Diffusion versus migration in North African prehistory

There is a little disagreement in the letters of this week's Science, about mtDNA evidence for migrations from West Asia into North Africa. This is in reference to a paper late last year by Olivieri and colleagues, that argued that both North African and European populations traced their ancestry ultimately to Upper Paleolithic people of the Levant.

That paper had this abstract:

Sequencing of 81 entire human mitochondrial DNAs (mtDNAs) belonging to haplogroups M1 and U6 reveals that these predominantly North African clades arose in southwestern Asia and moved together to Africa about 40,000 to 45,000 years ago. Their arrival temporally overlaps with the event(s) that led to the peopling of Europe by modern humans and was most likely the result of the same change in climate conditions that allowed humans to enter the Levant, opening the way to the colonization of both Europe and North Africa. Thus, the early Upper Palaeolithic population(s) carrying M1 and U6 did not return to Africa along the southern coastal route of the "out of Africa" exit, but from the Mediterranean area; and the North African Dabban and European Aurignacian industries derived from a common Levantine source.

That sets out the hypothesis: a migration from the Levant some 40,000 years ago spread these haplotypes into North Africa, at around the same time as in Europe.

I took some notes on this paper at the time, because of the real paucity of any comparative information on the "Dabban" industry that is the proposed archaeological correlate of this migration. I'm not going to go into it here; let's say I was skeptical at the time, not least because geneticists have a way of assuming that industries are much more "real" or extensive than the archaeology allows. I still think that the archaeology is weak, but it is certainly possible that some significant population movement happened.

The current letter and response are interesting, not because they differ in their interpretation of the haplotype distributions (which they do) but because their arguments are almost entirely in terms of archaeological and linguistic comparisons.

Forster and Romano argue that haplotypes can't really provide evidence for early population movement, because a relatively late migration could have carried much older haplotypes along with it, or they may have entered North Africa by diffusion without any major population movement.

They argue that archaeological and linguistic evidence favor more recent migration of populations as a major mechanism for the movement of gene lineages:

Three points lead us to believe that our younger chronology for the back-migration into northern Africa still merits consideration. First, the mtDNA trees reconstructed by Olivieri and colleagues are less than conclusive because they consist of phylogeographically mixed branches, which cause uncertainty in identifying the relevant founder nodes for genetic dating. Second, in our view the fact that the North African mtDNA marker types still correspond so closely with the Afro-Asiatic language zone argues against the existence of that correlation for tens of thousands of years. Third, cave art in the Sahara shows that in Neolithic times (around 5000 B.C.), the population of the Sahara was still of sub-Saharan African ancestry (see figure), whereas "Europoid" figures documenting the arrival of west Eurasians appear later in the cave art record (3).

In response, Olivieri and colleagues claim that the recent Holocene events -- although relatively well documented, are insufficient to explain older haplotype distributions, and that archaeological evidence also supports their point of view.

The principal problem with great syntheses of languages, genes, and figurines (or pots) is that they lump together different migrational and cultural processes and especially overstretch recent events of the Holocene, thereby downplaying or swamping the genetic signals that point to much earlier events of the Pleistocene (1, 2).

Personally, I don't know which hypothesis is correct; it seems to me that mtDNA haplotypes are never going to answer this kind of question. The question is about mechanisms of genetic dispersal.

Both hypotheses more or less agree about the current distribution of haplotypes. I say "more or less" because in fact, both hypotheses are interpreting these distributions post hoc -- they're not really testing hypotheses, they're just offering archaeological arguments in support of migrations they assume the mtDNA is documenting. Remember when I mentioned the "reality" of these archaeological industries? This is what I meant.

Regardless of what we think about the archaeology, these haplotype distributions still deserve some explanation. How did a 45,000-year-old haplotype spread from its apparent origin in the Levant across North Africa? Did it get there slowly and gradually by diffusion? Or did it come all at once in a long-distance movement of a group of people -- a so-called "folk migration"?

This question, diffusion versus folk migration, is of course a very old one. It remains central to all these considerations of recent genetic variation.

A couple of weeks ago, I had the extraordinary privilege of hearing two of the real leaders in paleoanthropology having precisely this argument. How much of recent evolution has been driven by folk migration, and how much by the diffusion of genes into standing populations?

These letters are a good illustration of the question, drawn out into a particular case. We'll be hearing more about this soon, I think.


Forster P, Romano V. 2007. Timing of a back-migration into Africa. Science 316:50-53. doi:10.1126/science.316.5821.50

Olivieri A, and 14 others. 2007. Timing of a back-migration into Africa. Science 316:50-53. doi:10.1126/science.316.5821.50

Olivieri A, and 14 others. 2006. The mtDNA legacy of the Levantine early Upper Palaeolithic in Africa. Science 314:1767-1770. doi:10.1126/science.1135566