to the whether or not cultivation began at the start of the Holocene in China. The imprecision in genetic dating, however, makes it about as likely that domestication took place around the start of the Yangshao period (~5000 BC), the period when we generally see the ending of the millennia long domestication process for the native China millets (

They are also quite imprecise about geography: are they suggesting a NW China (Xinjiang) or NE (Chifeng) source? And if either of these then discussion of early cord-marked pottery in South China (mostly south the Yangtze) is really not relevant, and yet they discuss this as though it indicates the use of hemp cords-- for which there is simply no evidence. 

Cannabis is undoubtedly an important crop brought into cultivation early in east Asia, esepcially for its medicinal and/or social uses, but becoming important for larger oily seeds and fibres over time. But in my view its development as a crop either parallels or is even inspired by the increasing importance of cultivation of other taxa, like the China millets. And this process could easily have played out multiple times-- perhaps in very different contexts in parts of central Asia or Jomon Japan, even amongst non-farming cultures. While genomic data will contribute to this, modelling such data really requires some calibration points in time and space, which will ultimately come from archaeobotany.