The Horse, the Wheel, and Language: How Bronze-Age Riders From the Eurasian Steppes Shaped the Modern World
Page 10
WAGONS AND THE ANATOLIAN HOMELAND HYPOTHESIS
The wagon vocabulary is a key to resolving the debate about the place and time of the Proto-Indo-European homeland. The principal alternative to a homeland in the steppes dated 4000–3500 BCE is a homeland in Anatolia and the Aegean dated 7000–6500 BCE. Colin Renfrew proposed that Indo-Hittite (Pre-Proto-Indo-European) was spoken by the first farmers in southern and western Anatolia at sites such as Çatal Höyük dated about 7000 BCE. In his scenario, a dialect of Indo-Hittite was carried to Greece with the first farming economy by pioneer farmers from Anatolia about 6700–6500 BCE. In Greece, the language of the pioneer farmers developed into Proto-Indo-European and spread through Europe and the Mediterranean Basin with the expansion of the earliest agricultural economy. By linking the dispersal of the Indo-European languages with the diffusion of the first farming economy, Renfrew achieved an appealingly elegant solution to the problem of Indo-European origins. Since 1987 he and others have shown convincingly that the migrations of pioneer farmers were one of the principal vectors for the spread of many ancient languages around the world. The “first-farming/language-dispersal” hypothesis, therefore, was embraced by many archaeologists. But it required that the first split between parental Indo-Hittite and Proto-Indo-European began about 6700–6500 BCE, when Anatolian farmers first migrated to Greece. By 3500 BCE, the earliest date for wagons in Europe, the Indo-European language family should have been bushy, multi-branched, and three thousand years old, well past the period of sharing a common vocabulary for anything.13
The Anatolian—origin hypothesis raises other problems as well. The first Neolithic farmers of Anatolia are thought to have migrated there from northern Syria, which, according to Renfrew’s first-farming/ language-dispersal hypothesis, should have resulted in the spread of a north Syrian Neolithic language to Anatolia (figure 4.7). The indigenous languages of northern Syria probably belonged to the Afro-Asiatic language phylum, like Semitic and most languages of the lowland Near East. If the first Anatolian farmers spoke an Afro-Asiatic language, it was that language, not Proto-Indo-European, that should have been carried to Greece.14 The earliest Indo-European languages documented in Anatolia—Hittite, Palaic, and Luwian—showed little diversity, and only Luwian had a significant number of speakers by 1500 BCE. All three borrowed extensively from non–Indo-European languages (Hattic, Hurrian, and perhaps others) that seem to have been older, more prestigious, and more widely spoken. The Indo-European languages of Anatolia did not have the established population base of speakers, and also lacked the kind of diversity that would be expected had they been evolving there since the Neolithic.
Figure 4.7 The spread of the first farming economy into Anatolia, probably by migration from the Core Area in northern Syria, about 7500 BCE. The first pioneer farmers probably spoke an Afro-Asiatic language. After Bar-Yosef 2002.
Phylogenetic Approaches to Dating Proto-Indo-European
Still, the Anatolian-origin hypothesis has support from new methods in phylogenetic linguistics. Cladistic methods borrowed from biology have been used for two purposes: to arrange the Indo-European languages in a chronological order of branching events (discussed in the previous chapter); and to estimate dates for the separation between any two branches, or for the root of all branches which is a much riskier proposition. Attaching time estimates to language branches using evolutionary models based on biological change is, at best, an uncertain procedure. People intentionally reshape their speech all the time but cannot intentionally reshape their genes. The way a linguistic innovation is reproduced in a speech community is quite different from the way a mutation is reproduced in a breeding population. The topography of language splits and rejoinings is much more complex and the speed of language branching far more variable. Whereas genes spread as whole units, the spread of language is always a modular process, and some modules (grammar and phonology) are more resistant to borrowing and spread than others (words).
Russell Gray and Quentin Atkinson attempted to work around these problems by processing a cocktail of cladistic and linguistic methods through computer programs. They suggested that pre-Anatolian detached from the rest of the Indo-European community about 6700 BCE (plus or minus twelve hundred years). Pre-Tocharian separated next (about 5900 BCE), then pre-Greek/Armenian (about 5300 BCE), and then pre–Indo-Iranian/Albanian (about 4900 BCE). Finally, a super-clade that included the ancestors of pre–Balto-Slavic and pre–Italo-Celto-Germanic separated about 4500 BCE. Archaeology shows that 6700–6500 BCE was about when the first pioneer farmers left Anatolia to colonize Greece. One could hardly ask for a closer match between archaeological and phylogentic dates.15 But how can the presence of the wagon vocabulary in Proto-Indo-European be synchronized with a first-dispersal date of 6500 BCE?
The Slow Evolution Hypothesis
The wagon vocabulary cannot have been created after Proto-Indo-European was dead and the daughter languages differentiated. The wagon/wheel terms do not contain the sounds that would be expected had they been created in a later daughter language and then borrowed into the others, whereas they do contain the sounds predicted if they were inherited into the daughter branches from Proto-Indo-European. The Proto-Indo-European origin of the wagon vocabulary cannot be rejected, as it consists of at least five classic reconstructions. If they are in fact false, then the core methods of comparative linguistics—those that determine “genetic” relatedness—would be so unreliable as to be useless, and the question of Indo-European origins would be moot.
But could the wagon/wheel vocabularies have been created independently by the speakers of each branch from the same Proto-Indo-European roots? In the example of *kwekwlos ‘wheel’, Gray suggested (in a comment on his homepage) that the semantic development from the verb *kwel–’turn’ to the noun wheel ’the turner’ was so natural that it could have been repeated independently in each branch. One difficulty here is that at least four different verbs meaning “turn” or “roll” or “revolve” are reconstructed for Proto-Indo-European, which makes the repeated independent choice of *kwel– problematic.16 More critical, the Proto-Indo-European pronunciations of *kwel- and the other wagon terms would not have survived unchanged through time. They could not have been available frozen in their Proto-Indo-European phonetic forms to speakers of nine or ten branches that originated at different times across thousands of years. We cannot assume stasis in phonetic development for the wheel vocabulary when all the rest of the vocabulary changed normally with time. But what if all the other vocabulary also changed very slowly?
This is the solution Renfrew offered (figure 4.8). For the wagon/wheel vocabulary to be brought into synchronization with the first-farming/language-dispersal hypothesis, Proto-Indo-European must have been spoken for thirty-five hundred years, requiring a very long period when Proto-Indo-European changed very little. Pre-Proto-Indo-European or Indo-Hittite was spoken in Anatolia before 6500 BCE. Archaic Proto-Indo-European evolved as the language of the pioneer farmers in Greece about 6500–6000 BCE. As their descendants migrated northward and westward, and established widely scattered Neolithic communities from Bulgaria to Hungary and Ukraine, the language they carried remained a single language, Archaic Proto-Indo-European. Their descendants paused for several centuries, and then a second wave of pioneer migration pushed across the Carpathians into the North European plain between about 5500 and 5000 BCE with the Linear Pottery farmers. These farming migrations created Renfrew’s Stage 1 of Proto-Indo-European, which was spoken across most of Europe between 6500 and 5000 BCE, from the Rhine to the Dnieper and from Germany to Greece. During Renfrew’s Proto-Indo-European Stage 2, between 5000 and 3000 BCE, archaic Proto-Indo-European spread into the steppes and was carried to the Volga with the adoption of herding economies. Late Proto-Indo-European dialectical features developed, including the appearance of “thematic” inflections such as o-stems, which occur in all the wagon/wheel terms. These late features were shared across the Proto-Indo–European–speaking region
, which comprised two-thirds of prehistoric Europe. The wagon vocabulary appeared late in Stage 2 and was adopted from the Rhine to the Volga.17
Figure 4.8 If Proto-Indo-European spread across Europe with the first farmers about 6500–5500 BCE, it must have remained almost unchanged until about 3500 BCE, when the wheeled vehicle vocabulary appeared. This diagram illustrates a division into just three dialects in three thousand years. After Renfrew 2001.
It seems to me that this conception of Proto-Indo-European contains three fatal flaws. First, for Proto-Indo-European to have remained a unified dialect chain for more than thirty-five hundred years, from 6500 to 3000 BCE, would require that all its dialects changed at about the same rate and that the rate was extraordinarily slow. A homogeneous rate of change across most of Neolithic Europe is very unlikely, as the rate of language change is affected by a host of local factors, as Sheila Embleton showed, and these would have varied from one region to the next. And for Proto-Indo-European only to have evolved from its earlier form to its later form in thirty-five hundred years would require a pan-European condition of near stasis in the speed of language change during the Neolithic/Eneolithic, a truly unrealistic demand. In addition, Neolithic Europe evinces an almost incredible diversity in material culture. “This bewildering diversity,” as V. Gordon Childe observed, “though embarrassing to the student and confusing on a map, is yet a significant feature in the pattern of European prehistory.”18 Long-established, undisturbed tribal languages tend to be more varied than tribal material cultures (see chapter 6). One would therefore expect that the linguistic diversity of Neolithic/Eneolithic Europe should have been even more bewildering than its material-culture diversity, not less so, and certainly not markedly less.
Finally, this enormous area was just too big for the survival of a single language under the conditions of tribal economics and politics, with foot travel the only means of land transport. Mallory and I discussed the likely scale of tribal language territories in Neolithic/Eneolithic Europe, and Nettles described tribal language geographies in West Africa.19 Most tribal cultivators in West Africa spoke languages distributed over less than 10,000 km2. Foragers around the world generally had much larger language territories than farmers had, and shifting farmers in poor environments had larger language territories than intensive farmers had in rich environments. Among most tribal farmers the documented size of language families—not languages but language families like Indo-European or Uralic—has usually been significantly less than 200,000 km2. Mallory used an average of 250,000–500,000 km2 for Neolithic European language families just to make room on the large end for the many uncertainties involved. Still, that resulted in twenty to forty language families for Neolithic Europe.
The actual number of language families in Europe at 3500 BCE probably was less than this, as the farming economy had been introduced into Neolithic Europe through a series of migrations that began about 6500 BCE. The dynamics of long-distance migration, particularly among pioneer farmers, can lead to the rapid spread of an unusually homogeneous language over an unusually large area for a few centuries (see chapter 6), but then local differentiation should have set in. In Neolithic Europe several distinct migrations flowed from different demographic recruiting pools and went to different places, where they interacted with different Mesolithic forager language groups. This should have produced incipient language differentiation among the immigrant farmers within five hundred to a thousand years, by 6000–5500 BCE. In comparison, the migrations of Bantu-speaking cattle herders across central and southern Africa occurred about two thousand years ago, and Proto-Bantu has diversified since then into more than five hundred modern Bantu languages assigned to nineteen branches, still interspersed today with enclaves belonging to non-Bantu language families. Europe in 3500 BCE, two thousand to three thousand years after the initial farming migrations, probably had at least the linguistic diversity of modern central and southern Africa—hundreds of languages that were descended from the original Neolithic farmers’ speech, interspersed with pre-Neolithic language families of different types. The language of the original migrants to Greece cannot have remained a single language for three thousand years after its speakers were dispersed over many millions of square kilometers and several climate zones. Ethnographic or historic examples of such a large, stable language territory among tribal farmers simply do not exist.
That the speakers of Proto-Indo-European had wagons and a wagon vocabulary cannot be brought into agreement with a dispersal date as early as 6500 BCE. The wagon vocabulary is incompatible with the first-farming/language-dispersal hypothesis. Proto-Indo-European cannot have been spoken in Neolithic Greece and still have existed three thousand years later when wagons were invented. Proto-Indo-European therefore did not spread with the farming economy. Its first dispersal occurred much later, after 4000 BCE, in a European landscape that was already densely occupied by people who probably spoke hundreds of languages.
THE BIRTH AND DEATH OF PROTO-INDO-EUROPEAN
The historically known early Indo-European languages set one chronological limit on Proto-Indo-European, a terminus ante quem, and the reconstructed vocabulary related to wool and wheels sets another limit, a terminus post quem. The latest possible date for Proto-Indo-European can be set at about 2500 BCE (chapter 3). The evidence of the wool and wagon/wheel vocabularies establishes that late Proto-Indo-European was spoken after about 4000–3500 BCE, probably after 3500 BCE. If we include in our definition of Proto-Indo-European the end of the archaic Anatolian-like stage, without a securely documented wheeled-vehicle vocabulary, and the dialects spoken at the beginning of the final dispersal about 2500 BCE, the maximum window extends from about 4500 to about 2500 BCE. This two thousand-year target guides us to a well-defined archaeological era.
Within this time frame the archaeology of the Indo-European homeland is probably consistent with the following sequence, which makes sense also in terms of both traditional branching studies and cladistics. Archaic Proto-Indo-European (partly preserved only in Anatolian) probably was spoken before 4000 BCE; early Proto-Indo-European (partly preserved in Tocharian) was spoken between 4000 and 3500 BCE; and late Proto-Indo-European (the source of Italic and Celtic with the wagon/wheel vocabulary) was spoken about 3500–3000 BCE. Pre-Germanic split away from the western edge of late Proto-Indo-European dialects about 3300 BCE, and Pre-Greek split away about 2500 BCE, probably from a different set of dialects. Pre-Baltic split away from Pre-Slavic and other northwestern dialects about 2500 BCE. Pre-Indo-Iranian developed from a northeastern set of dialects between 2500 and 2200 BCE.
Now that the target is fixed in time, we can solve the old and bitter debate about where Proto-Indo-European was spoken.
CHAPTER FIVE
Language and Place
The Location of the Proto-Indo-European Homeland
The Indo-European homeland is like the Lost Dutchman’s Mine, a legend of the American West, discovered almost everywhere but confirmed nowhere. Anyone who claims to know its real location is thought to be just a little odd—or worse. Indo-European homelands have been identified in India, Pakistan, the Himalayas, the Altai Mountains, Kazakhstan, Russia, Ukraine, the Balkans, Turkey, Armenia, the North Caucasus, Syria/Lebanon, Germany, Scandinavia, the North Pole, and (of course) Atlantis. Some homelands seem to have been advanced just to provide a historical precedent for nationalist or racist claims to privileges and territory. Others are enthusiastically zany. The debate, alternately dryly academic, comically absurd, and brutally political, has continued for almost two hundred years.1
This chapter lays out the linguistic evidence for the location of the Proto-Indo-European homeland. The evidence will take us down a well-worn path to a familiar destination: the grasslands north of the Black and Caspian Seas in what is today Ukraine and southern Russia, also known as the Pontic-Caspian steppes (figure 5.1). Certain scholars, notably Marija Gimbutas and Jim Mallory, have argued persuasively for this homeland for the last thirty years, each usin
g criteria that differ in some significant details but reaching the same end point for many of the same reasons.2 Recent discoveries have strengthened the Pontic-Caspian hypothesis so significantly, in my opinion, that we can reasonably go forward on the assumption that this was the homeland.
PROBLEMS WITH THE CONCEPT OF “THE HOMELAND”
At the start I should acknowledge some fundamental problems. Many of my colleagues believe that it is impossible to identify any homeland for Proto-Indo-European, and the following are their three most serious concerns.
Figure 5.1 The Proto-Indo-European homeland between about 3500–3000 BCE.
Problem #1. Reconstructed Proto-Indo-European is merely a linguistic hypothesis, and hypotheses do not have homelands.
This criticism concerns the “reality” of reconstructed Proto-Indo-European, a subject on which linguists disagree. We should not imagine, some remind us, that reconstructed Proto-Indo-European was ever actually spoken anywhere. R.M.W. Dixon commented that if we cannot have “absolute certainty” about the grammatical type of a reconstructed language, it throws doubt over “every detail of the putative reconstruction.”3 But this is an extreme demand. The only field in which we can find absolute certainty is religion. In all other activities we must be content with the best (meaning both the simplest and the most data-inclusive) interpretation we can advance, given the data as they now stand. After we accept that this is true in all secular inquiries, the question of whether Proto-Indo-European can be thought of as “real” boils down to three sharper criticisms:
Reconstructed Proto-Indo-European is fragmentary (most of the language it represents never will be known).