Web Gems: August 3, 2012

This is a new post genre on this weblog. “Web Gems” will feature smart thinking from the Comments sections of various other blogs followed by my no less bright commentary.

This week I have three good ones:

Andrew Oh-Willeke surprised today by writing the following:

“The C1-C3 date is still surprisingly low relative to O however. I would have expected a date long before the dates for O, not almost the same, given that there is C4 in Australia and C2 in Melanesia, for which a natural differentiation time would have been ca. 45,000 years ago, given that Denisovian admixture suggests that men with Y-DNA C would have been among those admixing with Denisovians and hence presumably in the first wave of modern human migration into the area, given the absence of Y-DNA O in Australia and New Zealand despite overlap on the mainland with C.”

He refers to the following dates obtained by Dienekes for the Y-DNA clades – C1-C3 (C1-P122 vs. C3-Z1453) 25,022; CF (C-M130 vs. F-P137) 47,379; CT (DE-M145 vs. CF-P143) 62,439 next to NO (N-M231 vs. O-P191) 32,467; O1-O2 (O1a-M119 vs. O2-M268); 26,145 O1-O3 (O1a-M119 vs. O3-P198); 26,303 O2-O3 (O2-M268 vs. O3-P198); 27,870 O3a1-O3a2 (O3a1-L465 vs. O3a2-P201) 18,765

Population isolation and high inbreeding rates must have an effect on the clades attested in the Americas. Clade C is a very upstream clade; correspondingly it has a wide geographic distribution and is well attested in such “relic” area as the Sahul. If hg C3 was indeed that young, we would have found hg O in the Americas, as hg O is attested all over Siberia and South East Asia. Hg O is not found in Australia either, and Australia has C4, a sister clade to C3. This indicates that the real dates for hg C3 must be on the order of 40-50,000 YBP.

“An aspiring linguist” Teo. observed the surprisingly low levels of linguistic diversity in Africa, which is inconsistent with the claim that Africa is the continent on which (behaviorally) modern humans originated:

“Not only that, sub-Saharan Africa as a whole is extremely undiverse and there doesn’t seem to be any good reason for that. Just compare it with New Guinea, which supposedly was populated long after Africa: NG was certainly not that isolated, either, having had contacts with groups from east Asia and Australia. And yet, NG is immensely more diverse than Africa. In a small stretch of land running the northern half of the island alone you can find more language families and isolates than in the whole of Africa (even accounting for further splits there). Has SS Africa been such a hotbed of cross-contact in comparison with NG – or much of Eurasia, for that matter? I don’t think so. So, again, either there was a “mass language extinction” event on a scale never seen again, or there is something very strange here. The linguistic assumptions may be wrong, the genetic assumptions may be wrong, or both….”

And later, in response to Onur’s argument that Africa is not alone in losing most of its linguistic diversity beginning from the Neolithic revolution: Eurasia too lost most of its linguistic diversity,”

“Onur, again, that is ok. But there are two problems: a) Eurasia had multiple neolithic “bursts” and the resulting crops and herds are easily adaptable from one end of the continent to the other under most climatic conditions, so a general “wipe-out” and a series of sweeping expansions (IE, Turkic, Sinnitc etc.) are not unexpected. I’m not sure the same is true of SS Africa. Even if there were pre-Bantu expansions, they don’t seem to have been as numerous or as successful as the multiple Eurasian ones.
But, even if that “wipe-out” scenario is possible, there is another problem:
b) Again, if language as we know it was an early phenomenon, the diversity in Africa was supposed to be much much higher than elsewhere pre-Neolithic, so that even a wipe-out should leave at least some hint of that former diversity intact in isolated pockets. We see very little of that. Why did the “linguistic culling” have to be so complete?
The Eurasian situation is supposed to be an already smaller diversity reduced to very little after countless Neol. Expansions for a long time, whereas the African one is supposed to be huge diversity reduced to very little after a handful of expansions (certainly fewer than Eurasia, and beginning later). Why? That is what is unique and requires explanation.”

It’s true that there were more language extinctions going on outside of Africa than in Africa in the past 40,000 years, including LGM, Neolithic replacements, not to mention the most recent, post-1492 language loss in the New World. PNG, the New World, Caucasus are the refugia that have retained ancient levels of linguistic diversity and population fragmentation. There are no comparable refugia in Africa. There’s a very simple explanation for this: Africa is not as ancient, as a behaviorally modern human occupied region, as Papua New Guinea, the New World or parts of Asia. As we can see from the distribution of the Y-DNA DE clade, it’s not found in America or Papua New Guinea suggesting that they had been colonized before the expansion of this clade in Africa and parts of Asia. Also, it’s precisely in the New World and PNG that we see the highest frequencies of Denisovan alleles. May be a coincidence or an artifact, but could be a systemic property related to the linguistic diversity phenomenon. Broadly speaking, East Africa has the largest linguistic diversity in Africa, with all major language families represented there via their most divergent branches (Kordofanian, Hadza, Sandawe, etc.). East Africa must have been then the route by which Africa was colonized by behaviorally modern humans. The question of course that looms large, under this scenario, is what to do with genetic diversity in Africa. The simplest solution is to consider all the particularly divergent branches in Africa as products of admixture with “anatomically modern humans” who, as Richard Klein would confirm, didn’t have languages and other aspects of modern human behavior. There may be other explanations (cycles of genetic divergence followed by admixture in Africa, increased mutation rate in Africa, selection, relaxation of inbreeding depression, etc.) but archaic admixture may be the easiest to visualize. In all the refugia areas outside of Africa (New World, PNG, Caucasus), we observe reduced levels of genetic diversity. In Africa, Hadza also show reduced genetic diversity. All of this may suggest that African diversity is a product of both a) archaic admixture and b) intraspecific admixture among various human populations. Hadza is an isolate that maintained the original high inbreeding rates, low heterozygosity, etc., while all other African populations absorbed a lot of gene flow.

Kosmo asked Razib a question:

“I’m curious about your thinking in regard to the lack of archaic MtDNA in modern human populations. With evidence for the diversity of ancient admixture mounting all the time, (Neanderthals, Denisovans, and now various African archaics) it seems that chance loss of archaic MtDNA haplotypes is less and less likely, and selective pressure must be invoked to explain their absence in living people. (Perhaps some selective advantage conferred upon haplogroup L and conserved in all daughter haplogroups thereafter– leaving all pre-L haplogroups relegated to the dust bin of history) If selection is a major force at work in MtDNA, does that not destroy the reliability of the clock?”

Razib gives a sensible answer (“i wonder as methods for autosomes get better and better we’ll stop using uniparentals for deep time studies. the other option could be that HVR is too mutable, and there’s convergence. so most lineages are extinct, period, and the coalescent just happens to not have the archaic branches”), but he doesn’t realize the implications. Kosmo continues:

“in regard to the missing MtDNA, I agree that Occam’s razor used to side with the idea that “the coalescent just happens to not have the archaic branches,” rather than invoking some special selection. But with the new studies that have come out recently, I wonder if thinkers in the field have had time to notice that the razor now seems to cut the other way. I claim no special insight, but with each new discovery of admixture, the odds go down that the archaic haplotypes were lost due to chance, again and again, while only the L-derived haplotypes were retained. Given selective neutrality, you could argue that somewhere close to 8% of Island New Guinea should have pre-L MtDNA. 4% of Eurasia. And now these new studies about African archaic admixture reveal yet further instances where L-derived haplotypes rose to 100% frequency at the cost of all other variants which admixture analysis, and an assumption of neutrality, tell us should likely be there. Chance could certainly account for the loss of certain mitochondrial lines within populations, but we must remember that each of these is a distinct case, so we must also consider the liklihood that they would happen lose exactly the same lines. I think Occam’s razor sides with selection now.”

And then,

“The first step should be to prove that the math has switched sides on the null hypothesis–and an alternate explanation to chance-loss is indeed required to explain the lack pre-L mitochondria. Ultimately, it’s a math problem, so it should be possible to prove which way Occam’s razor cuts. It could be I’m wrong, and no explanation is required, and stochastic processes do provide the necessary statistical spread. Why create complex explanations for the lack archaic MtDNA in modern populations if a simple explanation will do? I suspect a statistical analysis could be done to answer this. We just need to know the likelihood of a specific MtDNA line going extinct given x number of generations and y population size. In the case of Neanderthals, given something like a 4% starting frequency, how likely is it that chance alone could remove those haplotypes from the population? This seems like it could be calculated by somebody with the math chops to do it. (not me, alas) I suspect the answer would be that it is reasonably likely. Or at least not extremely unlikely. So perhaps there need be no mystery there. But then you also need to calculate the likelihood that another 5% of Denisovan MtDNA would happen to drop out of populations in New Guinea. And then you calculate the odds that all pre-L MtDNA would happen to drop out of the admixed African populations. And you have to multiply those odds together to get the likelihood that chance alone would remove all the pre-L haplotypes. And for each new discovery of admixture (the paper I read said the models predicted something like 9?) you have to keep multiplying. It doesn’t take long before the odds get pretty slim that chance alone could explain it. Each new instance of admixture is it’s own spaceship into the future, and we’re the aliens comparing the different colonies a thousand generations later. In any one colony, you might expect to lose a given mitochondrial line if it only starts out at a 5% frequency. But given enough colonies, those rare older clades should pop up somewhere. So far at least, they haven’t.”

In the past, Kosmo brought up Australian Mungo Man DNA in the context of pre-L mtDNA stage. I may only add that the phylogeographic “gaps” in the haploid phylogenetic trees continue well into more “downstream” clades. For example, there are no attested mtDNA pre-M and pre-N lineages in Africa; Africa only has derived branches such as hg M1, N1 and U6; there are no Y-DNA CF lineages in Africa. mtDNA M and N lineages and Y-DNA C lineages explode into a myriad of lineages in East Asia, East India and Southeast Asia, which is thousands of miles away from Africa.