There is a huge literature devoted to the Indo-European migrations and development of the Indo-European daughter languages, all of which is based ultimately on the reconstruction of the ancestral language, Proto-Indo-European.1080 Because the traditional reconstruction of the Proto-Indo-European phonological system embodies a fundamental mistake—first recognized implicitly in a brilliant article by Hermann Grassmann published in 1863—scholars attempting to draw conclusions about the nature of the protolanguage and the course of its development into the attested daughter languages (i. e., the daughter families, such as Germanic, Italic, Slavic, In-dic, and so on), all of which work depends on historical phonology above all, have in many cases drawn wrong conclusions. Despite Grassmann’s contribution,1081 he was not able to solve the fundamental problem with the reconstruction of Indo-European, mainly because he wrote before the discovery or invention of the phoneme.1082
The problem, as now acknowledged by all Indo-Europeanists, is that the traditional reconstruction of the Proto-Indo-European stop consonants
Has an unvoiced unaspirated series (e. g., *p, *t, *k), a voiced unaspirated series ([*b,] *d, *g), and a voiced aspirated series (*bh, *dh, *gh). When it became generally recognized that this is a typologically unlikely, if not impossible, phonological system that has other significant problems—most importantly, initial *b cannot be reconstructed in this theory of Proto-Indo-European—it was agreed that correction of the reconstruction was necessary. Several attempts have been made to solve the problem, and in fact they are a major topic of Indo-European linguistics—for example, in addition to Szemerenyi’s attempt, Gamkrelidze and Ivanov published a monumental work on the glottalic theory.1083 None of the proposals have worked, however, and none have achieved general acceptance, because they do not actually solve the problem. Although some prominent linguists have accepted Gamkrelidze and Ivanov’s proposal, it not only does not solve the problem at hand, it actually makes it worse.
The solution to the problem1084 is that the traditional three-way opposition of stops is an incorrect reconstruction from the point of view of the phonemic status of the phones involved. It has been known for most of a century, if not longer,1085 that the putative phonemes in question do not occur freely in all positions. Analysis of the accepted constraints shows that the two voiced series ([*b]: *bh, *d : *dh, *g : *gh) occur in complementary distribution and are thus allophones of a unitary voiced phoneme series (*b, *d, *g). They reflect the history of a temporary allophonic distinction that later became phonemic in some of the daughter languages, though in all attested languages that unnatural system has been changed to a natural two-way or four-way opposition of stops. The distinctions therefore can be reconstructed only for a temporary, convergent group consisting of languages that share the characteristic of having a reconstructible three-way opposition in the stops. A three-way system thus cannot be reconstructed to Proto-Indo-European, which had only a two-way phonemic opposition of stops—that is, *p : *b, *t : *d, *k : *g—and no missing *b. Because the other Indo-European daughter languages have either a two-way series in the stops, or a one-way series system (i. e., phonemic *p, *t, *k only) with residual evidence of an earlier two-way system, Proto-Indo-European could only have had a two-way phonemic opposition in its stops.
It is a fairly simple matter to show that, as a result, all known Indo-European languages belong to one of three Sprachbund-like groups, membership in which is determined by the number of categories in the attested or internally reconstructed phonemic stop systems of each daughter language family. Group A, the first-wave languages (with only unvoiced stop phonemes, though there is evidence of the former existence of both unvoiced and voiced stops), consists of Anatolian and Tokharian. Group B, the second-wave languages (with unvoiced, voiced, and voiced aspirate phonemes), consists of Germanic, Italic, Greek, Indic, and Armenian. Group C, the third-wave languages (with unvoiced and voiced stop phonemes), consists of Celtic, Slavic, Baltic, Albanian, and Iranian.1086
It is true that “Anatolian distinguishes inherited voiceless and voiced stops, though admittedly not word initially,”1087 and these distinctions in Anatolian may be reconstructible as such back to the proto language, but the remark “not word initially” is a key point. Tokharian, the other member of Group A along with Anatolian, also has some reflexes that suggest a two-way opposition of stops, but like Anatolian, no word-initial voiced stops. In other words, the two languages, as attested synchronically, do not have a phonemic distinction between voiced and unvoiced stops; the historical distinctions that are preserved word-internally are allophonic. The two daughter families thus belong in the same group from the point of view of both the fundamentally important linguistic phenomenon on which it is based, namely the distribution of the stops, as well as archaeology (on the basis of which both daughter languages appear to have migrated away from the Indo-European homeland around 2000 bc). The distinctions preserved in Anatolian and Tokharian do, however, support the two-way opposition of stops better known from Group C, and thus the reconstruction of the same bipolar system for Proto-Indo-Eu ropean.
Avestan and Vedic: Aspects of a Problem
It is frequently noted that Avestan, which is believed to be the “earliest-attested” form of Iranian, is astonishingly close to Vedic Sanskrit, the “earliest-attested” form of Indic, in phonology, morphology, syntax, and lexicon. In addition to these linguistic features, the contents and religious purposes of the texts in these languages are so remarkably similar in several respects, though of course radically different in overt doctrinal religious content, that based on their evidence it has been possible to reconstruct not only a Proto-Indo-Iranian language but a culture as well. More specifically, it is believed that the Avestan and Vedic texts preserve languages very close to what is thought to be an earlier Proto-Indo-Iranian language representing a stage midway between Proto-Indo-Eu rope an and the Proto-Indic and Proto-Iranian daughter languages. The Avestan and Vedic texts are thus thought to faithfully preserve the languages, and to a great extent the cultures, of the late Proto-Iranian and late Proto-Indic peoples respectively, if not the putative Proto-Indo-Iranians themselves. The texts themselves are now generally believed to have been transmitted orally, with extremely few later intrusions, from about three and a half millennia ago.1088
However, there are several problems with these views. First, the Avestan texts and the Vedic texts are actually attested from less than one thousand years ago.1089 The idea of dating them—as texts—to three or four millennia ago is romantic but hardly supported by much evidence. The usual practice of referring to the Avestan and Vedic texts as the earliest-attested forms of their respective languages is thus a gross distortion. The earliest-attested Avestan manuscripts are actually dated to the thirteenth century ad and are based on an archetype dated to only about three centuries before that.1090 By contrast, the Old Persian language is recorded from the middle of the first millennium bc. Yet, because of the Indo-Iranian theory, above all, Avestan is thought to be a much older form of Iranian, chronologically, than the Old
Persian texts.1091 Because the Avesta are the holy texts of the Zoroastrian religion, they are also believed to preserve references to the putative common Indo-Iranian pantheon and other common Proto-Indo-Iranian beliefs and cultural practices. However, the earliest-attested data on Iranian religious beliefs—including the early Old Persian i nscriptions—contain no reflections of Zoroastrianism per se. The belief system found in the early Avestan texts is not attested until Late Antiquity.
Second, the earliest truly attested forms of Iranian are the North (or “East”) Iranian and South (or “West”) Iranian words and phrases in Assyrian and Greek texts; the first actual texts written in an Iranian language are the Old Persian texts (inscriptions, clay tablets, and seals) of the sixth and fifth centuries Bc. These Iranian languages are very different from Avestan.
Third, Avestan is said to be an East Iranian language, but it has been shown that the now-usual arguments for placing it—and the home of Zoroaster—in Central Asia are chimerical. The apparent “East Iranian” features of Avestan are due to the influence of an East Iranian language during the transmission of the text.1092 The language cannot be placed firmly anywhere in the known Iranian-speaking world, or for that matter anywhen, until the Avestan texts were transcribed into Middle Persian script.
Fourth, it is extremely curious that except for Avestan, phonologically Iranian as a whole—in Old Persian and other early forms, in the Middle Iranian languages, and in the modern Iranian languages—is solidly, unquestionably, a Group C third-wave language, with a clear two-way phonemic opposition in the stops. Only Avestan contains occasional reflexes of a threeway stop system parallel to the system reconstructed on the basis of Vedic Sanskrit, which belongs to the Group B Sprachbund.
Finally, a major problem with Avestan that has so far apparently been overlooked would seem to vitiate, or at least call into serious question, both the traditional view of its linguistic relationship and the theories derived from it. As noted above, it has been remarked, “The Avestan speech is very closely related to Sanskrit,” so astonishingly close, in fact, that “we are able to transpose any word from one language into the other by the application of special phonetic laws.”1093 Avestan’s extensive case system and verbal conjugation system is not just similar to that of Vedic Sanskrit; it is almost identical to it. That is extremely odd. To demonstrate the similarity of the two languages, Indo-Iranian specialists have translated Avestan passages into Vedic Sanskrit, or “Old Indic”—for example, the following Avestan sentence (from Yast 10.6):1094
Avestan Old Indic
Proto-Indo-Iranian English gloss
Avestan Old Indic
Proto-Indo-Ira nian English gloss
Avestan Old Indic
Proto-Indo-Ira nian English gloss
Tdm amanvantdm yazatdm tam amanvantam yajatam *tam amanvantam yajatam This powerful deity,
SuTdm damohu savistam sUram dhamasu savistham *cUTam dhamasu cavistham strong, among the living the strongest,
MiOramyazai zaoOrabyd mitram yajai hotrabhyalt *mitrdm yajai jhautrabhyas Mithra, I honor with libations.
Due to this incredible, unprecedented closeness, Indo-Europeanists believe, “The Indo-Iranian languages clearly derive from an ancestor intermediate between Proto-Indo-Eu ro pe an and the earliest individual Ira nian and Indo-Aryan languages, i. e., one can reconstruct a Proto-Indo-Iranian language.”1095
However, the astounding closeness of Avestan and Vedic Sanskrit, together with the other points noted above, allows—or perhaps, demands—a very different conclusion. Avestan looks less like an Iranian language than like a phonologically Iranized Indic language.1096 The many inexplicable problems of Avestan and the culture thought to be represented in the text of the Avesta can be accounted for as an artifact of Iranians having adopted an oral religious text—clearly a heterodox one by comparison with the Vedas—from an Old Indic dialect. As required of Indic religious practitioners, they memorized it exactly, but in the process, or afterward, it underwent specifically Iranian sound shifts in the mouths of the Iranian-speaking oral reciters. As noted above, Avestan is known exclusively as a literary language of the Zoroastrian religion—it is not known where it was spoken or even if it was ever spoken at all (which seems unlikely)—and it is only attested quite recently.1097 Simple phonological change due to Iranian speakers attempting to preserve an Old Indic dialect text orally over a long period of time would thus explain virtually everything about Avestan. If nevertheless Avestan can still be shown without question to be a genuine Iranian language (which seems unlikely), it would have to constitute an independent sub-branch of its own. If not, and the language is removed entirely from the Iranian family tree, Iranian would then make internal linguistic sense as an Indo-European daughter family. The theory of a Proto-Indo-Iranian language—a striking exception to the otherwise exclusively radial, non-nodal Stammbaum of the Indo-European daughter families (despite the many attempts to construct other models)—would then have to be abandoned, along with much else based on the Indo-Iranian theory. In particular, theories about the culture of the putative Proto-Indo-Iranians and chronological theories concerning the movements of the Proto-Indic and Proto-Iranian peoples would need to be thoroughly revised, but so too would almost everything else in early Indic and Iranian studies.
The Indo-European Creoles
Each of the Indo-European daughter languages—the ancestors of the modern Indo-European languages—retains the bulk of the Indo-European basic lexicon and a significant amount of Indo-European morphology, but it has some local loanwords and, in particular, distinctive phonology. This distribution of features is characteristic of creoles. It must be understood that the term “creole” is not precisely delimited by specific features. It is used for everything from languages containing loanwords—and all known languages contain loanwords—to languages that have undergone major structural changes due to convergence with other languages.
In this book, “creole” is used to refer to languages that have undergone significant changes due to convergence with other languages, but not the kind of radical simplification of structure that is stereotypically said to characterize creoles, the usual (if not the only) example being Haitian Creole, a form of French. As many have noted, modern Indian English—the native speakers1098 of which have full English grammar and lexicon, with a very small number of Indian loanwords—has a phonological structure more akin to Indian languages than to English or other Germanic languages. Although some have claimed that this is a unique artifact of British colonial policies,1099 one must wonder why the same (actually, worse) policies in North America did not produce another creole there. Leaving aside the political aspects involved in such judgments, it is clear that in the former case the English speakers succeeded in imposing their language to some extent but not in eliminating the ruled people, unlike in the latter case. The result of the former was and is a creole. Much the same can be said of other modern Englishes spoken around the world in areas where English is an intrusive language, some with more “creolization” than others.
It is known from observed and recorded modern contact situations that creoles are produced in a very short period of time, not centuries or millennia. Languages are not spoken unchanged over millennia, nor do they take millennia to undergo major changes. That is, the daughter families of Indo-European could not have developed by glacially slow changes over millennia, as the old idea of Indo-European has it, and as most Indo-Europeanists still believe. Modern evidence, as well as modern research on languages undergoing change, shows that the traditional theory is typologically unprecedented and therefore, essentially, impossible. Languages do undergo some internal changes, very slowly, over time, but because these changes can never be isolated from external influences, it cannot even be shown that slow chronological change actually takes place purely on its own without external stimuli.1100 Nevertheless, leaving aside the probable fact of the latter type of change, it is unquestionably the case that major language shifts take place as a result of contact. The Indo-European daughter languages, or branches, are thus to a greater or lesser extent creoles, including even the very earliest recorded Indo-European languages: Hittite, Old Indic, and Mycenaean Greek. This is certainly not unusual. It has been said that “all mature languages are creoles.”1101
What is unusual is the idea that Indo-European, uniquely among the languages of the world, should have preserved its ancestral form (ProtoIndo-European) for thousands of years, then broke up purely via internal chronological change over more thousands of years, and finally developed into the attested daughter languages, all without any creolization. Creoliza-tion is explicitly rejected as a factor in the development of the Indo-European daughter languages1102 despite the fact that the daughter languages are mostly attested first in areas quite distant from the areas where the other daughter languages are first attested, and none of them are attested in the Proto-Indo-European homeland region until after they are attested elsewhere. That means the Indo-European speakers must first have settled in areas where other peoples already lived and mixed with them, producing different creoles of the inherited language, before their languages are first attested.
In addition, the astonishing fact (for the traditional theory) that none of the Indo-European daughter languages were spoken outside the world area where they are first attested cannot be overlooked, as it has been. Early Italic is unknown outside the region of Italy, Greek outside the region of Greece, Tokharian outside the Tokharian region of East Turkistan, and so forth.1103 Moreover, the spatial arrangement of the daughter languages according to isogloss information corresponds to their spatial arrangement geographically—that is, their attested earliest locations.1104 The traditional theory is typologically unprecedented anywhere in the world and does not accord with the evidence.
Each Indo-European daughter language—the protolanguage in turn of the Indo-European daughter families—is thus a creole, the result of immigrant speakers mixing with local people who spoke different languages. The immigrants’ Indo-European language was spoken by their local wives and children with a local accent and some grammatical changes, producing a dialect or creole which was simply an altered local version of the dominant Indo-European language.
The reconstructibility of Proto-Indo-European morphology has been seen as evidence of the incredible conservatism of Indo-European languages by comparison with other languages. But there is considerable evidence against this idea of Indo-European’s incredible—or more accurately, unbelievable—conservatism and slow phonological change over thousands of years: Hittite and the other Anatolian languages. The oft-repeated theory that the Anatolian languages were spoken in Anatolia for thousands of years before they were first recorded is based on the old idea of slow chronological change. Yet Anatolian languages and cultures are so full of local, non-Indo-European elements that it has been difficult to find any vestiges of Indo-Eu ropean religious beliefs and sociopolitical practices among them. How could they have adopted so much from non-Indo-Europeans, but somehow magically preserved a highly archaic “pure” Indo-European language, or an “Indo-Hittite” or “Pre-Proto-Indo-European” language, as some would have it? Because some of the complex morphophonological features reconstructed to Proto-Indo-European have been shown to be restricted to the Group B languages,1105 or should be so restricted—Proto-Indo-European is based largely on the early forms of those very languages (Greek, Latin, Germanic languages, and Sanskrit)—the absence of those features in Anatolian is not surprising. The putative conservativeness of Proto-Indo-European morphophonology is actually evidence for the recentness of the daughter languages’ separation. They may have diverged after the departure of the Proto-Anatolian speakers, but the appearance of the latter in Anatolia still cannot be dated much earlier than the nineteenth century bc. Significant phonological and lexical changes happened at the point in time (within one generation, or at most two generations) of the intrusion of an Indo-European group into areas where the local language was different, or when an Indo-European group was linguistically heavily influenced by a non-Indo-European group, as happened with the formation of Group B. The major structural changes distinguishing each daughter language from each other and from Proto-Indo-European thus did not take centuries to develop. Certainly some changes, once initiated, did take centuries to work themselves out, but that is a different matter. Observation of the way the phonology of daughter languages develop in modern times—Indian English being one of many well-known examples—indicates that creolization, as in the scenario presented here, is the main driving force.1106 The complexity of the changes in Indo-European would seem to be explained by the stages of the migrations, of which there were at least two for most, perhaps all, of what became the Indo-European daughter languages: first from the common homeland to some intermediate place or places (this is clearest in the case of Group B), and then again to the final destinations where the languages are first attested.
The Indo-Europeans, particularly the warrior segment of the population, had an extremely patriarchal, male-dominated society. In many cases, they and their mixed descendants were heavily outnumbered by the original inhabitants and eventually disappeared, leaving only some linguistic residue such as the names of their kings and gods and some other cultural words (as happened in the Mitanni kingdom and elsewhere in the ancient Near East), or even a few short inscriptions (as happened with many languages once spoken in Southern Europe). In other cases the Indo-Europeans imposed their language and maintained it long enough that it could be relatively well recorded. Both scenarios were played out time and again. The most important of these two processes for linguistic history is the second, because it provides sufficient material for careful reconstruction.1107
The relevance for Central Eurasian history is clear. The Indo-Europeans spread from their homeland in Central Eurasia to other parts of Central Eurasia as well as to parts of the Eurasian periphery. They accepted elements of the local cultures with which they mixed and also spread crucial aspects of their own culture. In so doing, they spread the earliest form of the Central Eurasian Culture Complex extensively enough that it survived and became the dominant culture of Central Eurasian peoples in protohistorical and early historic times, as described in the prologue and elsewhere in this book.
Languages in large parts of the world. Indo-European languages are dominant territorially today in every continent except Africa; demographically, the main exceptions are East and Southeast Asia.
World History
