Possible correlations between the histories of the major language families and major traditions within the archaeological record have exercised the minds of scholars since Gustav Kossinna, and Gordon Childe attempted early in the twentieth century to trace the archaeological record of the Indo-European languages. But long before the rise of archaeology as a research discipline, some of the major language families had already come into historical perspective through comparative linguistic research. This perspective is often claimed to have emerged when Sir William Jones in 1786 suggested that Greek, Sanskrit, Latin, Gothic, Celtic, and Old Persian were ’sprung from some common source.’ (Jones 1993).
In the twenty-first century, language history and the archaeological record can be studied in combination to recover history at two major (but clearly overlapping) levels:
(a) at the level of the individual language, ethnoling-uistic group, or historical community; and,
(b) at the level of the language family or major subgroup.
It is also possible to seek linguistic correlations for some archaeological complexes, particularly those which are sharply bounded and defined by consistent stylistic features, although this tends to become more difficult as the complex in question extends further back into prehistory and becomes more diffusely defined. Such correlations, for obvious reasons, also benefit from the assistance of written and translatable texts.
In general, it is a very difficult task to trace the identity into deep levels of prehistory of a specific ethnolinguistic or historical population (e.g. Celts, Greeks, Etruscans), unless one is dealing with a very isolated region or an island where one can assume there has been no substantial population replacement during the period in question. A good example of the latter would be certain Pacific islands, for example Easter Island or New Zealand, both fairly isolated since their first human settlements by Polynesians (Kirch and Green 1987). However, this entry is not primarily concerned with such society- or culture-specific correlations amongst language, history, and archaeology, but focuses instead on the study of languages as members of genetically constituted and evolving families, combined with the study of large-scale archaeological traditions as they spread, evolve, and interact through time and space.
Historical reconstruction at this level tends to be organized such that language families (e.g. Indo-European, Austronesian) are foregrounded as the major foci of enquiry, rather than archaeological complexes. This is because language families are usually more sharply defined and reveal much clearer patterns of genetic inheritance than do archaeological complexes. In such situations, archaeology tends to be used to support or refute historical linguistic questions (for instance, where was the Indo-European homeland located, what lifestyle did its inhabitants enjoy, and when?). However, some archaeological complexes of particularly wide distribution, internal homogeneity, and short time span (e.g., the Linearbandkeramik (LBK) early Neolithic of Central Europe, the Lapita cultural complex of the western Pacific) are also sometimes foregrounded as requiring a paleolinguistic identity. For instance, does the LBK correlate with Indo-European dispersal into Central Europe; does Lapita correlate with Austronesian dispersal through Melanesia into Polynesia? In order to understand how the data of historical linguistics and archaeology might be compared against each other in order to improve understanding of the human past, it is first necessary to state clearly the abilities and limitations of the two disciplines.
1. Language as a Source of Information on Human Prehistory
The branch of linguistics which is of most interest to prehistoric archaeologists is that known as comparative historical linguistics, in which the structures and vocabularies of present-day or historically recorded languages are compared in order to identify families, and subgroups within these families. The methodology of comparative linguistic reconstruction is precise. Like the methodology of cladistics, as applied in biology, its main goal is to identify shared innovations which can identify language subgroups. Such subgroups comprise languages which have shared a common ancestry, apart from other languages with which they are more distantly related. Languages which comprise a subgroup share descent from a common ‘protolanguage,’ this being in many cases a chain of related dialects. The protolanguages of subgroups within a family can sometimes be organized into a family tree of successive linguistic differentiations (not always sharp splits, unlike real tree branches), and for some families it is possible to postulate a relative chronological order of subgroup formation. For instance, many linguists believe that the separation between the Anatolian languages (including Hittite) and the rest of Indo-European represents the first identifiable differentiation in the history of that family. Likewise the separation between the Formosan (Taiwan) languages and the rest of Austronesian (Malayo-Polynesian) represents the first identifiable differentiation within Austronesian.
The vocabularies of reconstructed protolanguages (e.g., Proto-Indo-European, Proto-Austronesian) can sometimes provide remarkable details on the locations and lifestyles of ancient ancestral communities, with many hundreds of ancestral terms and their associated meanings reconstructible in some instances. There is also a linguistic technique known as glottochronology which attempts to date protolanguages by comparing recorded languages in terms of shared cognate (commonly inherited) vocabulary, applying a rate of change calculated from the histories of Latin and the Romance languages. But the rate of change varies with sociolinguistic situation, often a complete unknown in prehistoric situations. Glottochronology can be used only for recent millennia and for those languages which have not undergone intense borrowing from languages in other unrelated families. It is not a guaranteed route to chronological accuracy.
The other major source of linguistic variation, apart from modification through descent, is that termed by linguists ‘borrowing’ or ‘contact-induced change.’ This operates between different languages, and often between languages in completely unrelated families. Borrowing, if identified at the protolanguage level, can be as much an indicator of the homeland of a language family or subgroup as can genetic structure. It can also reflect important contact events in language history. (See Phylogeny and Systematics.)
2. Archaeology as a Source of Information on Human Prehistory
Archaeology is concerned mainly with the recovery and interpretation of the material remains of the human past, and the environmental contexts in which those remains were originally deposited. Such remains can be dated, and grouped into regional complexes of related components. Such complexes can then be compared with other complexes, and the natures of the boundaries between such complexes can be studied carefully. Some are sharply bounded, hence possible candidates for correlation with an ethnolinguistic group, others are simply nodes of relative homogeneity in a kaleidoscope of ever-shifting patterning. Archaeology alone cannot pinpoint ethnicity, unless of course it operates in an environment associated with literacy and the availability of written records (and even then ambiguity can plague interpretation, as in the modern debate in the UK archaeological literature about the definition and archaeological history of the Celts). Any correlations between the archaeological and linguistic records will always require care—prehistoric artifacts cannot talk!
3. How Can Language Family History and Archaeological Prehistory be Correlated?
Because languages change constantly through time, and because relationships between languages become ever fainter as we go back in time, it is assumed by most linguists that language family histories apply only to the past 8,000 to 10,000 years. At a greater timescale we enter the arena of’macrofamilies’ such as Nostratic and Amerind, concepts which cause rather vituperous debate amongst linguists because of their very ambiguity and elusiveness. Most of the examples discussed below relate to historical trends which have occurred since the beginnings of agriculture and which do not extend back as far as the macrofamily level.
Correlation of the archaeological and linguistic records is not always a simple matter because the two classes of data are conceptually quite discrete. However, correlations can be made when language family distributions correspond with the distributions of delineated archaeological complexes, particularly when the material culture and environmental vocabulary reconstructed at the protolanguage level for a given family correspond with material culture and its environmental correlates as derived from the archaeological record. Many reconstructed protolanguages, for instance, have vocabularies which cover crucial categories such as agriculture, domestic animals, pottery, and metallurgy, these all being identifiable in the archaeological record.
The concept of the language family is more sturdy than that of the archaeological culture. This is important, because linguists have come to a remarkable level of agreement on the classification of the world’s language families. Apart from a small number of Creoles, mostly a result of European colonization and population translocation, the vast majority of the world’s language families are clearly bounded in a classificatory sense and not beset with huge numbers of ‘mixed’ languages. As an example, the Indian subcontinent has been a region of interaction between the speakers of Indo-European and Dravidian languages for at least three millennia. Languages within these two families have borrowed extensively from each other, to the extent that the subcontinent is often referred to by linguists as a ‘linguistic area’ (a zone of widespread areal diffusion). Nevertheless, the subcontinent is not covered by languages which are half Indo-European, half Dravidian, chaotically mixed. This means that language families are coherent entities, capable of maintaining coherence and independence through long periods of time. As such, they are believed to carry traceable records of history and to be associated, in their origins, with homeland regions and processes of population dispersal. Where different language families meet, we can infer that different populations have met as well. Although some populations have changed their languages in the past, it is unlikely that language shift, as opposed to an actual dispersal of ancestral speakers of a protolanguage, can be the main mode of dispersal of a major language family. All of this means that the origins and dispersals of the protolanguages from which language families are created should correlate with major population movements, frequently on a scale which should be visible in the archaeological record.
4. Some Examples of Language Family Origin and Dispersal Histories with Claimed Archaeological Correlations
Some of the major language families of the Old World are shown in Fig. 1 (the American families are much more mosaic-like in distribution and cannot be mapped so easily). Also shown in Fig. 2 are regions where agriculture developed independently. Many archaeologists and linguists today recognize that many language families could owe their initial creations to population dispersal as a result of population growth following on from the development of agriculture. If this is so, then the homelands of these families can be expected to overlap with the regions of early agriculture, as indeed seems to be the case for the Middle East, China, and Mesoamerica. However, it is important to remember that many language families are associated totally with hunting and gathering populations, and presumably always have been, so their histories obviously will not involve this factor. Such hunter-gatherer families include Khoisan in southern Africa, the Australian languages (probably several families), Athabaskan and Eskimo-Aleut, and the languages of western North America and southern South America. Other families, such as Uralic, Algon-kian, and Uto-Aztecan, have both agricultural and hunter-gatherer populations. In some of these cases it is possible that former agricultural peoples have actually become hunters and gatherers in difficult environments (e.g., the Great Basin Uto-Aztecans).
It is also apparent that some languages and subgroups (but not whole language families) have been recorded as spreading over large distances in historical times, under conditions of statehood, religious evangelism, and colonialism. Thai, the Chinese languages, Arabic, and of course English and Spanish all come to mind here. In the twenty-first century, it is also apparent that lingua francas and national languages can spread rapidly as a result of educational policy, literacy, mass media, and sociolinguistic status; but it is more difficult to imagine such processes of language adoption as being of great significance amongst the small-scale societies of preurban prehistory. Nevertheless, many prehistorians have suggested that language replacement processes of this type, whereby people adopt a language deemed to be of high social status and abandon their original language, have been instrumental in the spread of some families. One such family is Indo-European, for which some linguists and archaeologists have long agreed on a homeland in the steppes north of the Black Sea, followed by a spread into Europe by Late Neolithic and Bronze Age pastoral peoples with domesticated horses and wheeled transport. According to the archaeologist Marija Gimbutas, these people undertook their migrations into Europe between 4500 and 2500 BC, dominating and absorbing the older Neolithic societies in the process.
This view of Bronze Age conquest and language replacement for Indo-European dispersal has been challenged by the archaeologist Colin Renfrew (1987, 1996), who opts instead for an association of early Indo-European with early Neolithic farming dispersal into Europe from Turkey. Nowadays, the idea that many of the major agriculturalist language families spread as a result of the early development of agriculture is taking firmer hold. Farmers typically have larger populations than hunter-gatherers, and if farmers are not enclosed by other farming populations i.e., if they live in a region surrounded essentially by low-density hunter-gatherers, then expansion is a likely outcome, exactly as in the European frontiers in Australia and western North America. So, while the agricultural dispersal hypothesis is no more ‘provable’ than any other hypotheses to explain language family origins, it does at least have the strong supporting factor of a historically proven mechanism which can allow and encourage population expansion to occur. Such expansion need not mean extinction of all hunter-gatherers. In many ethnographic situations, hunter-gatherers have survived in the interstices of agricultural or pastoralist landscapes, perhaps for millennia.
Put simply, the farming dispersal hypothesis would see the protolanguages for Indo-European, Semitic, Turkic, Sumerian, Elamite, and possibly Dravidian located in the wheat, barley, cattle, and caprine zone in the Middle East, with dispersals occurring mainly in the period between 6500 and 3000 BC. During this time mixed farming became widely established and the archaeological record tells us unambiguously that population was increasing in an overall sense quite rapidly (despite periodic environmental setbacks and short-term population retractions). Sino-Tibetan, Au-stroasiatic, Austronesian, Tai, and Hmong-Mien would all have begun their dispersal from the region of rice and millet cultivation in China, focused in the middle and lower Yellow and Yangzi valleys, between 5000 and 2000 BC (with Austronesians eventually colonizing the greater part of the Pacific). Niger-Congo (including Bantu) resulted from the development of agriculture in West Africa and the Sahel zone, mainly after 3000 BC, and perhaps following earlier pastoralist dispersals in northeastern Africa by Afroasiatic (Berber, Chadic, Cushitic) and Nilo-Saharan speakers. In the Americas, the Mayan, Otomanguean, Mixe-Zoque, Uto-Aztecan, and Chib-chan language families probably spread as a result of agricultural developments in Greater Mesoamerica after 3500 BC. In South America the picture is a little more diffuse, but some of the major Andean and Amazonian families might have spread as a result of the establishment of maize and manioc agriculture after about 2500 BC—examples here would include Quechua and Aymara, and lowland Amazonian families such as Arawak, Carib, and Tupi.
Archaeologically, these suggested language radiations associated with early agricultural societies should be reflected in the distributions of some very widespread archaeological complexes. In particular, it has been noted in many regions that the archaeological complexes of early agricultural phases are much more widespread and homogeneous in content than the highly regionalized complexes of later periods. This appears to be the case in early Neolithic Europe, East Asia, and the Pacific, and amongst the Early Formative cultures of the Americas.
5. Language Contact and Cultural Contact
Language and archaeology correlations can be sought not only for the origins and dispersal histories of language families, but can also reflect the contacts which take place from time to time between languages in different families and subgroups. For instance, the Austronesian speakers of New Guinea have been in intense contact for upwards of 2,000 years with the speakers of Papuan languages in several unrelated families. This has led to a great deal of contact-induced change and even language shift, and it is therefore not surprising to discover that distributions of material culture often cross-cut language boundaries. The archaeological record, however, suggests that quite strong differences in material culture would have distinguished Papuan and Austronesian societies 3,000 years ago, at the time of the Lapita archaeological spread through much of the western Pacific. The Lapita spread was probably associated with the initial Austronesian colonization of many of the western Pacific Islands, but it is significant that it appears to have avoided the island of New Guinea itself, where Austronesian speakers even today are found only in a few pockets of coastal distribution (Kirch and Green 2001).
Indeed, the linguist Robert Dixon (1997) has suggested that the overall history of the major language families can be separated into short periods of widespread dispersal, when the families are actually founded, interspersed with long periods like that described for New Guinea when populations of quite different linguistic and cultural origin interact, whether peacefully or belligerently. This hypothesis resembles the theory of punctuated equilibrium as applied to the biological evolution of species.
A final point to note is that, whereas archaeology and language history can often come together to throw independent light to support a plausible historical reconstruction, we often find that genetic data are not in full agreement. It is not the intention to discuss human genetics here, but it is perfectly obvious that not all of the speakers of some of the major language families are of tightly defined and geographically restricted genetic origin. For instance, the speakers of Austronesian languages range from Southeast Asians to Melanesians and Polynesians. The speakers of Indo-European languages range from northern Europeans to northern Indians. It is possible, but rather unlikely, that these differences represent no more than natural selection operating since the initial population dispersal which founded the language family in question. But it is far more likely that these differences reflect population mixing not always paralleled by an equivalent amount of language mixing. In other words, language families can have a life of their own, as can nodes of biological variation. Many population dispersals must have incorporated large numbers of the existing inhabitants of the newly settled regions, with consequent genetic effects stamped on later generations. This does not mean that there are no correlations between variations in language and biology in the human species, but we must be aware that any correlations will not always be clear-cut and obvious. They must be teased apart with care.
This field of archaeolinguistic research is not one in which we can expect absolute proofs for suggested correlations, particularly when dealing with prehistoric societies, but firm hypotheses are worthy of the research effort. The goals of archaeolinguistic research are laudable ones since they help us to interpret and understand so many fundamental developments and transitions in human prehistory.
