A massive data dive proves that languages and genes evolve together

Doesn't this perhaps show selection based on language? People who speaks similar languages are more likely to interact and reproduce, eventually narrowing the gene pool.

Linguists must be rolling their eyes at this study. If you want to study language evolution, phonemes are the last thing you should look at. Languages change, fast, and the sounds of a language change drastically over generations. Compare modern Icelandic and Old Norse (by which we usually mean Old Icelandic)—Icelandic is supposed to be one of the most conservative languages and is still quite close to Old Icelandic in spite of centuries of separation in areas like syntax and vocabulary. But if you compare the pronunciations, they are quite different. The sounds of a language are among its least stable aspects.

Similarities in pronunciation between languages spoken over the same region are part of a phenomenon well known to linguists and do not imply that the languages are related. For example in terms of a pronunciation feature called coda devoicing, Slovene dialects near the Austrian border will pattern with German, those near the Croatian border will pattern with Croatian, while Standard Slovene will have something in the middle. Slovene and German are only very distantly related.

There may be something there about closely related population groups ending up using similar sounds even if their languages are not closely related, but I doubt the researchers even thought about that and used phonemic distance as a shorthand for actual linguistic distance.

I am always very leery of big-ticket science published in the Proceedings of the National Academy of Sciences (PNAS), as this paper is. Many papers (including this one) are not subject to peer review because they use what has been called the "inside track" (http://www.nature.com/news/scientific-publishing-the-inside-...; though note that Nature is not a disinterested observer). This paper was "contributed" to PNAS, which means that the authors reached out a member of the National Academy of Sciences (usually a famous late-career scientist) who they believed would be sympathetic to the line of work, and asked this person to serve as contributor (essentially a type of action editor) on the piece. Importantly, this person (Marcus Feldman) happened to be a colleague of the first author (she appears to be a postdoc at Stanford, he is a professor at Stanford, and they have written papers together), but the conflict of interest is ignored. Unlike a normal action editor, however, the contributor is not expected to identify potential adversarial reviewers, or even to take negative reviews into account! As a result, a significant minority of PNAS "contributed" papers clearly wouldn't have made it past peer review elsewhere. Famously, the 'contributed' track has been widely abused (e.g., by Linus Pauling, who hoped to beat Crick & Watson to discover the structure of DNA--unfortunately, his structure is clearly wrong.)

I am not saying that this is the case for this paper. However, I note that there are no linguist authors, no linguist reviewers, and no linguist editors, involved in the process. And, as a linguist, I don't understand the motivation for this work. The presence of a phoneme in language X and language Y tells me next to nothing about the relationship between X and Y; there is no theory of "phoneme evolution" because the evolutionary metaphor is irrelevant at best (there is no analogue to selection). And we've known for centuries that unrelated languages in close contact may borrow phonemes from each other (retroflex t in South Asia being the parade example), and obviously close contact will lead to genetic transfer too.

This is interesting research, but it's also completely unsurprising. As human populations migrate, expand, collapse, invade, mix, and die out, both their culture and their genes will do all those same things.