The Horse, the Wheel, and Language: How Bronze-Age Riders From the Eurasian Steppes Shaped the Modern World

by David W. Anthony

  The part that is reconstructed is homogenized, stripped of many of the peculiar sounds of its individual dialects, by the comparative method (although in reconstructed Proto-Indo-European some evidence of dialect survives).

  Proto-Indo-European is not a snapshot of a moment in time but rather is “timeless”: it averages together centuries or even millennia of development. In that sense, it is an accurate picture of no single era in language history.

  These seem to be serious criticisms. But if their effect is to make Proto-Indo-European a mere fantasy, then the English language as presented in the Merriam-Webster Dictionary is a fantasy, too. My dictionary contains the English word ombre (a card game popular in the seventeenth and eighteenth centuries) as well as hard disk (a phrase that first appeared in the 1978 edition). So its vocabulary averages together at least three hundred years of the language. And its phonology, the “proper” pronuciation it describes, is quite restricted. Only one pronunciation is given for hard disk, and it is not the Bostonian hard [haahd]. The English of Merriam-Webster has never been spoken in its entirety by any one person. Nevertheless we all find it useful as a guide to real spoken English. Reconstructed Proto-Indo-European is similar, a dictionary version of a language. It is not, in itself, a real language, but it certainly refers to one. And we should remember that Sumerian cuneiform documents and Egyptian hieroglyphs present exactly the same problems as reconstructed Proto-Indo-European: the written scripts do not clearly indicate every sound, so their phonology is uncertain; they contain only royal or priestly dialects; and they might preserve archaic linguistic forms, like Church Latin. They are not, in themselves, real languages; they only refer to real languages. Reconstructed Proto-Indo-European is not so different from cuneiform Sumerian.

  If Proto-Indo-European is like a dictionary, then it cannot be “timeless.” A dictionary is easily dated by its most recent entries. A dictionary containing the term hard disk is dated after 1978 in just the way that the wagon terminology in Proto-Indo-European dates it to a time after about 4000–3500 BCE. It is more dangerous to use negative information as a dating tool, since many words that really existed in Proto-Indo-European will never be reconstructed, but it is at least interesting that Proto-Indo-European does not contain roots for items like spoke, iron, cotton, chariot, glass, or coffee—things that were invented after the evolution and dispersal of the daughter languages, or, in the metaphor we are using, after the dictionary was printed.

  Of course, the dictionary of reconstructed Proto-Indo-European is much more tattered than my copy of Merriam-Webster’s. Many pages have been torn out, and those that survive are obscured by the passage of time. The problem of the missing pages bothers some linguists the most. A reconstructed proto-language can seem a disappointing skeleton with a lot of bones missing and the placement of others debated between experts. The complete language the skeleton once supported certainly is a theoretical construct. So is the flesh-and-blood image of any dinosaur. Nevertheless, like the paleontologist, I am happy to have even a fragmentary skeleton. I think of Proto-Indo-European as a partial grammar and a partial set of pronunciation rules attached to the abundant fragments of a very ancient dictionary. To some linguists, that might not add up to a “real” language. But to an archaeologist it is more valuable than a roomful of potsherds.

  Problem #2. The entire concept of “reconstructed Proto-Indo-European” is a fantasy: the similarities between the Indo-European languages could just as well have come about by gradual convergence over thousands of years between languages that had very different origins.

  This is a more radical criticism then the first one. It proposes that the comparative method is a rigged game that automatically produces a proto-language as its outcome. The comparative method is said to ignore the linguistic changes that result from inter-language borrowing and convergence. Gradual convergence between originally diverse tongues, these scholars claim, might have produced the similarities between the Indo-European languages.4 If this were true or even probable there would indeed be no reason to pursue a single parent of the Indo-European languages. But the Russian linguist who inspired this line of questioning, Nikolai S. Trubetzkoy, worked in the 1930s before linguists really had the tools to investigate his startling suggestion.

  Since then, quite a few linguists have taken up the problem of convergence between languages. They have greatly increased our understanding of how convergence happens and what its linguistic effects are. Although they disagree strongly with one another on some subjects, all recent studies of convergence accept that the Indo-European languages owe their essential similarities to descent from a common ancestral language, and not to convergence.5 Of course, some convergence has occurred between neighboring Indo-European languages—it is not a question of all or nothing—but specialists agree that the basic structures that define the Indo-European language family can only be explained by common descent from a mother tongue.

  There are three reasons for this unanimity. First, the Indo-European languages are the most thoroughly studied languages in the world—simply put, we know a lot about them. Second, linguists know of no language where bundled similarities of the kinds seen among the Indo-European languages have come about through borrowing or convergence between languages that were originally distinct. And, finally, the features known to typify creole languages—languages that are the product of convergence between two or more originally distinct languages—are not seen among the Indo-European languages. Creole languages are characterized by greatly reduced noun and pronoun inflections (no case or even single/plural markings); the use of pre-verbal particles to replace verb tenses (“we bin get” for “we got”); the general absence of tense, gender, and person inflections in verbs; a severely reduced set of prepositions; and the use of repeated forms to intensify adverbs and adjectives. In each of these features Proto-Indo-European was the opposite of a typical creole. It is not possible to classify Proto-Indo-European as a creole by any of the standards normally applied to creole languages.6

  Nor do the Indo-European daughter languages display the telltale signs of creoles. This means that the Indo-European vocabularies and grammars replaced competing languages rather than creolizing with them. Of course, some back-and-forth borrowing occurred—it always does in cases of language contact—but superficial borrowing and creolization are very different things. Convergence simply cannot explain the similarities between the Indo-European languages. If we discard the mother tongue, we are left with no explanation for the regular correspondences in sound, morphology, and meaning that define the Indo-European language family.

  Problem #3. Even if there was a homeland where Proto-Indo-European was spoken, you cannot use the reconstructed vocabulary to find it because the reconstructed vocabulary is full of anachronisms that never existed in Proto-Indo-European.

  This criticism, like the last one, reflects concerns about recent inter-language borrowing, focused here on just the vocabulary. Of course, many borrowed words are known to have spread through the Indo-European daughter languages long after the period of the proto-language—recent examples are coffee (borrowed from Arabic through Turkish) and tobacco (from Carib). The words for these items sound alike and have the same meanings in the different Indo-European languages, but few linguists would mistake them for ancient inherited words. Their phonetics are non—Indo-European, and their forms in the daughter branches do not represent what would be expected from inherited roots.7 Terms like coffee are not a significant source of contamination.

  Historical linguists do not ignore borrowing between languages. An understanding of borrowing is essential. For example, subtle inconsistencies embedded within German, Greek, Celtic, and other languages, including such fleeting sounds as the word-initial [kn-] (knob) can be identified as phonetically uncharacteristic of Indo-European. These fragments from extinct non—Indo-European languages are preserved only because they were borrowed. They can help us create maps of pre—Indo-European place-names, like the place
s ending with [-ssos] or [-nthos] (Corinthos, Knossos, Parnassos), borrowed into Greek and thought to show the geographic distribution of the pre-Greek language(s) of the Aegean and western Anatolia. Borrowed non—Indo-European sounds also were used to reconstruct some aspects of the long-extinct non—Indo-European languages of northern and eastern Europe. All that is left of these tongues is an occasional word or sound in the Indo-European languages that replaced them. Yet we can still identify their fragments in words borrowed thousands of years ago.8

  Another regular use of borrowing is the study of “areal” features like Sprachbunds. A Sprachbund is a region where several different languages are spoken interchangeably in different situations, leading to their extensive borrowing of features. The most famous Sprachbund is in southeastern Europe, where Albanian, Bulgarian, Serbo-Croat, and Greek share many features, with Greek as the dominant element, probably because of its association with the Greek Orthodox Church. Finally, borrowing is an ever-present factor in any study of “genetic” relatedness. Whenever a linguist tries to decide whether cognate terms in two daughter languages are inherited from a common source, one alternative that must be excluded is that one language borrowed the term from the other. Many of the methods of comparative linguistics depend on the accurate identification of borrowed words, sounds, and morphologies.

  When a root of similar sound and similar meaning shows up in widely separated Indo-European languages (including an ancient language), and phonological comparison of its forms yields a single ancestral root, that root term can be assigned with some confidence to the Proto-Indo-European vocabulary. No single reconstructed root should be used as the basis for an elaborate theory about Proto-Indo-European culture, but we do not need to work with single roots; we have clusters of terms with related meanings. At least fifteen hundred unique Proto-Indo-European roots have been reconstructed, and many of these unique roots appear in multiple reconstructed Proto-Indo-European words, so the total count of reconstructed Proto-Indo-European terms is much greater than fifteen hundred. Borrowing is a specific problem that affects specific reconstructed roots, but it does not cancel the usefulness of a reconstructed vocabulary containing thousands of terms.

  The Proto-Indo-European homeland is not a racist myth or a purely theoretical fantasy. A real language lies behind reconstructed Proto-Indo-European, just as a real language lies behind any dictionary. And that language is a guide to the thoughts, concerns, and material culture of real people who lived in a definite region between about 4500 and 2500 BCE. But where was that region?

  FINDING THE HOMELAND: ECOLOGY AND ENVIRONMENT

  Regardless of where they ended up, most investigators of the Indo-European problem all started out the same way. The first step is to identify roots in the reconstructed Proto-Indo-European vocabulary referring to animal and plant species or technologies that existed only in certain places at particular times. The vocabulary itself should point to a homeland, at least within broad limits. For example, imagine that you were asked to identify the home of a group of people based only on the knowledge that a linguist had recorded these words in their normal daily speech:

  You could identify them fairly confidently as residents of the American southwest, probably during the late nineteenth or early twentieth centuries (six-gun and the absence of words for trucks, cars, and highways are the best chronological indicators). They probably were cowboys—or pretending to be. Looking closer, the combination of armadillo, sagebrush, and cactus would place them in west Texas, New Mexico, or Arizona.

  Linguists have long tried to find animal or plant names in the reconstructed Proto-Indo-European vocabulary referring to species that lived in just one part of the world. The reconstructed Proto-Indo-European term for salmon, *lók*s, was once famous as definite proof that the “Aryan” homeland lay in northern Europe. But animal and tree names seem to narrow and broaden in meaning easily. They are even reused and recycled when people move to a new environment, as English colonists used robin for a bird in the Americas that was a different species from the robin of England. The most specific meaning most linguists would now feel comfortable ascribing to the reconstructed term *lók*s- is “trout-like fish.” There are fish like that in the rivers across much of northern Eurasia, including the rivers flowing into the Black and Caspian Seas. The reconstructed Proto-Indo-European root for beech has a similar history. Because the copper beech, Fagus silvatica, did not grow east of Poland, the Proto-Indo-European root *bháo- was once used to support a northern or western European homeland. But in some Indo-European languages the same root refers to other tree species (oak or elder), and in any case the common beech (Fagus orientalis) grows also in the Caucasus, so its original meaning is unclear. Most linguists at least agree that the fauna and flora designated by the reconstructed vocabulary are temperate-zone types (birch, otter, beaver, lynx, bear, horse), not Mediterranean (no cypress, olive, or laurel) and not tropical (no monkey, elephant, palm, or papyrus). The roots for horse and bee are most helpful.

  Bee and honey are very strong reconstructions based on cognates in most Indo-European languages. A derivative of the term for honey, *medhu-, was also used for an intoxicating drink, mead, that probably played a prominent role in Proto-Indo-European rituals. Honeybees were not native east of the Ural Mountains, in Siberia, because the hardwood trees (lime and oak, particularly) that wild honeybees prefer as nesting sites were rare or absent east of the Urals. If bees and honey did not exist in Siberia, the homeland could not have been there. That removes all of Siberia and much of northeastern Eurasia from contention, including the Central Asian steppes of Kazakhstan. The horse, *ek*wo-, is solidly reconstructed and seems also to have been a potent symbol of divine power for the speakers of Proto-Indo-European. Although horses lived in small, isolated pockets throughout prehistoric Europe, the Caucasus, and Anatolia between 4500 and 2500 BCE, they were rare or absent in the Near East, Iran, and the Indian subcontinent. They were numerous and economically important only in the Eurasian steppes. The term for horse removes the Near East, Iran, and the Indian subcontinent from serious contention, and encourages us to look closely at the Eurasian steppes. This leaves temperate Europe, including the steppes west of the Urals, and the temperate parts of Anatolia and the Caucasus Mountains.9

  FINDING THE HOMELAND: THE ECONOMIC AND SOCIAL SETTING

  The speakers of Proto-Indo-European were farmers and stockbreeders: we can reconstruct words for bull, cow, ox, ram, ewe, lamb, pig, and piglet. They had many terms for milk and dairy foods, including sour milk, whey, and curds. When they led their cattle and sheep out to the field they walked with a faithful dog. They knew how to shear wool, which they used to weave textiles (probably on a horizontal band loom). They tilled the earth (or they knew people who did) with a scratch-plow, or ard, which was pulled by oxen wearing a yoke. There are terms for grain and chaff, and perhaps for furrow. They turned their grain into flour by grinding it with a hand pestle, and cooked their food in clay pots (the root is actually for cauldron, but that word in English has been narrowed to refer to a metal cooking vessel). They divided their possessions into two categories: movables and immovables; and the root for movable wealth (*peku-, the ancestor of such English words as pecuniary) became the term for herds in general.10 Finally, they were not averse to increasing their herds at their neighbors’ expense, as we can reconstruct verbs that meant “to drive cattle,” used in Celtic, Italic, and Indo-Iranian with the sense of cattle raiding or “rustling.”

  What was social life like? The speakers of Proto-Indo-European lived in a world of tribal politics and social groups united through kinship and marriage. They lived in households (*dómha), containing one or more families (*génh1es-) organized into clans (*wei-), which were led by clan leaders, or chiefs (*weik-potis). They had no word for city. Households appear to have been male-centered. Judging from the reconstructed kin terms, the important named kin were predominantly on the father’s side, which suggests patrilocal marriages (brides moved into the
husband’s household). A group identity above the level of the clan was probably tribe (*h4erós), a root that developed into Aryan in the Indo-Iranian branch.11

  The most famous definition of the basic divisions in Proto-Indo-European society was the tripartite scheme of Georges Dumézil, who suggested that there was a fundamental three-part division between the ritual specialist or priest, the warrior, and the ordinary herder/cultivator. Colors might have been associated with these three roles: white for the priest, red for the warrior, and black or blue for the herder/cultivator; and each role might have been assigned a specific type of ritual/legal death: strangulation for the priest, cutting/stabbing for the warrior, and drowning for the herder/cultivator. A variety of other legal and ritual distinctions seem to have applied to these three identities. It is unlikely that Dumézil’s three divisions were groups with a limited membership. Probably they were something much less defined, like three age grades through which all males were expected to pass—perhaps herders (young), warriors (older), and lineage elders/ritual leaders (oldest), as among the Maasai in east Africa. The warrior category was regarded with considerable ambivalence, often represented in myth by a figure who alternated between a protector and a berserk murderer who killed his own father (Hercules, Indra, Thor). Poets occupied another respected social category. Spoken words, whether poems or oaths, were thought to have tremendous power. The poet’s praise was a mortal’s only hope for immortality.

  The speakers of Proto-Indo-European were tribal farmers and stockbreeders. Societies like this lived across much of Europe, Anatolia, and the Caucasus Mountains after 6000 BCE. But regions where hunting and gathering economies persisted until after 2500 BCE are eliminated as possible homelands, because Proto-Indo-European was a dead language by 2500 BCE. The northern temperate forests of Europe and Siberia are excluded by this stockbreeders-before-2500 BCE rule, which cuts away one more piece of the map. The Kazakh steppes east of the Ural Mountains are excluded as well. In fact, this rule, combined with the exclusion of tropical regions and the presence of honeybees, makes a homeland anywhere east of the Ural Mountains unlikely.