Obviously, the Tocharians did not reach Chinese Turkestan by helicopter. They surely walked or rode there, and we have to assume that central Asia formerly had many other Indo-European languages that disappeared without the good fortune to be preserved by documents in secret chambers. A modern linguistic map of Eurasia (see page 228) makes obvious what must have happened to Tocharian and all those other lost Indo-European languages of central Asia. That whole area today is occupied by people speaking Turkic or Mongolian languages, descendants of hordes that overran the area from the time of at least the Huns to Genghis Khan. Scholars debate whether Genghis Khan's armies slaughtered 2,400,000 or only 1,600,000 people when they captured Harat, but scholars agree that such activities transformed the linguistic map of Asia. In contrast, most Indo-European languages known to have disappeared in Europe—like the Celtic languages Caesar found spoken in Gaul—were replaced by other Indo-European languages. The apparently European centre of gravity of Indo-European languages as of 1492 was actually an artifact of recent linguistic holocausts in Asia. If the PIE homeland really was centrally located in what became the Indo-European realm by 600 AD, stretching from Ireland to Chinese Turkestan, then that homeland would have been in the Russian steppes north of the Caucasus, rather than in Western Europe.

Just as the languages themselves gave us some clues to the time of PIE's break-up, so too they contain clues to the location of the PIE homeland. One clue is that the language family to which Indo-European has the clearest connections is Finno-Ugric, the family that includes Finnish and other languages native to the forest zone of north Russia (see map on page 228). Now it is true that the links between Finno-Ugric and Indo-European languages are enormously weaker than those between German and English, which stem from the fact that the English language was brought to England from northwest Germany only 1,500 years ago. The links are also much weaker than those between the Germanic and Slavic language branches of Indo-European, which probably diverged a few thousand years ago. Instead, the links suggest a much older propinquity between the speakers of PIE and of proto-Finno-Ugric. But since Finno-Ugric comes from the north Russian forests, that suggests a PIE homeland in the Russian steppe south of the forests. In contrast, if PIE had arisen much further south (say, in Turkey), the closest affinities of Indo-European might have been with the ancient Semitic languages of the Near East.

A second clue to the PIE homeland is the non-Indo-European vocabulary swept up as debris into quite a few Indo-European languages. I mentioned that this debris is especially noticeable in Greek, and it is also conspicuous in Hittite, Irish, and Sanskrit. That suggests that those areas used to be occupied by non-Indo-Europeans and were later invaded by Indo-Europeans. If so, the PIE homeland was not Ireland or India (which almost no one suggests today anyway), but it also was not Greece or Turkey (which some scholars still do suggest).

Conversely, the modern Indo-European language still most similar to PIE is Lithuanian. Our first preserved Lithuanian texts, from around 1500 AD, contain as high a fraction of PIE word roots as did Sanskrit texts of nearly 3,000 years earlier. The conservatism of Lithuanian suggests that it has been subject to few disturbing influences from non-Indo-European languages and may have remained near the PIE homeland. Formerly, Lithuanian and other Baltic languages were more widely distributed in Russia, until Goths and Slavs pushed the Baits back to their current shrunken domain of Lithuania and Latvia. Thus, this reasoning too suggests a PIE homeland in Russia. A third clue comes from the reconstructed PIE vocabulary. We already saw how its inclusion of words for things familiar in 4000 BC, but not for things unknown until 2000 BC, helps date the time when PIE was spoken. Might it also pinpoint the place where PIE was spoken? PIE includes a word for snow (snoighwos), suggesting a temperate rather than tropical location and providing the root of our English word 'snow'. Of the many wild animals and plants with PIE names (like mus meaning mouse), most are widespread in the temperate zone of Eurasia and help to pin down the homeland's latitude but not its longitude.

To me, the strongest clue from the PIE vocabulary is what it lacked rather than included—words for many crops. PIE speakers surely did some farming, since they had words for plough and sickle, but only one word for an unspecified grain has survived. In contrast, the reconstructed proto-Bantu language of Africa, and the proto-Austronesian language of Southeast Asia, have many crop names. Proto-Austronesian was spoken even longer ago than PIE, so that modern Austronesian languages have had more time to lose those old names for crops than have the modern Indo-European languages. Despite that, the modern Austronesian languages still contain far more old names of crops. Hence PIE speakers probably actually had few crops, and their descendants borrowed or invented crop names as they moved to more agricultural areas. That conclusion presents us with a double puzzle. Firstly, by 3500 BC farming had become the dominant way of life in almost all of Europe and much of Asia. That severely narrows down the-possible choices for the PIE homeland; it must have been an unusual area where farming was not so dominant. Secondly, it begs the question why PIE speakers were able to expand. A major cause of the Bantu and Austronesian expansions was that the first speakers of those language families were farmers, spreading into areas occupied by hunter-gatherers whom they could outnumber or dominate. For PIE speakers to have been rudimentary farmers invading a farming Europe turns historical experience on its head. Thus, we cannot solve the 'where' of Indo-European origins until we have come to grips with the hardest question: why?

In Europe just before the age of writing, there were not one but two economic revolutions so far-reaching in impact that they could have caused a linguistic steamroller. The first was the arrival of farming and herding, which originated in the Near East around 8000 BC, leapt from Turkey to Greece around 6500 BC, and then spread north and west to reach Britain and Scandinavia. Farming and herding permitted a large increase in human population numbers over those previously sustainable by hunting and gathering alone (Chapter Ten). Colin Renfrew, Professor of Archaeology at the University of Cambridge in England, recently published a thought-provoking book arguing that those farmers from Turkey were the PIE speakers who brought Indo-European languages to Europe.

My first reaction to reading Renfrew's book was, 'Of course, he must be right! Farming had to produce a linguistic upheaval in Europe, just as it did in Africa and Southeast Asia. This is especially likely since, as geneticists have shown, those first farmers made the biggest contribution to the genes of modern Europeans.

But—Renfrew's theory ignores or dismisses all the linguistic evidence. Farmers reached Europe thousands of years before the estimated arrival of PIE. The first farmers lacked, and PIE speakers possessed, innovations like ploughs, wheels, and domesticated horses. PIE is strikingly deficient in words for the crops that defined the first farmers. Hittite, the oldest known Indo-European language of Turkey, is not the Indo-European language closest to pure PIE, as one might expect from Renfrew's Turkey-based theory, but is instead the most deviant language and the one least Indo-European in its vocabulary. Renfrew's theory rests on nothing more than a syllogism: farming probably caused a steamroller, the PIE steamroller requires a cause, so farming is assumed to have been that cause. Everything else suggests that farming instead brought to Europe the older languages that PIE overran, like Etruscan and Basque.