It’s another outlet for the same research that I’ve posted in !linguistics@mander.xyz, so I’ll summarise it here.

There’s an archaeological culture called Yamnaya, dated from ~3000 BCE, found in the following region:

I’ll call that region “the steppes” here. To simplify it, the genetic pool of the people behind that culture is found all across Europe, Iran and northern India, in a way that you’d automatically associate it with Proto-Indo-European speakers, as the distribution seems to fit rather well the distribution of IE languages in Eurasia.

Well, there’s a problem with this Yamnaya = PIE speakers association. See, in central Anatolia there were also a bunch of Indo-European languages, nowadays extinct: Hittite, Luwian, Palaic, Lycian, etc. They’re called the “Anatolian” languages. (No, linguists are not a creative bunch.) If this association between the Yamnaya culture and PIE is correct, you’d expect to see at least some really old Yamnaya ancestry in that region, predating any Greek/Roman invasion… and you don’t. At least, not in meaningful amounts. Why is that?

Now let’s look at the article. They identified another group, that they’re calling the Caucasus Lower Volga (CLV). That group seems to have mixed with another group, generating the people behind the Yamnaya culture; and to have also migrated to central Anatolia.

But wait, which is the PIE speakers group then - the CLV or Yamnaya? Well… both. See, linguists have for a long time claimed that what we call “Proto-Indo-European” is at least two languages:

• Late PIE - ancestor of the modern Indo-European languages. Russian? Hindi? English? Italian? Albanese? Greek? Yes.
• Early PIE - ancestor of both Late PIE and the Anatolian languages.

That is mostly based on information that we have about Hittite, that seems to have diverged from other Indo-European languages early. For example, it conserves a bunch of really archaic features, such as a simple animate vs. inanimate gender system (Late PIE likely had a masculine vs. feminine vs. neuter gender system instead), and sounds only predicted to have existed in PIE (the “laryngeals”) before Hittite was discovered.

What’s more is that there was always some dispute on where Proto-Indo-European was originally spoken: Anatolia (near Hittite), the steppes (Yamnaya) or the Caucasus. Accordingly to this study, Early PIE was spoken near the Caucasus, and Late PIE in the steppes.

Personally I think that, while this is not outright revolutionary, it’ll have a big impact for Historical Linguistics.

For a start, it pretty much forces us to handle Early PIE and Late PIE separately; we need separated reconstructions for them. And we probably should stop calling them this way, and instead pick other names (like Proto-Indo-Celtic and Proto-Indo-Anatolian), or something like that.

Another matter is that, the more you try to fix PIE reconstruction oddities, the more it resembles the Northwest Caucasian languages - languages known for having a bazillion consonants, 2~3 vowels, and a mostly agglutinative grammar. Perhaps we should be looking more at Adyghe and its siblings when reconstructing Early PIE = Proto-Indo-Anatolian, for potential areal features.

A third matter are the laryngeals and the odd T D Dʰ stop system. Treating the laryngeals as consonants in Early PIE seems sensible; but in Late PIE, perhaps we should be treating them as ultrashort vowels instead; most descendants would get rid of them, and Greek would merge them into the short vowels. And it’s perfectly possible that Late PIE already had a Tʰ T D Dʰ system, like the one attested for Sanskrit.

Double reply addressing HN comments:

I haven’t read the papers in detail, but can someone explain how genetics can be used to trace spread of languages? For context, you don’t need population movements for a language to spread (it is similar to religion). See this article for a logical explanation

Languages don’t spread out of nowhere, they do it through groups of people interacting with each other. And those interactions are also bound to introduce at least some genetic admixture.

Because of that, you can use the presence or absence of genetic admixture in a place as weak evidence for / against the presence of a language there. On itself it is not fail-proof, you’d need further evidence to claim with certainty “no, [language] was not here”, but it helps.

Writings on artifacts and burial practices associated with DNA fragments found at the burial sites.]

No written record.

Fun facts, the most common words of Indo-European Family are surprisingly very similar across Sanskrit (S) <–> English (E) <–> German (G) [3].

And so they are in other Indo-European branches, if you know where to look like. Compare for example English and Spanish:

• foot, father vs. pie, padre - /f/ vs. /p/
• three, thrush (the bird) vs. tres, tordo - /θ/ vs. /t/
• horn, what, hundred vs. cuerno, que, ciento - /h/ vs. /k/ or /θ/~/s/

There’s plenty words like this, where Spanish uses an unvoiced stop while English uses a fricative. Pehausse Frotho-Chermanih speahers Because Proto-Germanic speakers eventually shifted those consonants this way.

There are some complications though. See what I said about English /h/ vs. Spanish /θ/~/s/? That’s in words where Latin was still using /k/, like “centum” /kentũ/. (Those words stick out like a sore thumb if you pick Italian instead, you’ll see /tʃ/ instead.)

My favorite part is that the most foundational swear words in modern Slavic languages are still recognizable from their PIE roots:

So is the most popular Romance foul word: PIE *(s)merdh₂ “stench, stinging [smell]”. Latin inherited it as “merda” (shit) and it’s still up there.

Could you explain in non-specialist language how similarities between these modern languages now has anything to do with their relationship from some earliest common ancestor? How is that explanation better than convergent evolution or overfitting hallucinations?

I almost forgot those are Hacker News comments. Thanks for reminding me.

Those sound changes are systematic, as I showed above with English vs. Spanish.