Biological dig for the roots of language

Nicholas Wade NYT Thursday, March 18, 2004


NEW YORK Once upon a time, there were very few human languages and perhaps only one, and if so, all of the 6,000 or so languages spoken today must be descended from it. If that family tree of human language could be reconstructed, a wonderful new window would be opened onto the human past.

Yet in the view of many historical linguists, the chances of drawing up such a tree are virtually nil and those who suppose otherwise are chasing a tiresome delusion. Languages change so fast, the linguists point out, that their genealogies can be traced back only a few thousand years at best before the signal dissolves completely into noise: Witness how hard Chaucer is to read just 600 years later.

But the linguists' problem has attracted a new group of researchers who are more hopeful of success. They are biologists who have developed sophisticated mathematical tools for drawing up family trees of genes and species. Because the same problems crop up in both gene trees and language trees, the biologists are confident that their tools will work with languages, too.

Their latest foray onto the linguists' turf is a reconstruction of the Indo-European family of languages by Russell Gray, an evolutionary biologist at the University of Auckland in New Zealand. The family includes such extinct languages as Hittite of ancient Turkey, and Tokharian, once spoken in Central Asia, as well as Indian languages and Iranian in one major branch and all European languages except Basque in another.

Gray's results, published in November in Nature with his colleague Quentin Atkinson, have major implications, if correct, for archaeology as well as for linguistics. The shape of his tree is unsurprising - it arranges the Indo-European languages in much the same way as linguists do, using conventional methods of comparison. But the dates he puts on the tree are radically older.

Gray's calculations show that the ancestral tongue known as proto-Indo-European existed about 8,700 years ago (give or take 1,200 years), making it considerably older than linguists have assumed is likely. The age of proto-Indo-European bears on a longstanding archaeological dispute. Some researchers, following the lead of Marija Gimbutas, who died in 1994, believe that the Indo-European languages were spread by warriors moving from their homeland in the Russian steppes, north of the Black and Caspian Seas, some time after 6,000 years ago.

A rival theory, proposed by Colin Renfrew of the University of Cambridge, holds that the Indo-Europeans were the first farmers who lived in ancient Turkey and that their language expanded not by conquest but with the spread of agriculture 10,000 to 8,000 years ago.

Gray's date, if accepted, would support the Renfrew position. Several linguists said Gray's tree was the right shape, but added that it told them nothing fresh, and that his dates were way off. "This method is not giving anything new," said Jay Jasanoff, a Harvard expert on Indo-European. As for the dates, he said, "The numbers they have got seem extremely wrong to me."

In the biologists' camp, however, there is a feeling that the linguists do not yet fully understand how well the new techniques sidestep the pitfalls of the older method. The lack of novelty in Gray's tree of Indo-European languages is its best feature, biologists say, because it validates the method he used to construct it.

Most historical linguists know a few languages very well but less often consider the pattern of change affecting many languages, said Mark Pagel, an evolutionary biologist at the University of Reading. "The field is being driven by people who are not confronted with the broad sweep of linguistic evolution," he said, "and is being invaded by people like me who are only interested in the broad sweep."

Glottochronology was invented by the linguist Morris Swadesh in 1952. It is based on the compiling of a core list of 100 or 200 words that Swadesh believed were particularly resistant to change. Languages could then be compared on the basis of how many cognate words on a Swadesh list they shared in common.

Cognates are verbal cousins, like the Greek podos and the English foot, both descended from a common ancestor. The more cognates two languages share, the more recently they split apart. Swadesh and others then tried to quantify the method, deriving the date that two languages split from their percentage of shared cognates. The method gave striking results, considering its simplicity, but not all of the findings were right. Glottochronology suffered from several problems. It assumed that languages changed at a constant rate, and it was vulnerable to unrecognized borrowings of words by one language from another, making them seem closer than they really were.

Because of these and other problems, many linguists have given up on glottochronology, showing more interest in an ingenious dating method known as linguistic paleontology.

The idea is to infer words for items in the material culture of an early language and to correlate them with the appearance of such items in the archaeological record. Cognates for the word wheel exist in many branches of the Indo-European family tree, and linguists are confident that they can reconstruct the ancestral word in proto-Indo-European. It is, they say, "k'ek'los," the presumed forebear of words like "chakras," meaning wheel or circle in Sanskrit, "kuklos," meaning wheel or circle in Greek, as well as the English word "wheel."

The earliest wheels appear in the archaeological record around 5,500 years ago. So the proto-Indo-European language could not have started to split into its daughter tongues much before that date, some linguists argue. If the wheel was invented after the split, each language would have a different or borrowed word for it.

The dates on the earliest branches of Gray's tree are some 2,000 years earlier than the dates arrived at by linguistic paleontology.

"Since 'wheel' is shared by Tocharian, Greek, Sanskrit and Germanic," said Bill Darden, an expert on Indo-European linguistic history at the University of Chicago, "and there is no evidence for wheels before the fourth millennium B.C., then having Tokharian split off 7,900 years ago and Balto-Slavic at 6,500 years ago are way out of line."

Gray, however, defends his dates, and points out a flaw in the wheel argument. What the daughter languages of proto-Indo-European inherited, he says, was not necessarily the word for wheel but the word "k'el," meaning "to rotate," from which each language may independently have derived its word for wheel. If so, the speakers of proto-Indo-European could have lived long before the invention of the wheel.