agriculture

Earlier this spring, I wrote about a paper by Brenna Henn and colleagues that presented new data on SNP variation in recent African hunter-gatherer populations [1] ("Population structure within Africa: has 'modern human origins' become a non sequitur?").

Another paper that came out this spring from the same research group is also very interesting. Christopher Gignoux, Henn and Joanna Mountain [2] examined the evidence for Holocene population growth in Europe, Africa and Southeast Asia, from within-haplogroup variability of mtDNA haplogroups. The idea is that earlier samples were not finely resolved enough to examine events of the last few thousand years, either because they included only small sequences (e.g., control region) with limited variation, or because they included whole mtDNA genomes with too few individuals to look at within-haplogroup coalescents. So here they add more individuals. It is still a small number (425 total) and so I expect that we will see better ones in the next few years.

The results are nonetheless useful because they provide some nice matches for the archaeology of early agriculture. For example, in Africa:

We find two periods of population expansion within our sample of lineages originating during the Holocene in western Africa. Although the majority of coalescent events occur during the Holocene, a number of lineages from this sample also coalesce during the Upper Paleolithic. The earliest growth begins at ≈38,000 ya (CI: 33,500–45,000 ya) (Table 1 and Fig. S1) and the second period begins at ≈4,600 ya (CI: 3,000–10,000 ya) (Table 1 and Fig. 1B). The correspondence between the timing of genetic evidence for a sharp increase in population size at 4,600 ya in our Holocene sample of sub-Saharan Africans and the archaeological evidence for origins of agriculture in western Africa is quite close (Fig. 1B and Table 1). In contrast, our southern African Upper Paleolithic sample representative of hunter-gatherers shows no growth over the past 20,000 y. We suggest Bantu-speaking farmers and other pastoralist groups migrated throughout southern Africa 2,000 ya (27) without impacting southern African mtDNA lineages (Fig. 1B).

We can't really understand the pattern of genetic variation within Africa without understanding when the population grew. In Africa, Middle Stone Age genetic variation must have been more extensive than that in other regions of the world. But the survival of that MSA variation to the present day depends on the demography of populations over the past 50,000 years. In a growing population, fewer lineages will be lost by random genetic drift. So if Gignoux, Henn and Mountain are right about the growth of West African populations by 35,000 years ago, we might expect that region to preserve some extensive variation from MSA times. That might explain why that population preserves very deep Y chromosome lineages [3]. Regarding only mtDNA, one might conclude that a historical paucity of migration between hunter-gatherer and agricultural groups would be the most important reason why MSA variation remains in the present-day African population. This has been the explanation for survival of deep mtDNA lineages in southern Africa, for example. The Y chromosome result and the current paper remind us that population growth can also preserve variation from earlier time periods.

I think this proposal of African population history matches very well the model that we assumed in our acceleration paper [4], which we based on the archaeological record. We suggested early population growth in Africa by 35,000 years ago followed by an agricultural expansion after 5000 years ago. The evidence for relatively late agricultural intensification, within the last 4000-5000 years in sub-Saharan Africa, is very clear archaeologically. Less clear: How big was the earlier, pre-agricultural human population? The LSA might correspond to a demographic intensification, generally after 45,000 years ago. Genetics has certainly seemed to support such a view, and we found it consistent with the evidence that positive selection had increased in rate much earlier in Africa than in other regions. Still, the more detailed study by Gignoux and colleagues helps to clarify this picture.

The results also show agricultural population growth to have been late in Southeast Asia.

Direct archaeological evidence for rice agriculture in southeastern Asia dates to only ≈4,400 ya in Thailand (28). Agriculture spread throughout Island Southeast Asia, with evidence of rice in Taiwan again dating to ≈4,400 ya. Our Southeastern Asian Holocene population size curve indicates expansion beginning ≈4,700 ya (CI: 3,000–5,700 ya) (Fig. 1C and Table 1).

Again, useful. I think we need to exert some effort making sure that the initial dispersal of people into South/Southeast Asia can be differentiated from the post-agricultural history. But assuming that Gignoux and colleagues are correct, it makes sense in an overall picture of slowly adapting early crops to tropical climate regimes, or replacing early domesticates with different ones in those areas.

I am less sanguine about their results for Europe. They show a gradual period of growth associated in time with the Younger Dryas (around 12,000 years ago), which could make sense in the archaeology. But I am not convinced that the "European" haplogroups here are really European to that time depth. We know that the Neolithic and post-Neolithic saw some large-scale shifts in the frequencies of mtDNA haplogroups in Central and Western Europe. Some Upper Paleolithic Europeans probably contributed mtDNA to this later population, but I have no confidence that the proportion was great enough to accurately infer the demography of that pre-Neolithic population. (This is also a problem with the current paper in Current Anthropology by Peter Rowley-Conwy. I'll discuss this sometime soon.)

The next frontier in reconstructing the population history of Europe will be ancient DNA. A good sample of Neolithic and pre-Neolithic whole mtDNA genomes would settle this question and allow inferences about the kind of demographic recovery Europe underwent after the Last Glacial Maximum.

An open question is to what extent the other populations have similar problems. The European population of today reflects West Asian population dynamics 10,000 years ago. The East African population today reflects West African population dynamics from before the Bantu expansion, possibly to a similar extent. The population of Southeast Asia reflects the population dynamics of early rice agriculturalists in South China. And so on.

Adding large-scale migration and partial population replacement to this kind of demographic analysis is not easy, but it will be essential if we want a better picture of how agriculture affected human populations. Considering these problems, I think it's easy to see why I started working on Holocene population dynamics. Evidence about Late Pleistocene populations, like MSA Africans and Neandertals, still lies within our genomes. But we see it through a lens. Holocene population dynamics -- movements and population growth -- distort that lens. If we don't account for those Holocene dynamics, we will conclude wrongly about the earlier dynamics.

I like this a lot, because this is what anthropology is really good for. We can bring a lot of archaeological and historical knowledge to bear on the question of post-agricultural population dynamics. But it's a deep, deep field with a lot of specialized literature.

References

The dawn of ironworking in Africa is a hot anthropological topic. My own interests in demographic growth and dispersals depends very closely on the chronology of ironworking in Africa, because the advent of iron may have enabled faster conversion of land to agriculture.

Many anthropologists believe that the dispersal of the Bantu languages may be traced to an agricultural explosion driven by iron technology. Others dispute this connection, raising doubts about whether the ironworking chronology can match the timing this dispersal. Both these have some wiggle-room in their dating, as do the times of introduction or domestication of various crop species.

For the purposes of our paper last year, it was sufficient to know that populations grew in Africa after roughly 2000 BC. But to test hypotheses about gene dispersal and selection among African populations -- data that are now available -- we have to be a bit more precise.

Last week's Science includes a summary article by Heather Pringle, which discusses the controversy over the chronology of African ironworking.

Now controversial findings from a French team working at the site of Ôboui in the Central African Republic challenge the diffusion model. Artifacts there suggest that sub-Saharan Africans were making iron by at least 2000 B.C.E. and possibly much earlier--well before Middle Easterners, says team member Philippe Fluzin, an archaeometallurgist at the University of Technology of Belfort-Montbéliard in Belfort, France. The team unearthed a blacksmith's forge and copious iron artifacts, including pieces of iron bloom and two needles, as they describe in a recent monograph, Les Ateliers d'Ôboui, published in Paris. "Effectively, the oldest known sites for iron metallurgy are in Africa," Fluzin says.

Some researchers are impressed, particularly by a cluster of consistent radiocarbon dates.

And, as you might expect:

Others, however, raise serious questions about the new claims.

The article casts the debate as an opposition between a diffusionist hypothesis (metallurgy entered Africa from the Near East) and local development. That's appropriate, since this pattern of opposition is one of the oldest stories in archaeology. But I'm more interested in the dates and resulting population dynamics. How did technology relate to demographic growth, and how were genes affected by these processes?

The article makes the early development of ironworking in Africa seem very credible, particularly if the only other option is a late introduction via Carthage or the Nile corridor. It is not obvious how much of the apparent controversy is about the early dates from this one site in particular, and how much is about the presence of pre-first-millennium BCE ironworking generally. Critics raise various scenarios for the contamination of radiocarbon dates by old carbon. This always reminds me about how much error may lie within Paleolithic dates if we have to worry about contamination in Iron Age sites!

Well, more on this issue later.

References:

Pringle H. 2009. Seeking Africa's first Iron Men. Science 323:200-202. doi:10.1126/science.323.5911.200