A social prehistory of European languages.

Robb, John

Consistent to most views of Indo-European in later European prehistory is a genetic focus. The blanket of related languages across Europe marks an equal human spread -- whether of steppe warriors, Beaker burialists or slashing-and-burning farmers. What if the languages are reconstructed using other premisses than this 'genealogical' view?

Introduction

All language prehistories are hypothetical models largely constructed from their premisses. In this paper I will present a prehistory of the European languages based on rather different premisses from those usually used, with the goal of exploring some unexamined linkages between prehistoric society and language. The present moment is especially propitious for such an effort. Within the last decade, several relatively sophisticated models of Indo-European languages have been proposed, relating them to social phenomena such as trade and elite dominance and to the ecological and demographic consequences of the spread of agriculture (Renfrew 1987; Ehret 1988; Sherratt & Sherratt 1988; Zvelebil & Zvelebil 1988). In a similar vein, I will outline a hypothetical prehistory of European languages, as their evolution may have been shaped by successive waves of social change. While the results are both speculative and over-generalized, they may serve to stimulate our archaeological imagination by viewing old questions from new angles.

Language history involves studying two complementary aspects of language, genetic origins and sociolinguistic processes. In effect, the former traces the raw material presented to each generation of speakers in terms of its historical derivation, while the latter studies the processes of selection and modification which reshape and transmit this material to the next generation. In historical linguistics, the only work to consider Indo-European languages other than genetically seems to be that of Trubetzkoy (1939). In archaeology, students of Indo-European have focused almost exclusively on attempting to trace the historical continuity of Indo-European as a particular group of language lineages; this includes traditional archaeologists as well as bioanthropological work (Cavalli-Sforza et al. 1993) and recent archaeological work (Renfrew 1987; 1992) applying processual social models to genetic linguistic questions. In this paper I will follow an opposite approach, and ignore the question of where modern European languages came from in favour of the question of what happened to them en route. At the present moment, this approach affords two advantages. Much important linguistic change happens through ways other than the simple lineal transmission of a group's language to its offspring groups. Broadening the topic thus may make linguistic models more accurate and more relevant to processual social models. Secondly, the genetic model also carries certain theoretical and methodological baggage. For instance, it is difficult to use it prehistorically without also arguing that language, ethnic group and material culture coincide reliably enough for the latter to serve as an archaeological key to the former over long spans of time; this argument tends to commit the prehistorian to excluding social processes in which these phenomena follow different paths (a fact largely responsible for the conspicuous absence of language and ethnicity in New Archaeological treatments of European prehistory). It is liberating to consider language prehistory without the burden of tracing a specific lineage, simply as the prehistory of European languages. Reconceptualizing the 'Indo-European problem'

How can language patterns such as the Indo-European expansion be conceptualized in social terms? The pervasive influence of the genetic method of comparative linguistics must be recognized immediately. The genetic method compares languages to determine their structural similarities, which it attributes to a documented or hypothetical common ancestor. It can thus only produce scenarios of original unity followed by subsequent divergence (cf. Renfrew 1987; Thomason & Kaufman 1984; Terrell 1988). Moreover, since we know progressively less about earlier linguistic landscapes, it is not surprising that we imagine modern languages branching from isolated, single ancestors. These facts, combined with archaeological assumptions about the unity of material culture, linguistic community and ethnic identity, are largely responsible for the traditional narrative of large-scale folk migrations outward from an Indo-European 'homeland'.

While it is true that all languages and cultures have historical antecedents, sociolinguistic processes can affect the relations between the two in many ways. Linguistic variability gives speakers resources for expressing solidarity and distinction along many dimensions. Speakers continually choose among alternative forms of speech, on scales from the inflection of a vowel to the adoption of a primary linguistic identity, and linguistic change results from the shifting balance of these choices. Language as strategic choice thus responds not only to ethnicity but to political, economic and cultural factors as well. Vectors of language replacement, for instance, can include mass migration, migration of limited groups (as in elite dominance models), migration of individuals not organized into groups (as through intermarriage) and shifting language choices among multilingual individuals for a number of reasons (cf. Ehret 1988). Languages in contact also influence each other through the formation of pidgins and creoles, the spread of trade or gender-specific languages, and borrowing and convergence among genetically distinct tongues. As Thomason & Kaufman (1988) point out, the effects of language contact can be dramatic. Over historical periods of time, these forces can result in language replacement with or without migrations, in the formation of new languages of uncertain ancestry and in the formation of regionally convergent linguistic systems, Sprachbund (Hoch 1991: 494).

On the micro-scale of tracing individual language histories, these factors complicate the task considerably. Without large-scale political organization, causal factors underlying sociolinguistic choices in one region may not be important elsewhere. Moreover, speech communities, self-identified ethnic groups, political units and material culture assemblages may not coincide, even when ethnic identity, language and material style all serve as boundary markers. As Hill (1978) points out, the 'dialect-tribe' model is far from universal; perhaps the limiting case of divergence between co-residence and language affiliation was Bara society in the Upper Orinoco basin, where exogamous clans traditionally spoke different languages, resulting normally in marriages between native speakers of different languages (Jackson 1983). However, the same factors open up the possibility of analysis on the macro-scale. By treating the linguistic map as an array of individual cases which respond to economic, social and political circumstances, we may be able to formulate hypotheses about how general characteristics of the map might change in response to these circumstances even when individual cases are indeterminate. For this reason, at the present state of knowledge, it is probably more useful to formulate the problem as the prehistory of European languages rather than as the 'Indo-European problem'.

Social trends affect the linguistic map ultimately through language extinctions and production, and I suggest that the balance between the two can provide a good general gauge of their effects. Normal social processes result in both language extinction and creation, creating language 'turnover' over time. One effect of this is that, at any point, the linguistic landscape will be populated with a few relatively large language families, created through social processes which are undirected on the macro-scale (Robb 1991). A second implication is that the rate of language turnover due to social processes will change according to the nature of the social processes behind it. When social processes remain relatively constant, individual language families may rise and fall stochastically, but the overall form of the linguistic map is likely to stay the same. Social evolution on the continental scale, in contrast, is likely to change both the fates of individual families and basic features of the map itself -- its patchiness, grain and time-depth. Over the course of social evolution, then, changes in language ecology -- the fit between language and its social environment -- may be observable through trends in language demography. Historical parameters for a hypothetical prehistory of European languages

Both the spatial and temporal constraints of European language prehistory are vague. Spatially, in spite of a century of research, the regions where Indo-European languages were spoken 5000 years ago are still mostly a matter of conjecture. Chronological aspects are uncertain for several reasons. The time-depth of reconstructed proto-ancestors depends, somewhat artificially, upon how much we know about them, and historic accounts of Indo-European expansions often do not specify the kind of language spoken indigenously. For instance, it was assumed that Greek entered Greece early in the 1st millennium until the decipherment of Linear B demonstrated that a Greek language was spoken there earlier. Moreover, the best-known case, Indo-European, is atypical; probably the great majority of European language lineages experienced quite different historical trends ending in extinction or survival of only a few languages. What historical evidence exists, however, gives an idea of the general form of European language prehistory. The Indo-European (IE) languages were dispersed widely before the 2nd millennium, as IE languages are historically known in the early and mid 2nd millennium in the Aegean, southwest Asia and south Asia. On the other hand, their penetration is unlikely to have been very thorough before this date, as non-IE languages are known from much later dates in Italy, Spain, the Baltic and Russian frontier of Indo-European, the Caucasus and Anatolia. (Documented non-IE languages include Etruscan, Basque, Iberian, Tartessian, Estonian, Finnish, Urartian, Sumerian, Hurrian, Hattic and Mitannian. Languages which may also have been non-IE include Pictish, Lepontic and Ligurian.) The location of non-IE languages around the fringes of Europe may reflect the contours of an expansion (Mallory 1989). However, these areas also either have some early historic information available (i.e. the southern and eastern fringes) or may have been relatively unaffected by historical linguistic imperialism and hence have preserved linguistic isolates to a later date (i.e. the northeastern fringe). We really do not know to what extent Europe north of the Alps harboured non-IE islands similar to Basque or Etruscan in late prehistory. The general historic trend suggested by the IE languages is of a widespread but patchy early distribution, followed by subsequent stages of both 'filling in' and expansion at the periphery. More typical lineages would have followed the converse pattern. At several millennia before the historic record begins, therefore, the map may have included a patchy mixture of languages; the intervening millennia would have seen the rise and consolidation of one group of languages and the decline of others.

The Palaeolithic/Mesolithic

Palaeolithic and Mesolithic hunter-gatherers lived at population densities much lower than later groups, and networks of nomadic groups occupied large territories; for instance, Bogucki (1988: 43) postulates only 14 Mesolithic 'social territories' in Europe north of the Alps. Moreover, contact with other groups would have been vital for information on resources (Gamble 1982; Whallon 1989) and as a source of mates (Wobst 1974), and hunter-gatherers living at low densities typically maintain fluid relations with neighbouring populations as ecological insurance against local shortfalls. These considerations suggest that languages and language families would have been spoken over relatively wide, flexibly bounded territories. For example, while horticulturists in temperate Native North American spoke dozens of languages, Inuit populations from Greenland to Siberia spoke 5-10 dialects of only two language families, Eskimo and Aleut (Woodbury 1984). Similarly, all Australian languages south of the northern coast belong to one language family. A second characteristic of populations living at low population densities is that they often do not conform to the model of one tribe-one dialect (Hill 1978). Instead, a variety of linguistic features allow linkages among groups. Languages may be differentiated through easily-learned lexical items while sharing a common phonological system, multilingualism is widespread, and linguistic exogamy is also known. Hunter-gatherer languages sometimes also form dialect continuums, instead of sharply delimited linguistic zones, as in the North American Boreal Forest and Arctic (Rhodes & Todd 1981; Woodbury 1984). Krauss & Golla (1981: 68-9) note: Attempts to classify the Athapaskan languages into historically meaningful subgroups have not met with success . . . the principal difficulty arises from the fact that Athapaskan linguistic relationships, especially in the Subarctic area, cannot be adequately described in terms of discrete family tree branches. This is because intergroup communication has ordinarily been constant, as no Northern Athapaskan language or dialect was ever completely isolated from the others for long . . .

Whatever the language boundaries, the network of communication in the Northern Athapaskan dialect complex is open-ended. It is probably worth noting that, even in 1980, perhaps most Northern Athapaskans live with only other Athapaskan speakers as neighbors and rarely hear a language that is not Athapaskan. People from adjacent communities usually expect to be able to understand one another's speech, if not immediately then surely after some practice. Local dialects and languages are important as symbols of social identity; but the native expectation these differences, even across relatively vast differences, will not be barriers to communication gives the Northern Athapaskan speaker a distinctively open and flexible perception of his social world. As this suggests, the Palaeolithic language map would probably have featured few language families, each spoken over a wide region. Pleistocene populations may also have suffered periodic 'bottlenecks' due to glaciations or other climatic changes which would further reduce cultural and linguistic diversity. Over the 25,000-30,000 years of the Upper Palaeolithic, the linguistic scene is likely to have undergone cyclical change; individual languages and language families would have expanded and contracted while the overall distribution remained more or less constant. Through the accumulation of lineage extinctions involved in such linguistic turnover, one or another language spoken in the Early Upper Palaeolithic would eventually become the ancestor of all modern European languages, fissioning to found a growing language family (Robb 1991). The total number of languages spoken at any given point, however, would probably fluctuate around an average tied primarily to population density in that particular millennium. In the Mesolithic, with incipient sedentism and intensive exploitation of more localized resources in some areas, this scenario would begin to evolve towards the broad changes described below. Earlier Neolithic: sedentism, population growth and ethnogenesis

Broad homogeneous horizons of Early Neolithic material culture (Impressed Ware/LBK) represent the widespread, rapid diffusion of a Neolithic way of life, if not actual migrations. Renfrew (1987) has argued that this wave of diffusion carried with it not only a new subsistence economy but the Indo-European languages as well. While specific aspects of this position have been criticized (Zvelebil & Zvelebil 1988; Mallory 1989), Renfrew's use of Ammerman & Cavalli-Sforza's (1984) 'wave of advance' model accurately ties language change to the demographic effects of neolithization. Whether expanding Neolithic populations carried their languages into new territories or acculturating Mesolithic populations retained theirs, sedentism and population growth would have affected language distributions in complex ways.

Agriculture allows a greatly increased population density at the cost of reduced mobility. Four direct effects would have been critical for language. For the first time, local groups would have been demographically self-sufficient. Subsistence farming communities could also have satisfied most or all of their material needs within a few dozen kilometres' range of home. While long-distance trade is always evident archaeologically to some degree, inter-group communication would have dropped and become narrower in scope, and long-distance trade may have been mediated through a long series of immediate neighbours. Simultaneously, group solidarity would have become focused sharply within groups with a stable, concretely bounded membership and a specific territorial base. These conditions -- the formation of small, self-sufficient territorial groups in a landscape of rising population density -- would have favoured a wave of ethnogenesis and probably of language formation as well. This is suggested both by communicative considerations and by North American linguistic geography. As a communicative signal, the choice of a primary language identity, like cranial deformation, is relatively inflexible, apparent in virtually all contexts of social life and difficult for adult outsiders to acquire or feign. These 'design features' lend it well to the purpose of symbolizing identities which are general to many contexts, life-long and exclusive to individuals sharing a common territory and history; identities tied to specific situations, contexts or within-group distinctions might be symbolized by more flexible material media or smaller-scale linguistic choices. In the context of neolithization, language may have been 'activated' as a stylistic marker, resulting in differentiation among small tribal groups distinguishing and identifying themselves by their local dialects (as Laycock (1982) argues, the dynamics of language change are different in languages with few speakers than in widely spoken language, and the primary motivation underlying Malanesia's famous linguistic diverity is to use language as a badge of social identity -- even to the point that 'the maximum number of persons in Papua New Guinea speaking exactly the same language is about 500' (1982: 34).). On the macro-scale, the relation between sedentism, population density and linguistic diversity is well illustrated by the distribution of native North American languages (Voegelin & Voegelin 1966), in which the size of language family territories varies inversely with latitude from Arctic and Boreal forest hunter-gatherers to temperate and sub-tropical horticulturists; in place of the broad zones of mutual intelligibility characterizing northern mobile populations, European explorers often reported that woodland horticultural groups spoke a different language every 50 km (Crawford 1975a; 1978). The Northwest Coast and coastal California may represent an extreme situation of high local population densities and linguistic fragmentation, with, in the Northwest Coast, some linguistic features associated with economic intensification and trade (shared linguistic features among genetically distinct languages, development of a trade jargon) (Thompson & Kinkade 1991; Hill 1978).

Over several millennia of Neolithic life, a given Mesolithic language may have had numerous offspring, and the resulting landscape might look like those of New Guinea or the Amazon basin, with patchworks of languages of different families spoken by a few thousand persons each. The result would be a dramatic increase in the number of languages spoken compared to the Palaeolithic and Mesolithic. The proliferation of new languages probably kept language family extinctions to a minimum, except for those actually supplanted by languages associated with an intrusive Neolithic, and many currently recognizable language families may have been 'founded' by ramification from Mesolithic ancestors in this period. The Later Neolithic-Bronze Age: intensification, prestige economies and language consolidation

The Late Neolithic, Copper and Bronze Ages have provided the second and traditionally more popular candidate for Indo-European migrations. Common archaeological features include wide stylistic horizons, economic intensification, the rise of prestige systems in political life and of an associated iconography of weaponry, and possible changes in kinship. These appear to have taken place within relatively acephalous societies. Interpretation of these changes as evidence of expanding patriarchal Indo-European hordes (Gimbutas 1973; 1980) is far unanimous. Other archaeologists have ascribed them to the rise of stratified societies (Gilman 1981; 1991), the rise of a prestige-oriented society (Shennan 1982), economic intensification (Sherratt 1981) and the consequences of a male-oriented gender ideology (Robb 1992).

Feil (1987) discusses the effects of economic intensification of language groups in Highland Papua New Guinea. Traditional groups in the Eastern Highlands lived in small, isolated communities practising mixed subsistence agriculture; languages among them were typically spoken by a few thousand individuals. In contrast, in the Western Highlands and particularly around Mount Hagen, culturally similar groups cultivated sweet potatoes intensively in order to feed pigs for elaborate exchange networks. Communities as a whole were larger (although this larger group size would have been difficult to discern archaeologically, as the relief from endemic warfare in the exchange-oriented Western Highlands allowed populations to live in small dispersed settlements, more convenient for pig-raising, instead of in the nucleated, often fortified villages typical of the Eastern Highlands) (Feil 1987). Dialects, languages and language families were significantly larger, and some dialects were spoken by up to almost 70,000 people. The reason for this greater size was clear. It was politically advantageous to incorporate more producers and exchange partners into a larger system, and the resulting success in ceremonial exchange reinforced the ideological ties binding groups together. The increased number of speakers of Western Highlands languages reflects both larger overall group size and more extensive within-group ties based upon production and exchange. As this suggests, as regional exchange systems supported by secondary products developed, fewer languages would have been spoken in the same population or territory. One consequence of the secondary products revolution (Sherratt 1981) would thus be the gradual consolidation of fewer, larger language groups. Two collateral TABULAR DATA OMITTED effects would tend to magnify this difference over time. As Thorpe & Richards (1984) point out, societies with prestige-oriented economies tend to be more aggressively expansive than those regulated by pervasive ritual authority. Secondly, once differentials in intensification and population density existed, larger language groups would be stochastically less likely to become extinct than smaller, less-intensified groups, and would be likely to exert political and economic forces over them and to assimilate their speakers if they became defunct.

Beside the general effects of intensification and prestige economies, several special types of linguistic interaction are relevant to this period.

Regional exchange languages

Both increased long-distance trade and colonization of interstitial areas such as highlands would result in more contact across linguistic frontiers. Multi-lingualism is common in such situations, and where a number of groups habitually interact, one language may be recognized as a trade language. In the southeastern United States (Crawford 1975a; 1978), for example, French and English observers were often impressed by the bewildering variety of tribal languages they found there. Groups communicated through bilingual individuals scattered among villages. In several regions, however, single languages such as Creek exerted a centripetal attraction; typically such a language belonged to a group which was centrally located and powerful, had important trade connections or led a regional confederation. In these cases, the language took on added political functions and became a common means of communication among speakers of different languages. As a further development, in one case, such a central language spawned a pidgin. Mobilian trade jargon probably originated before European contact as a Chickasaw-based pidgin used in coastal and riverine trade on the Gulf Coast (Crawford 1975a; 1978; Haas 1975; Dreschel 1977; 1981). Trade languages were common throughout North America, including a Virginia Algonquian language used by the Powhatan confederacy, a pidginized form of Delaware, Occaneeche, Peoria, Ojibway, Cree, Dakota, Comanche, Navaho, Hupa, and a pidginized version of Chinook (Goddard 1971; Dreschel 1981; Taylor 1981; Rhodes 1982; Thomason 1983). Some of these may represent post-contact developments for trade with Europeans; Delaware pidgin was developed among a Delaware-led confederation trading with the Dutch and Swedes (Goddard 1971; Dreschel 1981), while the use of Ojibway and Cree as lingua francas was sponsored by the French and English fur trades respectively (Rhodes 1982). Others, such as the trade jargon used by the Chinook to communicate with trade partners up and down the coast and with captives incorporated as slaves, probably developed indigenously (Thomason 1983; cf. Dutton 1978; 1983 for indigenously developed New Guinea trade pidgins associated with particular trade networks).

For European prehistory, regional exchange languages may have had several significant effects. When a regional language is present, languages are often status-ranked, with bilingualism becoming one-way (e.g. native speakers of low-status languages learn regional high-status languages but native speakers of the high-status language do not learn other languages) (Rhodes 1982). In times of demographic or political change, a language used for regional communication would thus stand a good chance of supplanting other languages during group fission or fusion, spreading it through the conversion and incorporation of speakers (Rhodes 1982: 2, 8):

Most of the speakers of Ottawa that I have been working with are descendants not of Ottawa but of Potawatomis and Chippewa, i.e. American Ojibwas. . . . we have no guarantees that any of the people we work with in the 20th century are speaking the same language as their ancestors, because a language shift can come about, not through a relatively visible mass assimilation like the Mancouten (Goddard 1978: 670), but, as in the case of the Chippewa and Potawatomi speakers of Ottawa in Michigan and Ontario, by silently slipping into the trade language. Specific trade networks would provide vectors for expansion, as Sherratt & Sherratt (1988) have suggested for Indo-European. Finally, in such situations, linguistic accommodation in syntactic and phonological features among a region's languages may lead to the formation of a Sprachbund in which common morphological and phonological features spread among genetically distinct, converging languages. In grammatical simplification during contact, features held in common are likely to be retained preferentially, and the resulting convergence may muddle genetic reconstructions of language relatedness (Thomason & Kaufman 1988).

Prestige languages

As Helms (1983) argues, one of the bases of elite power is exotic knowledge, and this may include knowledge of exotic languages, particularly as these would be associated with the ability to trade, to conduct politics and to form external contracts. An elite language may be introduced by invasion or intermarriage of a small group at the pinnacle of society, as in the 'elite dominance' model (Renfrew 1987; 1992). It may also be created or acquired by a native group, whether definitively dominant or not, as a restricted marker of their status and special abilities. Prestige languages may also be used for ritual speech within a regional belief system, and in many cases, prestige languages are the internal face of external trade languages.

Archaeologically, prestige languages may have been associated with the visible rise of regional horizons of prestige goods during the Later Neolithic, Bronze and Copper Ages. As Shennan (1986) notes, such changes may have spread among neighbouring polities by chain reaction. Linguistically, inter-group prestige chaining may have resulted in rapid spreads of languages, prestigious linguistic features or of the vocabulary of status (such as the IE cognate for 'king' (Latin rex, Celtic rix, Vedic rg), sometimes cited as evidence of archaic Indo-European social structure, cf. Sherratt & Sherratt 1988: note 12; and Benveniste 1974).

Gender-associated languages

Males and females in most societies speak in different ways about different things, and variation seems to derive from the cultural connotations of gender. In the most extreme case of gendered languages recorded, that of the Island Caribs shortly after European contact, males used a recognized, distinct sub-language among themselves which women were forbidden to use when speaking among themselves (Taylor & Hoff 1980; Davis & Goodwin 1990). Similar gender-associated languages are known in other groups such as the Yanomama (Hill 1978), the Chukchi and even within Sumerian society (Diakonoff 1976). Sociolinguistic studies of English (Trudgill 1972) have associated levelled gender distinctions in society with reduced speech differences, and gender divergence would presumably also be reflected linguistically. This is probably in fact the case with Island Carib; the men's language probably originated as an Arawak-based trade pidgin with heavy influence from mainland Carib languages before being adopted as a gender marker (Farr 1993).

Archaeologically, the 3rd and 2nd millennia see the rise of weapon symbolism, of hunting art and ploughing symbolism which may be linked to a heightened male gender ideology (Robb 1992). Such an ideology may well have been expressed in speech in phonological and paralinguistic differences, if not through more extensive linguistic features. Moreover, given that men may have established their claim to prestige and power through warfare, exchange and inter-group politics, these three situations would almost certainly overlap. Knowledge of an inter-group exchange language, for instance, may have both established one as elite within the group and allowed gender-specific distinctions (for example Salisbury 1962).

In 3rd- and 2nd-millennium Europe, intensification and prestige systems are well-documented, and trade, prestige and gender-associated languages are likely to have occurred in some areas. As Sherratt & Sherratt (1988) suggest, these developments probably helped spread the Indo-European languages. Bronze Age sociolinguistic developments would have intensified in the later Bronze and Iron Age, as their social causes -- economic intensification, increasing polity size, developing gender and social stratification -- accelerated throughout Europe. Increasingly active and mercantile trade along coasts and rivers would encourage the formation of trade languages, and control of exotic or esoteric languages may have formed a necessary part of an aristocrat's claim to social position in stratified societies such as those modelled by Frankenstein & Rowlands (1972). Among these varied, localized trends, several continent-wide trends would have been established. First, the size of speaking units would have generally increased, resulting in fewer overall language communities and in relatively greater linguistic stability for economically intensified groups than for unintensified ones. Secondly, increased communication among groups, combined with increased differentiation within them, would have begun to cut across linguistic isolation, establishing regional groups united by the use of common exchange networks, prestige symbols, gender ideologies, and possibly trade languages or high-status languages. In the long run, such bridging would have encouraged linguistic homogeneity within a region, both by favouring partial or complete language replacement and by motivating lexical and grammatical borrowing among languages. The overall effect on language distributions would have been to reduce diversity via extinctions, and to consolidate and expand the survivors. This would have taken place across the linguistic map; the Indo-European languages would have been only one among a number of families developed and spread by these processes.

The 1st millennium: historical processes

Language extinction is an epidemic effect of state organization and civilization (Robins & Uhlenbeck 1991), and this has probably been so from the start. European societies on the expanding margins of civilizations were affected in many ways (Rowlands et al. 1987). By the 3rd millennium Mesopotamian civilizations had established trading outposts in the Levant, by the mid 2nd millennium Levantine and Mycenaean traders ventured throughout the eastern and central Mediterranean, and by the end of the 1st millennium the entire Mediterranean and most of Europe north of the Alps had either been incorporated within the expanding civilized core or been deeply affected by interactions around its periphery.

Among their direct effects, empires forcibly incorporated other societies and moved individuals and populations long distances, often mixing speakers in situations such as slavery in which the imperial language or another lingua franca such as Aramaic (Paper 1982) was the only common tongue. Empires spread religious systems, literacy and associated prestige languages over vast areas. Civilizations exported manufactured goods and exotic luxuries, and archaeological horizons such as the Italian orientalizing period testify to the striking effects produced in societies marked by internal prestige competition. In cultural terms, empires often simplify the language map by promoting a single native language for trade or conversion, as an adjunct to a homogenized category of 'native'. Quechua, for instance, was spread by Spanish missionaries to Andean populations who never adopted it from its original speakers, the Incas (Heath & Laprade 1982). In political terms, the development of centralized multi-ethnic polities with developed bureaucracies would have been a key threshold for linguistic extinctions. In the Roman Empire, for instance, with the exception of Britain (in which romanization was confined to an urban elite), Latin replaced native languages in regions in which Romans created large-scale administrative and industrial structures (Italy, Gaul, Spain, Dacia); it failed to take hold where such structures already existed and the linguistic changes they generated had already occurred (Greece, Asia Minor, Palestine, Egypt, North Africa). As ethnohistoric examples demonstrate, the indirect effects of contact with civilization may precede and carry far beyond actual zones of colonization (Wolf 1982; Gailey & Patterson 1988). For instance, native groups trading with the core, such as Celtic aristocrats in France (Crumley 1974; Nash 1987; Haselgrove 1987) would also have experienced further the effects of economic intensification and political stratification, as well as transmitting them to their own hinterlands via long-range trade to procure slaves, metals and other commodities. Roman intervention also both allowed pro-Roman groups such as the Aedui to dominate their neighbours, resulting in the formation of fewer but more powerful native groups on the periphery (Haselgrove 1987: 112), and provoked the formation of anti-Roman alliances of unprecedented size. The formation of large tribal confederacies such as the Samnites and expansionism such as the protohistoric movement of Britons into Britain may reflect such turmoil around the periphery of the civilized world.

As the history of almost every European country demonstrates, these effects combined to reduce linguistic diversity dramatically. In Italy, where the ethnolinguistic record is more complete than in most of Europe, inscriptional evidence of the 6th-5th centuries BC attests to the presence of about 12-15 languages belonging to 5-10 branches of at least four major language families (Celtic, Italic, Greek and Etruscan) (de Voto 1978; Pulgram 1958; 1978). By the end of the first century BC only two languages are attested, Latin and Greek, though others may have continued to be spoken for some time. Linguistic change may have taken place in two distinct phases, though evidence is especially scanty for earlier periods. From early in the 1st millennium, Italian societies formed a periphery influenced by trade with Greek and Near Eastern civilizations. The only directly documentable linguistic effect of this is the spread of writing systems and Greek loan words into Etruscan, Latin and Italic dialects (de Voto 1978: 57). However, it is likely that the advent of expansionist and mercantile class-stratified urban societies bore linguistic consequences such as the elimination of smaller groups via their assimilation into larger ones. The second phase corresponds to the political and military incorporation of various Italian societies within the civilized core. As Rome came to dominate Italy, it did not follow a policy of forcible linguistic assimilation, and native tongues such as Oscan and Umbrian are attested by inscriptions dating to several centuries after Roman annexation. However, as the official language of a centralized empire, Latin was the key to administrative power, to trade and to cultural prestige. It would also have replaced local languages for everyday use in Rome's Italian colonies, in multi-ethnic cities and large farms increasingly run by slave labour, in military service and in the political life of increasingly enfranchised subject populations. While other native Italian tongues were treated with indifference by the Romans, they became increasingly consigned to the role of basilect in a diglossic situation and eventually died out, leaving at most some regional imprint on later Italian dialects.

Thus historical processes would have both spread languages from the core outward and caused a wave of language extinctions within and around the margins of civilization. Languages and language families would have been directly eradicated by political or cultural take-overs as well as wiped out in the faster and larger-scale social flux around the periphery. They would have vanished as their speakers were absorbed into more resilient native groups, or would have been replaced by state languages and state-sponsored native languages. Ironically, the linguistic distribution reported by Classical observers may have evolved only shortly before they arrived, and partly as a result of the same processes which brought them there.

Conclusions: implications for Indo-European studies

There are really two distinct trends in European language prehistory. In the first, genetic trend most known European languages can be related historically within a common language family, Indo-European. In the second trend hypothesized here and deriving from sociolinguistic models, after a post-Palaeolithic wave of language formation, from the Later Neolithic through the present there has probably been an accelerating reduction in linguistic diversity throughout Eurasia. On the macro-scale, this reflects two key factors, population density and subsistence, and political organizations of increasing scale. In this reconstruction of Eurasian language prehistory, relatively few language groups occupied large, loosely bounded territories during the Palaeolithic and Mesolithic. Neolithic sedentism and population growth brought a wave of ethnogenesis in a social context within which language would have been a prominent identity marker for new, well-bounded groups tied to small territories. From the Later Neolithic onwards, language extinctions outstripped replacement; this was linked to the rise of economic intensification, gender and rank stratification and inextricable regional economic and political relationships. Historical processes emanating from the first civilizations accelerated this wave of language extinctions greatly. This long-term trend is the primary reason why only a few language families are spread over vast distances in Eurasia. That one should be Indo-European rather than another family seems mostly due to historical accident. An argument could perhaps be made that Indo-European's location in Old World political geography, along the northern periphery of expanding civilizations and later partially within this core, was the critical factor. Indo-European thus followed an expansive trajectory similar to the Semitic languages, in contrast to more marginal groups such as Finno-Ugric.

While this reconstruction is speculative, it suggests several points for further consideration.

Beginning from the premiss that language, social process and material culture interact dynamically has both theoretical costs and benefits. Taking this premiss seriously may undermine our ability to pursue specific genetic questions beyond relatively recent prehistory; it may be simply indeterminable whether a given IE lineage pre-dates the 2nd-3rd millennium in a given area, where the IE homeland was, and so on. In return, considering prehistoric languages as formed by sociolinguistic processes may allow us to explain language distributions using other models as well as mass migration and based upon general characteristics of the archaeological record rather than strict ceramic-ethnic correlations. It also allows us to form interesting hypotheses about language--society interactions in prehistory. For instance, broad archaeological horizons at the beginning and the end of the Neolithic probably both represent periods of linguistic change, but of very different kinds. Two further points concern archaeological and historical sources of evidence on ancient languages. If the reconstruction here is correct, the ethnolinguistic record known from Classical observers and generally late inscriptional evidence should be treated with all the caution due to any ethnohistory. For instance, it seems unlikely that only one language, Gaulish, would have been spoken in a territory the size of France peopled by small decentralized societies. The ethnolinguistic record probably reflects both the lack of linguistic interest of Classical observers and historical processes affecting linguistic patterns and coinciding with their arrival. The late prehistoric record, if we could observe it, would probably show much greater diversity within groups like Celtic, and possibly many more islands of non-IE languages. Finally, language has always been read into the archaeological record through proxy concepts such as ethnicity. Somewhat counter-intuitively, using social processes as a linking concept may allow more articulation between linguistic process and archaeological evidence rather than less. Substantial work has been done on material culture as communicative media, and language may be integrated in these models. For instance, it may be possible to verify trends such as ethnogenesis or prestige-related intensification or to understand when communities will draw linguistic boundaries tightly or relax them, as well as to predict concurrent or divergent social trends in other, archaeologically visible media such as ceramics. Hypotheses about linguistic behaviour and social processes, even if not entirely verifiable, can thus contribute towards a complete model of prehistoric societies.

Acknowledgements. I am grateful to Starr Farr and Alex Barker for discussion of the Island Carib and Mobilian linguistic situations, to E.W. Robb for general discussion of linguistic matters, and to Geoff Emberling, Lynn Fisher and C. Loring Brace for critical discussion of the manuscript. All sins of omission and commission remain my own work.

References

AMMERMAN, A. & L. CAVALLI-SFORZA. 1984. The Neolithic transition and the genetics of populations in Europe. Princeton (NJ): Princeton University Press.

BENVENISTE, E. 1974. Indo-European language and society. London: Faber & Faber. BOGUCKI, P. 1988. Forest farmers and stockherders: early agriculture and its consequences in north-central Europe. Cambridge: Cambridge University Press.

CAVALLI-SFORZA, L., A. PIAZZA & P. MENOZZI. 1993. Demic expansions and human evolution, Science 259: 639-46.

COOPER, R. (ed.). 1982. Language spread: studies in diffusion and social change. Bloomington (IN): Indiana University Press.

CRAWFORD, J. 1975a. Southeastern Indian languages, in Crawford (1975b): 1-120.

(Ed.). 1975b. Studies in southeastern Indian languages. Athens (GA): University of Georgia Press.

1978. The Mobilian trade language. Knoxville (TN): University of Tennessee Press.

CRUMLEY, C. 1974. Celtic social structure. Ann Arbor (MI): Museum of Anthropology. Anthropological Paper 54.

DAVIS, D. & R.C. GOODWIN. 1990. Island Carib origins: evidence and nonevidence, American Antiquity 55: 37-48.

DE VOTO, G. 1978. The languages of Italy. Chicago (IL): University of Chicago Press.

DIAKONOFF, I.M. 1976. Ancient writing and ancient written language: pitfalls and peculiarities in the study of Sumerian, in S.J. Lieberman (ed.), Sumerological studies in honor of Thorkild Jacobsen: 99-121. Chicago (IL): University of Chicago Press.

DRESCHEL, E. 1977. Metacommunicative functions of Mobilian Jargon, an American Indian pidgin of the lower Mississippi River region, in G. Gilbert (ed.), Pidgin and creole languages: 433-44. Honolulu (HI): University of Hawaii Press. 1981. A preliminary sociolinguistic comparison of four indigenous pidgin languages of North America with notes towards a sociolinguistic typology in American Indian linguistics, Anthropological Linguistics 23: 93-112.

DUTTON, T. 1978. Language and trade in central and southeast Papua, Mankind 11: 341-53.

1983. Birds of a feather: a pair of rare pidgins from the Gulf of Papua, in E. Woolford & W. Washabaugh (ed.), The social context of creolization: 77-105. Ann Arbor (MI): Karoma.

EHRET, C. 1988. Language change and the material correlates of language and ethnic shift, Antiquity 62: 564-73.

FARR, S. 1993. Gender and ethnogenesis in the early colonial Lesser Antilles. Paper prepared for the 15th Congress of the International Association for Caribbean Archeology, San Juan, Puerto Rico.

FEIL, D. 1987. The evolution of Highland Papua New Guinea societies. New York (NY): Cambridge University Press.

FRANKENSTEIN, S. & M. ROWLANDS. 1972. The internal structure and regional context of early Iron Age society in southwest Germany, Bulletin of the Institute of Archaeology, University of London 15: 73-112.

GAILEY, C. & T. PATTERSON. 1988. State formation and uneven development, in J. Gledhill, B. Bender & M. Larsen (ed.), State and society: 77-90. London: Unwin Hyman.

GAMBLE, C. 1982. Interaction and alliance in Palaeolithic society, Man 17: 92-107.

GILMAN, A. 1981. The development of social stratification in Bronze Age Europe, Current Anthropology 22: 1-24.

1991. Trajectories towards social complexity in the later prehistory of the Mediterranean, in T. Earle (ed.), Chiefdoms: power, economy, ideology: 146-68. New York (NY): Cambridge University Press.

GIMBUTAS, M. 1973. The first wave of Eurasian steppe pastoralists into Copper Age Europe, Journal of Indo-European Studies 5: 277-338.

1980. The Kurgan Wave 2 (c. 3400-3200 BC) into Europe and the following transformation of culture, Journal of Indo-European Studies 8: 273-315.

GODDARD, I. 1971. The ethnohistorical implications of early Delaware linguistic materials, Man in the Northeast 1: 14-26.

HAAS, M. 1975. What is Mobilian?, in Crawford (1975b): 257-65.

HASELGROVE, C. 1987. Culture process on the periphery: Belgic Gaul and Rome during the late Republic and early Empire, in Rowlands et al. (1987): 104-24.

HEATH, S.B. & R. LAPRADE. 1982. Castilian colonization and indigenous languages: the cases of Quechua and Aymara, in Cooper (1982): 118-47.

HELM, J. (ed.). 1981. Handbook of North American Indians 6: Subarctic. Washington (DC): Smithsonian Institution Press.

HELMS, M. 1983. Ulysses' sail. Princeton (NJ): Princeton University Press.

HILL, J. 1978. Language contact systems and human adaptations, Journal of Anthropological Research 34: 1-26.

HOCH, H. 1991. Principles of historical linguistics. 2nd edition. New York (NY): Mouton de Gruyter.

HOLM, J. 1988. Pidgins and creoles. Cambridge: Cambridge University Press.

JACKSON, J. 1983. The fish people: linguistic exogamy and Tukanoan identity in norhtwest Amazonia. New York (NY): Cambridge University Press.

KRAUSS, M. & V. GOLLA. 1981. Northern Athapaskan languages, in Helm (1981): 67-85.

LAYCOCK, D. 1982. Melanesian linguistic diversity: a Melanesian choice? in R.J. May & H. Nelson (ed.), Melanesia: beyond diversity 1: 33-8. Canberra: Research School of Pacific Studies, Australian National University.

MALLORY, J. 1989. In search of the Indo-Europeans. London: Thames & Hudson.

NASH, D. 1987. Imperial expansion under the Roman Republic, in Rowlands et al. (1987): 87-103.

PAPER, H. 1982. Language spread: the ancient Near Eastern world, in Cooper (1982): 107-17.

PULGRAM, E. 1958. The tongues of Italy. Cambridge (MA): Harvard University Press. 1978. Italic, Latin, Italian: 600 BC to AD 1260. Heidelberg: Carl Winter Universitatsverlag.

RENFREW, C. 1987. Archaeology and language: the puzzle of Indo-European origins. London: Cape.

1992. Archaeology, genetics and linguistic diversity, Man 27: 445-78.

RHODES, R. 1982. Algonquian trade languages, in Papers of the 13th Algonquian Conference: 1-10. Ottawa: Carleton University Press.

RHODES, R. & E. TODD. 1981. Subarctic Algonquian languages, in Helm (1981): 52-66.

ROBB, J. 1991. Random causes with directed results: the Indo-European spread and the stochastic loss of lineages, Antiquity 65: 287-91.

1992. Gender ideology and social evolution in prehistoric Italy. Paper presented at the Society for American Archeology meetings, Pittsburgh.

ROBINS, R. & E. UHLENBECK (ed.). 1991. Endangered languages. New York (NY): St Martin's Press.

ROWLANDS, M., M. LARSEN & K. KRISTIANSEN. 1987. Centre and periphery in the ancient world. New York (NY): Cambridge University Press.

SALISBURY, R. 1962. Notes on bilingualism and linguistic change in New Guinea, Anthropological Linguistics 4(7): 1-13.

SHENNAN, S. 1982. Ideology, change and the European Bronze Age, in I. Hodder (ed.), Symbolic and structural archaeology: 155-61. New York (NY): Cambridge University Press.

1986. Interaction and change in third millennium BC western and central Europe, in J. Cherry & C. Renfrew (ed.), Peer polity interaction: 137-48. New York (NY): Cambridge University Press.

SHERRATT, A. 1981. Plough and pastoralism: aspects of the secondary products revolution, in I. Hodder, G. Isaac & N. Hammond (ed.), Pattern of the past: 261-305. Cambridge: Cambridge University Press.

SHERRATT, A. & S. SHERRATT. 1988. The archaeology of Indo-European: an alternative view, Antiquity 62: 584-95.

TAYLOR, A. 1981. Indian lingua francas, in C. Ferguson & S. Heath (ed.), Language in the USA: 175-95. Cambridge: Cambridge University Press.

TAYLOR, D. & B. HOFF. 1980. The linguistic repertory of the Island Carib in the seventeenth century: the men's language -- a Carib pidgin?, International Journal of Linguistics 46: 301-12.

TERRELL, J. 1988. History as a family tree, history as an entangled bank: constructing images and interpretations of prehistory in the South Pacific, Antiquity 62: 642-57.

THOMASON, S. 1983. Chinook jargon in areal and linguistic context, Language 59: 820-70.

THOMASON, S. & T. KAUFMAN. 1988. Language contact, creolization and genetic linguistics. Berkeley (CA): University of California Press.

THOMPSON, L. & M.D. KINKADE. 1990. Languages, in W. Suttles (ed.), Handbook of North American Indians 7: Northwest coast: 30-51. Washington (DC): Smithsonian Institution Press.

THORPE, I. & C. RICHARDS. 1984. The decline of ritual authority and the introduction of beakers into Britain, in R. Bradley & J. Gardiner (ed.), Neolithic studies: 67-78. Oxford: British Archaeological Reports. British series 133.

TRUBETZKOY, N. 1939. Gedanken uber das Indogermanenproblem, Acta Linguistica Copenhagen 1: 81-9.

TRUDGILL, P. 1972. Sex, covert prestige, and linguistic change in the urban British English of Norwich, Language in Society 1: 179-95.

VOEGELIN, C. & F. VOEGELIN. 1966. Map of North American Indian languages. Seattle (WA): University of Washington Press.

WHALLON, R. 1989. Elements of cultural change in the Upper Palaeolithic, in P. Mellars & C. Stringer (ed.), The human revolution: 433-54. Edinburgh: Edinburgh University Press.

WOBST, H.M. 1974. Boundary conditions for Paleolithic social systems: a simulation approach, American Antiquity 39: 147-78.

WOLF, E. 1982. Europe and the people without history. Berkeley (CA): University of California Press.

WOODBURY, A. 1984. Eskimo and Aleut languages, in D. Damas (ed.), Handbook of North American Indians 5: Arctic: 49-63. Washington (DC): Smithsonian Institution Press.

ZVELEBIL, M. & L. ZVELEBIL. 1988. Agricultural transition and Indo-European dispersals, Antiquity 62: 574-83.