Indian English is a Prakrit, not a creole, says linguist Peggy Mohan. This part jumped out at me:
You mention that women were likely barred from writing Vedic poetry. But weren’t there some Vedic poetesses?
[Philologist] Michael Witzel thinks no. It’s possible, but we do know that women were not allowed to speak Sanskrit after some time. Witzel says they were men speaking as women. It fits the pattern of women being different, speaking an earlier language.
In Vedic ceremonies, their role is extremely limited. It’s almost like the women were a separate ethnic group. And they had to be tolerated because there would be no children without them, but they were the people who, on another day, you were fighting on the battlefield.
The latest work I’ve seen shows even in northern Pakistani groups steppe-derived mtDNA is present at around 10% of the total. You can look at the raw results for the Sintashta in Narasimhan et al., and the discontinuity in the Y vs. mtDNA is striking.
One thing I want to moot is that the sex bias was even stronger as Indo-Aryans moved out of the Indus Valley. My hypothesis is that sons of mixed background would be more likely to migrate east and south than “pure-blood” Indo-Aryans. The imbalance between Y chromosomes and mtDNA is much higher east and south than in Pakistan.
Hypothesis has been proposed that verbal ability is less demanding than spatial ability on brain volume. So men’s advantage over women is on spatial ability, not verbal one. No surprise that STEM graduates are mostly males.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2668913/
People of smaller brain volume tend to do well on functions depending on verbal ability. As the research indicates, brain volume still matters for the quality in the end. However aggressive personality might compensate for lack of quality with first step advantages. High quality always wins at end. This pattern applies numerous phenomena in this world. Women with smaller brain volume is discussed in the article.
Witzel’s the “Munda languages before Indo-European” guy? Long term diglossia of men and women in the same ethnic group seems pretty out there.
@Razib: Can you please elaborate on what pure-blood Indo-Aryans would be more likely to stay back? I didn’t quite get it.
@vsds
I think the notion is that there would be more prestige living and working and ruling in “headquarters” where there was an established IE high culture in place, than there would be living and working in the periphery/branch office/frontier which was less civilized.
The eldest sons from the IE wives would have gotten first pick and had more to lose by adventuring into new lands than the younger sons/sons of local concubines who didn’t “look the part” compared to their higher prestige half-siblings. So, they had worse prospects staying put and more to gain relative to that by striking out to conquer new lands despite the ever present risk of failure.
In Europe, during primogeniture you see something similar between heirs apparent and younger children and children from cadet lines, with fairly elite upbringings but no future. Columbus, the man given credit for “discovering” the Americas, for example, had a name that was a generic name assigned to orphans (comparable to the name “Snow” of Jon Snow in Game of Thrones).
@ogwilleke: Thanks that makes sense.
But a possible counter to that could be that the Indo-Aryans saw conquering as a virtue. So there is more glory to be gained from conquest than administration ?
@Matt “Witzel’s the “Munda languages before Indo-European” guy? Long term diglossia of men and women in the same ethnic group seems pretty out there.”
Witzel’s thrown some curve balls, and his decision to name the Harappan substrate in Sanskrit “Para-Mundi” is surely among them, although he’s largely repented from that. He’s also done some very good work, however, and the same body of work’s insight that the early Rig Vedic substrate in not Dravidian is among that. So, I wouldn’t dismiss it out of hand.
It isn’t entirely unprecedented. The tribes in the Amazon that were first found to have Paleo-Asian ancestry actually practice linguistic exogamy – i.e. people only marry people who don’t speak the same language as they do. Mixed language married couples (esp. “mail order brides”) aren’t unknown today. In between, Henry V’s wife didn’t speak the same first language as he did although they were taught each other’s languages from tutors as children. Hell, my son’s current serious girlfriend doesn’t have the same first language as he does, although they are both fluent in his language.
It does seem implausible that girls and boys of the same couple would speak different languages from each other, although that would be less surprising once Sanskrit became a secondary liturgical language, along the line of ancient Hebrew in Jewish communities before its revival as a living language, with Hattic, Sumerian and Latin being other examples. Because it wouldn’t be surprising if girls were left out of anything requiring formal instruction in that era.
It doesn’t seem implausible, however, that a conquering IE elite would continue to speak their native language evolving somewhat through contact effects, while senior subjects who learn IE as a second language in connection with their jobs might have daughters who didn’t learn the IE language but are married to IE elite members for multiple successive generations, making a lack of shared language between husbands and wives common for several generations before the indigenous bourgeois eventually undergoes language shift to IE completely.
As I posted in a comment to another post, I’d love to see more relevant ethnography and historical fiction (e.g. from the households of the Conquistadors in South America, or from non-Arabic/Turkic speaking people conquered by linguistically Arabic or Turkic peoples during Islamic expansions), but I wouldn’t completely rule it out.
@vsds
“But a possible counter to that could be that the Indo-Aryans saw conquering as a virtue. So there is more glory to be gained from conquest than administration?”
Like so many things in this kind of field, there are good arguments to make *ex ante* for lots of different possibilities that contradict each other because it isn’t a hard science. But when we find out what really happened, we can try to concoct *post hoc* narratives to explain them.
The steppe women helped to delete the complete linguistic of Indian women(no remnants) leaving Dravidian side. Despite the fact, we dominate the landscape and I will never appreciate Sanskrit(burn the language).
I speak some sort of dardic language (pre-gandhari). I am R30b1 mtdna.
For my example word for hand: vocabulary is “hatta, hatta dasto, hapto.’
My father’s ancestors only practiced “vedic” , when their daughters were married in that family.
ohwilleke, yeah, that’s probably a fair comment re Witzel being actually good. I guess that was me rather basely indirectly “throwing shade” on a proposal I thought I found a bit farfetched.
I could see some kind of short term diglossia, and it makes sense Sanskrit becoming a frozen liturgical language which women were excluded from using makes some sense and there’s a vernacularized version that they speak… But it would be hard for me to imagine a situation where women are, without ongoing immigration into the community, raised to speak one language from an unrelated stock or anything like that. Esp for hundreds of years! It seems without parallel. And there are plenty of religious practices that exclude women without really needing to, y’know, have a perception of them as a separate tribe somehow (despite the males being raised by them).
(I do wonder if a long-term situation of IE with regards to possibly having a particularly integrating / “hungry” form of female exogamy relates to the early steppe star-founder effect that probably preceded the IE expansions. How much of later myths and to some extent practices were set down by probably early pattern in early steppe community of expansion of a few male lines which probably drew in women from a much larger set of sources *within* the steppe environment… I say myth as there wasn’t an X:A sex bias detected in the recent paper showing entry of steppe ancestry into Italy, but the Romans had a legend about abduction of female outgroup to found the community.)
I think the notion of an initially diglossic language situation where in some situations children are raised speaking one language from mother then learn another in adulthood would seem to offer some possibly plausible story for Basque or something like this, where some prominent persons/males have less interest in learning a second language as older children or see rejecting or not much using the adult language which they’re less proficient in as a way to assert their personal power. Or variations on this theme.
Not related to your comment but generally on sex bias in South Asia, I’m wondering if many more high quality ancient genomes will eventually provide enough SNPs on the X-chromosome to kill the standard error on qpAdm runs on it down to the same levels as the autosome, which is required to really do the estimations of sex-bias from X:A properly. The AGDP project has a Sintashta guy I1053 at a high level of coverage, and quite a few of the samples from Russia_MLBA_Krasnoyarsk who were largely placed by Narasimhan into Central_Steppe_MLBA (“We assign the main cluster of17 individuals the Krasnoyask_MLBA split label and theCentral_Steppe_MLBA analysis labelas they are genetically homogenous to other sites across the Central Steppe”). So again I wouldn’t be totally surprised if this was somewhere “in the works”.
The Narasimhan paper tried to do that with the X chromosome on the capture data – came out with female bias from the steppe ancestry in present-day 1000 Genomes South Asian populations which are Gujaratis, Pakistani Punjabis, Bengalis, Tamils and Telugus, who came out together 13.7% steppe on X vs 11.6% on autosomes, which requires a fairly substantial shift. But were unable to distinguish that from any other situation since the bounds of 5% standard error swamp the result, and both the fractions of overall Steppe ancestry in South Asia and the skews in the X generated by even quite extreme bias are low enough that this can’t tell us much.
(Comparably, the situation in African-Americans which would seem relatively extreme only generated a detectable difference of 24% autosomal European ancestry vs 20% X Chromosome ancestry, or 4% – https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4289685/. So on half the overall ancestry, but in a social history roughly parallel to Black Americans you might only expect a difference of 2%, or 12% vs 10%?)
(There also hasn’t been too much excess segment sharing on the autosome in IBD chunks between South Asians and Europeans, compared to the X, as far as I know.)
Of course, it may be that these high-quality ancient genomes are not very useful for this purpose, and mainly just give rare and disease associated alleles restricted to some particular sample, which while medically extremely interesting would not tell us much more about population relationships.
Mtdna and y-dna and this X:A bias should all tell the same story (it’s hard to see plausible situations where they get out of whack), but I guess I’d regard the X:A bias as the most definitive test of the *entire* historical dynamics. I still have an irrational mistrust of the uniparental markers.
“I’m wondering if many more high quality ancient genomes will eventually provide enough SNPs on the X-chromosome to kill the standard error on qpAdm runs on it down to the same levels as the autosome, which is required to really do the estimations of sex-bias from X:A properly.”
At any point in time, the autosome is always going to have on average 35x times as much data as X-chromosomes (for gender balanced sample sets), with a lower standard error due to statistical error. The systemic error should predominate fairly quickly, however, because the larger your sample size, the less of an impact on statistical error component of standard error additional data provides. To cut a 5% statistical margin of error to 2.5% requires four times as large of a sample. There first few samples matter a lot, but soon, the statistical error bars for autosomal and X-chromosomes should get pretty similar. Systemic error shouldn’t scale meaningfully with sample size, its a function of the quality of your lab equipment and not how many samples you have. At even modest sample sizes, systemic error should start to dominate over sample size related statistical error.
“I still have an irrational mistrust of the uniparental markers.”
Not many researchers do multidisciplinary research, or even integrate their Y-DNA data, the global autosomal data, their ancestry informative autosomal data, their X-chromosome data, their mtDNA data, their carbon-14 data, their strontium data, their bone measurement data, their ancient DNA and their modern DNA data into a combined analysis.
But just integrating X-chromosome data and mtDNA ancient DNA data into a single analysis and finding models that fit (even if this is hard to do analytically, you can do Monte Carlo modeling and the great thing about genetics is that we can computationally model it almost perfectly and it still doesn’t take nearly the computing power of cosmology and particle physics calculations), you can pretty significantly reduce standard errors in your model based on joint analysis of the X-chromosome and mtDNA data, ruling out a lot of outlier possibilities with the two logically related data sets that can be correlated with each other at the individual sample level. My mathematical intuition is that with the right analysis, you can probably cut the standard error in half with existing data just by correlating X-chromosome and mtDNA data into a single analysis that also makes both sets of conclusions more robust against “unknown unknowns” in addition to reducing the standard error that you can quantify.
Also, if you are interdisciplinary, this tends to allow for demographic narratives that are easier to fit than overly conservative WEIRD influenced constraints that paleogenetics analysis is prone to, that rule out in the guts of the assumptions in the models possibilities that stray too far from contemporary or recent historical norms. The variation in number of women having children with a given man, for example, has a much broader range in the Bronze and Iron Ages than in did in the Medieval era, with more men having lots and lots of baby mamas, and more men shut out of opportunities to reproduce entirely. Standard modeling also tends to assume away population structure that was almost surely present.
Given that South Asia is far and away the world record holder when it comes to the magnitude and completeness of its non-geographically enforced endogamy (1500 years with almost no exogamy in Jati), it wouldn’t be stunning if South Asia was also pushing the envelope in some other respects, including linguistic ones. There is a case to be made for South Asian exceptionalism.
Ultimately, I don’t think any of that is going to change the basic picture very much, however. It will just nudge the fine details of the parameters and will localize those parameters better.
@ohwilleke, that’s an interesting suggestion to use multiple datasets as a check (I think?) on systematic error from one method. I guess one thing with adding mtdna as an additional independent dataset to increase p is, well, how confident are we that mtdna is a sufficiently neutrally evolving line?
This is vivid in my mind not just because of the extreme examples (Neanderthal mtdna replacement) but because of a recent study last week, with an example news report here – https://www.cam.ac.uk/research/news/mothers-can-influence-offsprings-height-lifespan-and-disease-risk-in-unexpected-ways-through-their – The study, involving data from more than 350,000 participants in the UK Biobank (UKBB), found that genetic variants in mitochondrial DNA passed to offspring could increase the risk of developing different conditions (examples given type 2 diabetes (T2D) and multiple sclerosis) , as well as influence characteristics such as height and lifespan. There was also evidence that some changes in mitochondrial DNA were more common in people with Scottish, Welsh or Northumbrian genetic ancestry, implying that mitochondrial DNA and nuclear DNA (which accounts for 99.9% of our genetic make-up) interact with each other.. They don’t seem to think the mtdna-nuclear correlation is due to population structure… (“It looks like our mitochondrial DNA is matched to our nuclear DNA to some extent – in other words, you can’t just swap the mitochondria with any donor, just as you can’t take a blood transfusion from anyone,” explained Professor Chinnery.)
If you have these very low level selection effects and nuclear incompatabilities (and I might expect those to be somewhat larger in a genetically more divergent population admixture scenario?), then over about 3000 years (90 generations), you might still get some compounding of these differences into something that has a meaningful effect if you’re comparing, like 10% ancestry to 5% mtdna or something and drawing a conclusion.
(Some folk also think the relative rise of mtdna H in post-Neolithic Europe is some kind of selective phenomena, for example.)
Hence I guess why I would prefer if we could get like 200,000 SNPs on the X or something like that. (I think per this paper – https://www.biorxiv.org/content/10.1101/2020.07.02.184507v1.full – “1,055,186 autosomal SNPs and also 49,711 X chromosome positions available in the Lazaridis et al ancient dataset” (roughly the 1240k dataset?)).
That said, the X might have these issues as well – it’s notable that modern humans today have no Neanderthal mtdna, but also have “gene deserts” of Neanderthal ancestry on the X, and these happened quite quickly, being evident even in early paleolithic Zlaty Kun / Bacho Kiro. The X chromosome is massively conserved compared to other parts of the genome, isn’t it? Mtdna is probably still worth looking at. But hopefully clearly neutrally evolving sites can be identified for either part of the genome.
(Y-dna also has some potential issues over long-term, like this – https://elifesciences.org/articles/65420 – which indicates that the R1a1-M458 haplogroup is prone to a deletion causing azoospermia. Though expect that to be more overruled by social processes more often.)
Is K1a a steppe mtDNA or IVC or neither? Not really sure where mine comes from.
My Y H is pretty easy. Probably IVC or some tribal AASI dude
@Matt I would be quite surprised if, in the kind of contexts where you are contemplating using it, intrinsic selective fitness differences of mtDNA haplogroups and X chromosome DNA weren’t dwarfed by cultural selection factors in the short run.
Low magnitude intrinsic selective fitness advantages accumulate over many generations, which make these a worry if you are trying to determine relative ancestry components from a hundred or more generations prior from modern population genetics, but for intrinsic selective fitness issues to deviate far from an intrinsically selective fitness neutral model so close to the time of admixture in ancient DNA would require a very strong intrinsic selective fitness effect that would virtually wipe out the less selectively fit mtDNA haplogroups by modern times. The intrinsic selective fitness effects of any mtDNA haplogroup present in both modern and ancient populations can be very strongly bounded in mangitude.
Also, in the kind of joint analysis that I’m contemplating, essentially what you are doing is reducing standard error by suppressing the tails of possible X chromosome and mtDNA variation probability distributions, which have an outsized effect on the standard error. This is because even if one extreme narrative is hypothetically possible in one of these subsets of data, it usually isn’t possible in the other one. The benefits of this kind of association should be pretty robust.
The British example you cite involves intrapopulation variation with many centuries of panmixia. You don’t have that in Indo-Aryan v. Harappan admixture models with the admixture of two long separated populations.
FWIW, I think that the lack of Neanderthal mtDNA in modern humans is mostly “culturally driven” not by intrinsic selective fitness. I think Neanderthal-human hybrids were raised in their mother’s tribes and the Neanderthal mom hybrids vanished when their species went extinct.
@thewarlock
mtDNA K1a is probably neither South Asian nor Steppe in origin. Its modern distribution geographically is suggestive of an Anatolian and Levantine distribution prior to the Neolithic revolution, that was a minor component of the Neolithic revolution’s expansion from these regions into Europe and North Africa. But it has dispersed at low frequencies into South Asia, Iran, Central Asia and even China. But, is correlated in some distributions with Y-DNA R1b.
Realistically, you are looking a narrative of a steppe man taking a wife from a conquered Neolithic farmer community derived remotely from Western Anatolian farmers, who ends up migrating to South Asia where his daughter marries a local Y-DNA H man and they have a child (although probably with these events spread over many generations and being far more twisted, rather than in a neat and tidy three generation sequence).
@ohwilleke, yes, I don’t think that would pose a problem when looking at a transect close to time of admixture, though sampling problems might be more significant close to time of admixture. Our Swat_IA sample had no R1a, and I don’t know that it had either any steppe mtdna, despite having steppe ancestry. Some have suggested this is due to sampling limitations due to burial practices that don’t reflect the entire admixing populations.
Also, for a relatively modern example (about 1/5-1/10 the time depth) on the topic this paper by Zaidi and collaborator may interest you – https://www.nature.com/articles/s41559-018-0766-1
“Here, we explored potential mitonuclear incompatibility in six admixed human populations from the Americas: African Americans, African Caribbeans, Colombians, Mexicans, Peruvians and Puerto Ricans. By comparing nuclear versus mitochondrial ancestry in these populations, we first show that mitochondrial DNA (mtDNA) copy number decreases with increasing discordance between nuclear and mtDNA ancestry. The direction of this effect is consistent across mtDNA haplogroups of different geographic origins. This observation indicates suboptimal regulation of mtDNA replication when its components are encoded by nuclear and mtDNA genes with different ancestry. Second, while most populations analysed exhibit no such trend, in African Americans and Puerto Ricans, we find a significant enrichment of ancestry at nuclear-encoded mitochondrial genes towards the source populations contributing the most prevalent mtDNA haplogroups (African and Native American, respectively)”.
They also found in 4/6 populations, a similar level of mtdna was present as estimated from X:A ratio (mostly in populations with very low levels of female ancestry from Europeans), but also an effect in the samples from Colombian and Puerto Rican with intermediate European ancestry had some different mtdna ratios of Native American, European mtdna distributions than suggested by the X:A ratios, and ran a statistical test which they believed excluded it being due to random sampling / drift (in their roughly 100 samples for each). None of their results replicated in all populations of course.
Re Neanderthals, Zaidi also notes, “a recent study has also shown that nuclear-encoded mitochondrial genes are significantly underrepresented in Neanderthal, but not Denisovan, introgressed regions in the human genome. Such a depletion is expected because there is no evidence of introgression of Neanderthal and Denisovan mtDNA into mod-ern humans. Therefore, if admixture between Neanderthals/Denisovans and humans results in mitonuclear incompatibility, selection should favour human-specific alleles at nuclear-encoded mitochondrial genes” which could reflect either causality (mtdna selected out because of nuclear incompatibility, or selection towards modern human nuclear-encoded mitochondrial variants due to modern human mtdna).