Monday, July 20, 2015

Review: The Indo-European Controversy (Pereltsvaig, Lewis)

A Review: The Indo-European Controversy: Facts and Fallacies in Historical Linguistics (Asya Pereltsvaig, Martin W. Lewis)

A few years ago, a team of researchers lead by Russell Gray and Quentin Atkinson presented a mathematical model for the spread of language families. Applying this model in reverse to the Indo-European languages supported the Anatolian hypothesis, a minority position on the location of the Indo-European urheimat.

For some reason, this was widely published in media, and the paper Mapping the Origins and Expansion of the Indo-European Language Family appeared at the same time in the journal Science. Gray and Atkinson have made very vocal and powerful claims about their findings: 'decisive support [for the Anatolian hypothesis]' is among the various things they have said about their own work.

Pereltsvaig and Lewis go over the results in depth, and find them highly lacking: they find numerous problems in the geographical spread that it presents, including multiple instances where the sanity of the model is excruciatingly questionable. They present the evidence we have against the Anatolian hypothesis (and even more clearly, the evidence we have against the Gray-Atkinson version of the Anatolian hypothesis) and all the difficulties it brings with it, as well as the evidence we have for the Steppe hypothesis.

They also present the evidence that has lead most linguists to accept Steppe Hypothesis instead. 

The argumentation is persuasive and clear, well-nigh undeniable. This leads to an important question: how did Science let a paper that is so rife with unsound historical linguistics pass peer review? It turns out that linguists did peer review it, and Science ignored their judgment, because their negative comments did not pertain to the maths of the model - clearly, having a mathsy model is a guarantee that the mathsy model is correct in Science's view?

Publications such as Business Insider either repost bad science from the Gray-Atkinson team, or add their own even worse spin to it. Consider their version of the Gray-Atkinson animated map. This is, allegedly, how "Language" spread across Europe. In linguistics, "Language" signifies the general phenomenon, the fact that humans can communicate in a complicated system. So if we are to take Business Insider's video title seriously, this is how the ability to speak spread in Europe, and all the current language families were the first languages spoken in their areas. 

Pereltsvaig and Lewis point out a very real problem: other scientists apparently do not take linguistics seriously, and we are facing a rise of armchair philosophers who disdain empiricism in favour of cute models (at least when going outside of their own field - i.e. Gray and Atkinson probably understand how to be scientific in their own field, but when working with language, the computational model seemingly blinds them to empirical facts). 

This, in turn, is coupled with the modern phenomenon of clickbaiting, where the most attention-attracting claim is more likely than other claims to pull in ad money, and thus scientific claims are propagated online not by their likelihood of being accurate, but by how tittilating they are. This is a genuine problem, and needs to be curbed.

Pereltsvaig's and Lewis's book is less combative than this review, although at times it does take vigorous swings at the Gray-Atkinson teams publications. It is a good read, and gives a lot of information about historical linguistics and especially Indo-European historical linguistics. A certain glimpse into issues in the philosophy of science can also be gleaned. It is well written: both clear, enjoyable and relevant, and it does something wonderful: fight pseudoscience.

1 comment:

  1. Thank you for a great review and for emphasizing the important contributions of our book! We tried to contain our combativeness and the general dismay at the sort of work we've criticized so it's great to hear that we've succeeded. What a great phrase about "the sanity of the model [being] excruciatingly questionable" :)