alan little's weblog

origins of the indo-european languages

29th December 2003 permanent link

Continuing our late December theme of heated-but-probably-ultimately-futile historical controversies involving the term “Aryan”: apparently some geneticists think they have convincingly traced the origin of Indo-Euopean languages to around 9 to 10 thousand years ago in Anatolia. The original Nature article is subscription only; the abstract is here. And just like the last time a non-linguist came to this conclusion (the archaeologist Colin Renfrew in his 1987 book Archaeology and Language : The Puzzle of Indo-European Origins ), apparently linguists are unimpressed.

utter bollocks

part of their results are the fruit of begging the question, and the rest (circa 80%) is at the mercy of the “glottochronological clock”, which is as accurate as a sundial at night … the basic tenets of glottochronology are rubbish
Jacques Guy on sci.lang (original messages)

all their time estimates come from fourteen data points, deliberately chosen to have wide error bars … I should note that the selection of languages is dodgy as well
a commenter at languagehat

The geneticists in question are Russell Gray and Quentin Atkinson of the University of Auckland in New Zealand. They reached their conclusions by applying statical techniques normally used in evolutionary biology to linguistics. Linguists seem to have two basic issues with their conclusions. One is that they published in Nature – a journal with no linguistics expertise – and in the form of a five-page letter with nowhere near enough detail for a serious assessment of their methods. The other is that, from what little their letter describes of their method, it appears to be similar to a discredited technique in historical linguistics known as glottochronology, dating from the 1950s. Glottochronology attempted to date when languages that are known or assumed to have a common ancestor diverged by measuring differences in their vocabulary. Unfortunately the key assumption on which it was based – that the rate of vocabulary change over time in all languages is known and constant – is completely groundless. Gray and Atkinson are aware of this problem and claim that the statistical technique they used is somehow immune to it, but don’t provide any convincing explanation of how. It also seems to be unclear whether they did basic credibility checks, like using their technique to date historically known language splits (e.g. French, Italian and Spanish from Latin, Old English from Old German).

The most detailed explanation of the apparent problems that I’ve found is by Bill Poser in Language Log, a group weblog that is actually written by real linguists (oh no, not another bloody interesting weblog to read). There is another detailed critique by Peter Daniels in the sci.lang newsgroup – the whole thread is worth reading.

My attention was originally drawn to this by the gene expression weblog. The folks there seem to think the general hostility to Gray & Atkinson from linguists is at least partly a tribal reaction to outsiders trespassing on their turf. From what I’ve read, I disagree. It looks very much to me like the linguists are right – Gray & Atkinson at the very least have made some wild claims based on an unproven technique, and provided nowhere near enough explanation of what they have done to actually convince anybody.

Update on chariots: in the course of reading up on Gray & Atkinson, I came across references to chariots dated to 2100BC having been found in the southern Urals. A while ago I wrote at length about Robert Drews’ theories on chariots and Bronze Age warfare, and why I think the Rig Veda cannot be much older than western scholars generally think it is – i.e. somewhere around 1500 BC. The Urals chariot finds knock a rather large hole in Drews’ theory about the Greeks, which is based on the belief that the chariot was invented (a) south of the Caucasus and (b) not much earlier than about 1800 BC. I don’t see that it materially affects my opinion on the Vedas, though – chariots being known before 2100 BC on the Steppes still isn’t any kind of evidence that they were in use hundreds or thousands of years before that in northern India. More on this later.


related entries: Language Yoga

all text and images © 2003–2008