Phylogeny vs reticulation in prehistory.

Bellwood, Peter

The debate

We were standing in front of a painting with two white water lilies. I stepped a little closer to the picture and looked at it. It was then that I noticed that the lilies were nothing but blobs and blotches of paint. But when I stepped away again, they turned into real water lilies floating in a pond - magic?

BJORK & ANDERSON 1987: 14-15

It takes only a limited perusal of the ethnographic record to make the observation that human languages, cultures and gene pools do not co-vary in absolute unison. Politically-motivated attempts to make them co-vary, as we see in current arenas of ethnic conflict, always end in tragedy and disaster. Despite this, it is also apparent that cases of relatively strong co-variation have occurred in some instances in the past human pattern. A number of geneticists have recently presented evidence which they believe indicates a relatively close relationship between language families and genetic geography on a world scale (Cavalli-Sforza et al. 1988; 1994; Ruiz-Linares et al. 1995; Chen et al. 1995). Despite criticism, on grounds which vary from technical through theoretical to purely ethical (e.g. Bateman et al. 1990; Nei & Choudhury 1993; Moore 1995; Pluciennik 1996), the underlying logic of some degree of correlation between large-scale linguistic and genetic entities cannot easily be overturned.

This paper, not dealing with genetic data per se, focuses on an allied perspective from archaeology and linguistics. This perspective should clarify aspects of historical trajectory essential for any balanced consideration of such large-scale issues of correlation. It involves the respective roles of reticulation and phylogeny in the generation of cultural and linguistic patterns of variation. Reticulate models, as recently presented by anthropologists and archaeologists, stress the importance of continuing processes of interaction between contemporary communities. Phylogenetic models, stressing commonality of descent, imply dispersal by culturally and linguistically related populations from common origins in circumscribed homeland regions. The intention of this paper is to counter the growing tendency in the archaeological and anthropological literature to see only reticulate models as representing cultural reality, with phylogeny as a red herring in human cultural affairs (e.g. Terrell 1988; Rowlands 1994; Moore 1994a; 1994b). Both are necessary: reticulate models, if applied alone to explain long-term patterns of prehistoric cultural and biological variation, can overlook the large-scale patterning visible on continental and millennial scales in the linguistic and archaeological records.

Reticulate processes of pattern-formation

In two recent papers, Moore (1994a; 1994b) has asked how 'ethnic groups' - the named ethno-linguistic entities studied by ethnographers have been generated throughout the human past and in the ethnographic record. Moore criticizes 'bifurcatory' (phylogenetic) models and promotes 'rhizotic' (reticulate) ones. Reticulate processes focus on a continuous and relatively uncoordinated shifting of linguistic, cultural and biological boundaries through assimilation, intermarriage, borrowing and diffusion. According to Moore, they represent the true foundations of ethnogenesis; so interaction and borrowing, rather than the transmission of inherited traits, have always been the main sources of human cultural and biological variation.

Reticulate models have a distinguished pedigree. As stated by Jean Hiernaux over 30 years ago, referring to biological populations (Hiernaux 1964: 42):

Even if we could reconstruct the intricate succession of mixtures that contributed to each living population, the final picture would look like a reticulum more than a tree, and a reticulum defies dichotomizing subdivision.

A similar model was adopted by Morton Fried (1975) for the genesis of pre-state societies. Today the concept, albeit in varying guises, is commonplace. For instance, in the area of Oceanic prehistory it is expressed by John Terrell (1986; 1988), a contributor to a recent debate which in part confronts reticulate versus phylogenetic models as applied to the ethnographic record of northern New Guinea (Welsch et al. 1992; Moore & Romney 1994; Welsch & Terrell 1994; Roberts et al. 1995; Terrell et al. in press). In other current archaeological literature, the basic reticulate model is seen, although not always couched in the terminology used by Moore and other cultural anthropologists, in papers by Upham (1994: Southwestern archaeology, with strong parallels drawn with Atkinson's 1989 reticulate model of the development of ethnicity in northern Uganda); Rowlands (1994: Childean theory and creolization in archaeology); Dewar (1995: 'reticulate' versus 'dendritic' processes in the cultural history of Madagascar); and Gosden & Pavlides (1994: interaction-based models in the Arawe Islands, Papua New Guinea). Reticulate models are frequently invoked in discussing interfaces between archaeological and linguistic perspectives on prehistory; for instance, by Ehret (1988: material correlates of language and ethnic shift), Yoffee (1990: Indo-European prehistory) and Hall (1990: Bantu prehistory). One can hardly disagree with the principles of reticulate cultural differentiation as stated by these authors; humans flourish in interactive groups, and total isolation of any human group has been very rare in prehistory.

But is reticulate interaction the only process through which the generation of all human cultural and linguistic variation has occurred? Admittedly, many case-studies of interactive ethnogenesis in the ethnographic record are situated in cultural landscapes of most ambiguous phylogeny. If this ambiguity is often the case now, then perhaps it has always been the case, meaning that common origin has never been a significant factor in cultural patterning and any potential traces of it are simply the mimicking results of interaction. But is the ethnographic situation of the past few centuries a sufficient analogy for all human situations in the past?

Phylogenetic processes of pattern formation

Despite the current popularity of reticulate models, many favour phylogenetic models to explain in part the development of human variation from a base-line of shared ancestry (e.g. Kirch & Green 1987; Mace & Pagel 1994). In addition, language family subgrouping classifications, when based on the identification of shared innovations, are also inherently phylogenetic. Phylogenetic models, often expressed in 'family tree' form, are constrained by a requirement that genetic relationship cannot result entirely from the in situ convergence or mixing of unrelated entities; they require some degree of radiative dispersal of an originally more localized and unified entity. Logically, phylogenetic differentiation should occur most clearly during and after rapid and widespread dispersals of relatively homogeneous ethno-linguistic populations, in situations where interaction with the homeland region diminishes with distance. Indeed, the existence of truly phylogenetic relationships within an array of cultures and languages must imply some form of migration/dispersal trajectory at source. Similarities resulting from interaction alone are phenotypic, not phylogenetic.

Of course, there is a danger of circular reasoning here - traces of phylogeny are stated to imply movement from an ancestral source, and traces of such movements imply phylogenetic relationships. One can perhaps avoid this circularity by comparing traces of phylogenetic relationship and population dispersal revealed by the patterns of data within two quite independent disciplines - linguistics and archaeology. Language subgrouping relies on a strict comparative methodology to identify the cognate shared innovations which structure phylogenetic relationships. Sometimes it also uses phenotypic techniques such as lexico-statistics and mass comparison which rely on matrices of similarity coefficients, but here also the conclusions drawn are often stated to reflect phylogeny. Language families arise from the diversification of genetically related languages, not by convergence or by the formation of pidgins and creoles. Indeed, if one adopts the phylogenetic perspective (i.e. the view of 'genetic' relationship) favoured by most linguists for the definition of a language family, then one can hardly argue around the statement of Kaufman (1990: 21):

The fact that a set of languages can be shown to be genetically related entails that there was a real protolanguage, spoken by a particular group of people, in a particular region, at a particular time period.

I know of no convincing argument to refute this view.

Archaeology will never be likely to offer such transparent clues to phylogeny as those we see enshrined in the genetic structure of a language family, but archaeological expressions of material culture dispersal can at least be epiphenomena of cultural phylogeny. Archaeology and linguistics reveal very different structural perspectives on human prehistory - we cannot just 'mix them together' to know why human cultures differ, but we can use both independently to draw inferences when they intersect positively.

The importances of ethnography and scale

Reticulate models can explain some situations of small-scale ethnogenesis perfectly well, but on large geographical scales beyond the range within which human groups can be expected to interact directly they fall short as sufficient explanations for the totality of human variation (even allowing for indirect interaction via networks and chains). A differentiation, albeit with overlap, between large scale 'phylum-forming' and small-scale 'ethnic group-forming' processes surely lies at the heart of the issue under debate.

Those who propose reticulate interpretations to explain past human variation refer to parallels drawn from the vast repertoire of the ethnographic and ethnohistoric literature (e.g. Moore 1994a: 931 for Plains Indians; Welsch et al. 1992 for northern New Guinea). But a problem lies in the nature of this repertoire, which necessarily offers a multitude of small-scale observations on societies already under the influence, whether beneficial or malignant, of colonial authorities. Some of these societies were (and still are) in retreat from pre-colonial conditions of greater freedom for action and healthier demographic profiles. Ethnographers did not witness these pre-colonial conditions, and more to the point of this paper no ethnographer was ever around to witness all societies across, for instance, Asia or North America, evolving from 10,000 BP to the 20th century. The success of an analogy depends very much on the actual range of observations upon which it is based; extending the ethnographic corpus to include ancient Old World and recent Western colonial and linguistic history can give a quite different picture of how human cultural diversity is generated.

The large-scale patterns in the records of archaeology and comparative linguistics give perspectives, completely independent of those derived from ethnography, on the changing patterns of human affairs through many millennia across whole continents and archipelagoes, in many cases incorporating anonymous information residues about societies long extinct for whom we have no specific and individual records whatsoever. Only these disciplines can give the crucial clues to ancient phylogenetic dispersals; much comparative ethnography has a subject-matter rather too dependent upon reticulate interaction.

On phylum formation from a linguistic perspective

It is proposed here that large-scale language family relationships can reflect the existence of phylogenetic arrays in human terms, and demonstrate the reality of large-scale past population dispersals (Bellwood 1995b; in press a). It is quite inconceivable that processes of reticulate interaction could produce such demonstrably phylogenetic entities as language families from an array of genetically unrelated ancestral forms (implying a complete lack of inter-language comprehensibility). The vast majority of languages which are classified within such families as Indo-European, Austronesian, Bantu, Sino-Tibetan, Uto-Aztecan, Arawakan or Algonquian (plus many more), spread over thousands rather than merely hundreds of kilometres, are generally agreed to belong to these families by the majority of linguists. They also fall into sub-groups for which phylogenetic relationships, usually expressed in 'family tree' (or 'family rake') form and based on distributions of uniquely shared innovations, can be demonstrated. In other words, these language families are internally well subgrouped and externally well bounded, even though processes of borrowing and pidginization do mean that, from time to time, languages occur which may be genetically unclassifiable. Yet careful perusal of the linguistic literature indicates that true pidgins and creoles (as opposed to languages which have simply borrowed heavily from each other) are rare in numerical terms, except in specific circumstances generated during the colonial period of the last few centuries (Dyen et al. 1992: 12).

The essential question addressed here is whether these major language families have developed

1 through linguistic diffusion alone amongst relatively unmoving human populations (i.e. by convergence), or

2 have spread through language shift by populations previously speaking unrelated languages, or

3 have spread through dispersal of actual speakers of ancestral languages within these families.

Of these three processes, only the third requires actual population dispersal - the first two involve only diffusion. Naturally, all three processes work to some degree hand-in-hand, but it is unreasonable to assume that all three have worked in even proportions in all historical situations.

The first diffusion-based process on the required scale lacks empirical support in the ethnographic and historical records. No linguist has proposed a convincing mechanism to explain language family distributions on the scales of those just listed using convergence principles alone. Trubetzkoy's brief attempt to explain Indo-European in this way can be regarded as no more than an unsupported hypothesis (Trubetzkoy 1939).

The second shift-based process requires a language family to have spread through widespread adoption, through complete language shift via bilingualism, of a relatively uniform ancestral language or set of related dialects through unmoving populations. In theory, such progressive language shifts could occur through prestige or conquest. In some modern nations we see national languages spreading at the expense of local vernaculars owing to literacy, mass education and mass-media communications. But for small-scale preliterate societies such processes do not apply. Indeed, in societies where between-group contact is very frequent, as amongst the trading societies of ethnographic western Melanesia, we often find that linguistic diversity is relatively unaffected by constant inter-communication; people become bilingual where necessary and languages undergo borrowing, not shift and extinction (except in rare cases - see Ross 1991 and Dutton 1995 for some Melanesian situations). In the prehistoric situation it is impossible to visualise a driving and constant socio-cultural factor which could spread (for example) some early Indo-European target language(s) across an area extending ultimately from Ireland to Bangladesh amongst small-scale Neolithic or Bronze Age societies (and multilingualism by itself certainly could not account for this). Even if such a factor existed one still has to handle the problem of linguistic interference in the target languages themselves through the shift process (see Thomason & Kaufman 1988). If unmoving and presumably ethnically diverse populations over an area of sub-continental size gradually adopted an incoming set of target languages, eventually abandoning their natal vernaculars yet without being incorporated into a single political entity with a widespread network of frequent communication, we need to ask if the linguistic end-product would, after a few millennia, be a recognizable language family. Ethnography and history cannot answer this question on the scale required, but history can at least tell us that when large-scale language spread has happened, it has not happened by this means.

The third process, based on population dispersal, can draw support from parallels with episodes of European colonization after AD 1500 and, for quite different and negative reasons, from the records of ancient Old World conquest empires. As with convergence and language shift we have no ethnographic examples of population dispersal on the scale under examination, but in this case we do have historical ones from the recent past. Recent recorded histories of language spread make it perfectly obvious that population dispersal has been the main vector of large-scale language dispersal at the whole-population vernacular (not just the elite or lingua franca) level. This is true of English in North America and Australasia. It is not true of English in India, where only a small percentage of the total population is literate in English (Wardhaugh 1987: 12 gives an estimate of 2%), despite its widespread use in printed form. For a variety of reasons discussed by Crosby (1986), European colonization of densely populated tropical countries such as India and Indonesia was not successful, hence the inability of European languages to become dominant vernaculars.

Like English, Spanish was also spread mainly by colonization in the first instance (e.g. Hale & Harris 1979: 182 for New Mexico). While Spanish attitudes allowed considerable intermarriage with native peoples, the fact remains that many Amerindian communities which still have strong demographic and cultural profiles, in Mesoamerica and much of the Andes, keep their native languages (Heath & Laprade 1982; cf. Schooling 1990 for native language 'tenacity' in the face of French in New Caledonia). They have no more converted universally to English and Spanish than the English converted universally to French in the centuries of Norman domination.

Concerning such limitations of the 'elite dominance' model for language spread, one should also bear in mind Dyen's observation (1990: 219) that '. . . conquest and subsequent heavy contribution by the language of the conquerors to that of the conquered is an exceptional event rather than a rule' (see also Wurm 1991). Although they conquered, widespread colonization was an ingredient lacking in the histories of many ancient empires such as those of the Romans, Persians and Mongols. Latin was in use as a whole-population vernacular in only about one-third of the geographical extent of the Roman Empire in its latter days, reasoning from the development of the Romance languages and taking into account the continuing vernacular uses of languages such as Greek, Coptic, Aramaic, and Germanic and Celtic languages in the north. This occurred despite the use of Latin as a language of government over the whole of the empire and despite the power of Roman rule and literacy for over four centuries. Those areas where Romance languages eventually became the vernaculars of total populations were close to the heart of the empire (Italy, Iberia and Gaul, with more remote Dacia/Romania) and favoured for colonization by Roman citizens, especially retired soldiers and officials (Brosnahan 1969). These settlers surely communicated in Latin. Rome was not predominantly a colonizing empire and as a result many regions, some close to the centre like Greece and Illyria, maintained other vernaculars throughout the Roman period.

Likewise, the spread of Arabic involved mainly the outwards colonization of farming and urban populations (not nomads) from the Arabian Peninsula into southwest Asia and northeastern Africa (Holt et al. 1977). Where Arab colonization did not occur, even with a rapid adoption of the Islamic religion, it appears that Arabic likewise did not spread as the dominant vernacular. Indonesia today has the world's largest Moslem population, but the main linguistic influence of Arabic, like the earlier Sanskrit, is confined to loans into Austronesian vernaculars such as Javanese and Malay.

Another observation to support population dispersal as the main vector of major language family spread concerns timing on the millennial scale. The greater part of world language family distribution before AD 1500 (excluding colonial spreads of European languages) was already in place geographically before regional historical records began, that is, in prehistoric time. Written history might explain the spreads of Latin and Greek, of Quechua and Chinese, but written history does not account for the distribution of the whole Indo-European family from Ireland to Bangladesh, of Uto-Aztecan from Mexico City to Idaho, or of Austronesian more than half way around the world from Madagascar to Easter Island. These language dispersals, amongst the most remarkable large-scale cultural patterns to result from human behaviour, were the achievements of small-scale prehistoric societies, in many cases of hunter-gatherers (e.g. Athapaskans, Eskimo-Aleuts and some of the Uto-Aztecans) or early agriculturalists (Bellwood 1995a). They are absolutely without parallel in the record of ethnography, a record too narrow to instruct at that scale.

Comparative linguistics and historical/ethnographic observation together indicate that the major language families were derived from much smaller areas than they now occupy, ultimately from one or a few closely related ancestral languages. They have been spread mainly by processes of population dispersal, basically leading to bifurcatory or radiative processes of ethnic differentiation, fuelled by factors such as distance and adaptation to differing environmental circumstances, but overlain by continuous reticulate types of interaction. These dispersals have not uniformly occurred at the same rate, through time or across space. There have been times when population expansions have occurred very insistently, sometimes via slow demographic expansion, other times via more explosive radiations. At times of non-dispersal, reticulate processes of interaction have doubtless dominated in prehistory, just as they dominate today in the ethnography of areas such as the Great Plains, Melanesia or Borneo (Tillotson 1994). It is not that either dispersal-based or reticulate processes of ethnogenesis have dominated human history in any total sense; both have occurred to differing degrees in different times and places. The past has not witnessed a uniformitarian trajectory in human affairs all over the globe.

Archaeological circumstances of human population expansion

Why should certain language families have spread with population dispersal over such vast distances in prehistoric times? The initial radiations of human populations into previously human-free regions such as Australia, the Pacific Islands and the Americas must surely have led to many episodes of cultural and biological differentiation into phylogenetically related units - if the early colonists were culturally and biologically related in the first instance, a likely possibility if they were drawn from source regions of relatively restricted extent. These initial radiations are not my main theme, since I regard them as almost self-evident examples of phylogenetic population differentiation, especially in the cases of recent and well-understood radiations such as that of the Polynesians (Kirch & Green 1987).

Instead, the core concept discussed here revolves around agricultural expansion during those early centuries in various parts of the world when successful systems of food production replaced hunter-gather precursors (Bellwood 1984-5; 1991; 1994; 1995a; 1996a; 1996b; in press a; in press b; Renfrew 1987; 1992a; 1992b; 1994; Ammerman & Cavalli-Sforza 1984; Cavalli-Sforza et al. 1993; Barbujani et al. 1995). One can extract certain significant generalizations from a number of disciplines which greatly affect our understanding of the prehistory of the past 10,000 years in those temperate and tropical latitudes where environmental factors have allowed agricultural production to take place. Some of these generalizations are as follows:

1 Primary agricultural development (i.e. local uninfluenced evolution from forager to farmer) has probably only occurred, according to the archaeological and botanical records, in a very few regions of the world (southwestern Asia, central and southern China, New Guinea Highlands, Mesoamerica, northern Andes, eastern Woodlands of the USA, perhaps central Africa). Agriculture has spread from these regions mainly (but not entirely) by demographic growth of the agricultural populations themselves (cf. Barbujani et al. 1995 for Europe) rather than by adoption of agriculture by pre-existing and unmoving hunter-gatherers (Bellwood 1990; 1994 and in press a). This statement is based on surveys of the relevant archaeological, linguistic and genetic records and on the records of forager/agriculturalist interaction available in the ethnographic literature. Some archaeologists regard agriculture as easily diffusable to 'receptive' foragers (e.g. Gebauer & Price 1992: 8), but evidence suggests that such foragers have mostly existed only in agriculturally marginal regions, especially in rich maritime environments (such as parts of coastal Europe) where agriculture was restricted by geomorphic or climatic factors and where existing foragers were able to maintain sufficient demographic balance against incoming agriculturalists to allow for successful interaction and diffusion of ideas and techniques.

2 The languages within geographically extensive language families (such as Austronesian, Indo-European, Sino-Tibetan) carry linguistic traces of common ancestry and dispersal from a homeland region. They cannot logically result from convergence of phylogenetically unrelated forms but must have spread, to the great extents attained by examples such as the above, by means predominantly of population colonization. They have not spread by language shift alone, even though processes of language shift have operated intensively in many localized group-to-group circumstances recorded ethnographically and historically.

3 Language family homelands, as determined from subgroup relationships based on comparative method reconstructions of shared innovations, proto-language vocabularies, and lexicostatistical calculations (the latter subject to many provisos owing to varying rates of change), indicate that many of the major agriculturalist language families originated in relatively restricted areas of the earth's surface, in areas - not coincidentally - where agriculture originated from primary foraging base-lines (e.g. southwest Asia for Indo-European, Turkic, Elamite (with Dravidian?), Sumerian, Semitic and perhaps Afro-Asiatic as a whole; central and southern China for Tai-Kadai, Austronesian, Austroasiatic, Hmong-Mien and possibly Sino-Tibetan; Mesoamerica for Mayan, Mixe-Zoque, Otomanguean and perhaps Uto-Aztecan: Bellwood 1994; in press b). Regions of agricultural origin are regions where the geographical distributions of many different language families intersect, and also where the origin zones of many of those intersecting language families were located. Over time, populations have tended consistently to move out from such zones of primary agriculture through demographic growth, rather than in; their languages have moved outwards with them (the exceptions do not negate the general pattern at the language family level). Conversely, regions which are not known from the archaeological record to be areas of primary agricultural origin, e.g. peninsular South Asia (central and southern India and Sri Lanka), are also not the origin regions of the language families which now occupy them (the Indo-Aryan, Munda and Dravidian language groups did not originate in central and southern Peninsular India according to the majority of comparative linguists).

Shifts from complex foraging to systematic agriculture, especially in the cereal-based regions of primary agricultural origin, led to rapid and in some cases explosive population growth (cf. Cavalli-Sforza 1983), a population growth archaeologically visible if we accept the rapid growths in sizes and densities of sites consequent upon the establishment of identifiable agricultural activities (cultivation and animal husbandry) in regions such as the Levant, China, the northern Andes and Mesoamerica. Many of these populations presumably began relatively quickly (depending on the nature of the particular regional trajectory from foraging to farming), as a result of insistent internal population growth, to colonize adjacent areas inhabited only by foragers, who in many cases were probably incorporated into the larger agricultural populations (thus imposing a reticulate flavour to the whole process). After relatively short periods of time (averaging perhaps between one and two millennia) some of these expansions could have reached the huge extents attained by groups such as Austronesians (Bellwood et al. 1995), Indo-Europeans (Renfrew 1987), and Uto-Aztecans (many of these apparently switching back to foraging in the marginal climates of the Great Basin), all before any historically recorded major language expansions had taken place. Historical expansions of individual languages - e.g. Thai, Burmese, Malay and Chinese in eastern Asia, Quechua and Nahuatl in the Americas - have been numerous, but in no case can these historical movements account for the total distribution of a major family of languages.

We can also see many examples of the expansion of relatively homogeneous packages of material culture in the regional archaeological records of early agriculture. This kind of homogeneity could be a result of phylogenetic patterning; unfortunately the archaeological record per se is unable to give unambiguous evidence for spreads of population as opposed to spreads of material culture and stylistic elements through interaction. Examples include the Danubian of central Europe; Lapita and Island Southeast Asian red-slip pottery assemblages in the Philippines/Indonesia and Oceania; Mesoamerican and Andean Formative cultures (Ford 1969: still a very significant survey despite recent criticisms, as by Hoopes 1994); and much of the Chinese and Southeast Asian early Neolithic (Bellwood 1995a; Higham 1996). In many regions, as time progressed forwards from periods of agricultural beginnings, so archaeological patterns became more fragmented. This is certainly true of Southeast Asia and the western Pacific (compare, for instance, Lapita cultural homogeneity at c. 3000 BP with the highly variable archaeological and ethnographic picture of the past 2000 years in Melanesia), and it appears true of the agricultural regions of the Americas, although statistical demonstrations of such trends are still only in infancy (see Zeitlin 1994 for a thought-provoking commencement). This early agricultural-phase homogeneity in so many regions of the world archaeological record can hardly document diffusion alone, simply because diffusion alone via some kind of 'interaction sphere' model through unmoving, pre-existing and diverse hunter-gatherer populations could seemingly not lead to all the shared elements of iconography and style which make the archaeological complexes listed above so strikingly homogeneous.

Some of the ethnolinguistic expansions suggested in this paper were undoubtedly variable and erratic in speed and periodicity; Austronesian colonization seems to have been very rapid in eastern Indonesia and the western Pacific, but delayed by a millennium or so into eastern Polynesia (Bellwood 1987: 56-7) and on to the mainland of Southeast Asia (Malay Peninsula and southern Vietnam: Blust 1984-5). Likewise, the Indo-European colonization of northern India (beyond the Harappan region) apparently took place some millennia after movement into eastern Europe, perhaps owing to ecological factors involved in the move from a winter-rainfall western Asian environment into a summer-rainfall subtropical monsoon environment, with consequent economic stresses for agriculturalists. Speed can be read not only from the distribution of absolute dates for comparable sections of the archaeological record, but also from the subgroup relationships of the language families themselves. Tree-shaped hierarchical structures of subgroup relationships should reflect a slower outward dispersal and continuing internal contact since they depend on relatively long periods of regional stasis and interaction in order to accumulate the shared subgroup innovations which produce the hierarchical tree shape. A rapid spread over a large geographical region should yield a linguistic family tree of a 'rake' shape, with very little hierarchical differentiation at the highest level. Language families which remain resistant to overall hierarchical classification, such as Indo-European (Dyen et al. 1992), Uto-Aztecan (Ruhlen 1987: 237) and many of the Melanesian languages which belong to the Oceanic sub-group of Austronesian (Ross 1989; Pawley & Ross 1995), might have spread rapidly in the first instance, with early founder populations moving quickly into situations of diminished contact with speakers of other languages within the same family.

Conclusion

The linguistic and archaeological records, considered in parallel, show widespread expansions by early agriculturalists, expansions hard for us to visualize today since recorded history has seen only those major expansions set in train by European colonization, and ethnographic dispersals of relatively minor overall extent such as the Iban expansion in Borneo. We cannot assume that the Neolithic past is just a stripped-down version of the ethnographic record of human behaviour. Linked population and language expansions probably occurred after the regional beginnings of agriculture in many areas on a scale broadly like recent colonial dispersals from the British Isles and Iberia. Large-scale and fairly integrated colonizations did happen in prehistory; human cultures and languages can, to varying degrees dependent upon time and space coordinates, be organized in phylogenetic arrays. The generation of human diversity in the past has not been entirely reticulate and dependent on processes of in situ interaction between peoples of different ethno-linguistic background. Neither has it been entirely radiative and dependent on adaptation in isolation. But to rule out phylogeny as of any significance in the patterning of difference and similarity between human cultures is surely no more than a 'whimsical view', in the words of Mace & Pagel (1994: 563).

Acknowledgements. Earlier versions of this paper were read by Andrew Pawley and Bob Dixon and both provided useful linguistic comments, although the overall viewpoint is the author's own. Thanks also to three referees for pointing out a few inconsistencies.

References

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

ATKINSON, R.R. 1989. The evolution of ethnicity among the Acholi of Uganda: the precolonial phase, Ethnohistory 36: 19-43.

BARBUJANI, G., R.R. SOKAL & N.L. ODEN. 1995. Indo-European origins: a computer-simulation test of five hypotheses, American Journal of Physical Anthropology 96: 109-32.

BATEMAN, R., I. GODDARD, R. O'GRADY, V.A. FUNK, R. MOOI, W.J. KRESS & P. CANNELL. 1990. Speaking of forked tongues: the feasibility of reconciling human phylogeny and the history of language, Current Anthropology 31: 1-24.

BELLWOOD, P. 1984-5. A hypothesis for Austronesian origins, Asian Perspectives 26: 107-17.

1987. The Polynesians. 2nd edition. London: Thames & Hudson.

1990. Foraging towards farming: a decisive transition or a millennial blur?, Review of Archaeology 11(2): 14-24.

1991. The Austronesian dispersal and the origin of languages, Scientific American 265(1): 88-93.

1994. An archaeologist's view of language macrofamily relationships, Oceanic Linguistics 33: 391-405.

1995a. Early agriculture, language history and the archaeological record in China and Southeast Asia, in C.-T. Yeung and B.W.-L. Li (ed,), Archaeology in Southeast Asia: 11-22. Hong Kong: University Museum & Art Gallery.

1995b. Language families and human dispersal, Cambridge Archaeological Journal 5: 271-4.

1996a. Early agriculture and the dispersal of the Southern Mongoloids, in T. Akazawa & E. Szathmary (ed.), Prehistoric Mongoloid dispersals: 287-302. Oxford: Oxford University Press.

1996b. The origins and spread of agriculture in the Indo-Pacific region, in D. Harris (ed.), The origins and spread of agriculture and pastoralism in Eurasia: 465-98. London: UCL Press.

In press a. Prehistoric cultural explanations for the existence of widespread language families, in P. McConvell & N. Evans (ed.), Understanding ancient Australia: perspectives from archaeology and linguistics. Melbourne: Oxford University Press.

In press b. Prehistory of the Indo-Malaysian archipelago. 2nd edition. Honolulu (HI): University of Hawaii Press.

BELLWOOD, P., J.J. Fox & D. TRYON (ed.). 1995. The Austronesians: historical and comparative perspectives. Canberra: Department of Anthropology, Australian National University.

BLUST, R. 1984-5. The Austronesian homeland: a linguistic perspective, Asian Perspectives 26(1): 45-68.

BJORK, C. & L. ANDERSON. 1987. Linnea in Monet's garden. Stockholm: R&S.

BROSNAHAN, L.F. 1969. Some historical cases of language imposition, in J. Spencer (ed.), Language in Africa: 7-24. Cambridge: Cambridge University Press.

CAVALLI-SFORZA, L.L. 1983. The transition to agriculture and some of its consequences, in D.J. Ortner (ed.), How humans adapt: 103-26. Washington (DC): Smithsonian Institution.

CAVALLI-SFORZA, L.L., A. PIAZZA, P. MENOZZI & J. MOUNTAIN. 1988. Reconstruction of human evolution: bringing together genetic, archaeological and linguistic data, Proceedings of the National Academy of Science of the USA 85: 6002-6.

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

(Ed.). 1994. The history and geography of human genes. Princeton (NJ): Princeton University Press.

CHEN, J., R. SOKAL & M. RUHLEN. 1995. Worldwide analysis of genetic and linguistic relationships of human populations, Human Biology 67: 595-612.

CROSBY, A. 1986. Ecological imperialism. Cambridge: Cambridge University Press.

DEWAR, R. 1995. Of nets and trees: untangling the reticulate and dendritic in Madagascar's prohistory, World Archaeology 26: 301-18.

DUTTON, T. 1995. Language contact and change in Melanesia, in P. Bellwood et al. 1995: 192-213.

DYEN, I. 1990. Homomeric lexical classification, in P. Baldi (ed.). Linguistic change and reconstruction methodology: 211-30. Berlin: Mouton de Gruyter.

DYEN, I., J.B. KRUSKAL & P. BLACK. 1992. An Indoeuropean classification: a lexicostatistical experiment, Philadelphia (PA): American Philosophical Society. Transactions 82(5).

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

FORD, J.A. 1969. A comparison of formative cultures in the Americas. Washington (DC): Smithsonian Institution. Contribution to Anthropology 11.

FRIED, M. 1975. The notion of tribe. Menlo Park (CA): Cummings.

GEBAUER, A.B. & T.D. PRICE. 1992. Foragers to farmers: an introduction, in A.B. Gebauer & T.D. Price (ed.), Transitions to agriculture in prehistory: 1-10. Madison (WI): Prehistory Press.

GOSDEN, C. & C. PAVLIDES. 1994. Are islands insular? Landscape vs seascape in the case of the Arawe Islands, Papua New Guinea, Archaeology in Oceania 29: 162-71.

HALE, K. & D. HARRIS. 1979. Historical linguistics and archaeology, in A. Ortiz (ed.), Handbook of North American Indians 9: Southwest: 170-77. Washington (DC): Smithsonian Institution.

HALL, M. 1990. Farmers, kings and traders. Chicago (IL): University of Chicago Press.

HEATH, S.B. & R. LAPRADE. 1982. Castilian colonization and indigenous languages: the cases of Quechua and Aymara, in R.L. Cooper (ed.), Language spread: 118-47. Bloomington (IN): Indiana University Press.

HIERNAUX, J. 1964. The concept of race and the taxonomy of mankind, in A. Montagu (ed.), The concept of race: 29-45. New York (NY): Free Press.

HIGHAM, C. 1996. Archaeology and linguistics in Southeast Asia: implications of the Austric hypothesis, Bulletin of the Indo-Pacific Prehistory Association 14: 110-18.

HOLT, P.M., LAMBTON, A.K.S. & B. LEWIS (ed.). 1977. The Cambridge history of Islam. IA. Cambridge: Cambridge University Press.

HOOPES, J.W. 1994. Ford revisited, Journal of World Prehistory 8: 1-49.

KAUFMAN, T. 1990. Language history in South America, in D.L. Payne (ed.), Amazonian linguistics: 13-73. Austin (TX): University of Texas Press.

KIRCH, P.V. & R.C. GREEN. 1987. History, phylogeny and evolution in Polynesia, Current Anthropology 28: 431-56.

MACE, R. & M. PAGEL. 1994. The comparative method in anthropology, Current Anthropology 35: 549-64.

MOORE, C. & A.K. ROMNEY. 1994. Material culture, geographic propinquity, and linguistic affiliation on the north coast of New Guinea: a reanalysis of Welsch, Terrell & Nadolski (1992), American Anthropologist 96: 370-96.

MOORE, J.H. 1994a. Putting anthropology back together again: the ethnogenetic critique of cladistic theory, American Anthropologist 96: 925-48.

1994b. Ethnogenetic theory, National Geographic Research and Exploration 10(1): 10-23.

1995. The end of a paradigm, Current Anthropology 36: 530-31.

NEI, M. & A.K. CHOUDHURY. 1993. Evolutionary relations of human populations on a world scale, Molecular Biology and Evolution 10: 927-43.

PAWLEY, A.K. & M. ROSS. 1995. The prehistory of the Oceanic languages: a current view, in Bellwood et al. (ed.): 39-74.

PLUCIENNIK, M. 1996. Genetics, archaeology and the wider world, Antiquity 70: 13-14.

RENFREW, C. 1987. Archaeology and language. London: Jonathan Cape.

1992a. Archaeology, genetics and linguistic diversity, Man (n.s.) 27: 445-78.

1992b. World languages and human dispersals: a minimalist view, in J.A. Hall & I.C. Jarvie (ed.), Transition to modernity: 11-68. Cambridge: Cambridge University Press.

1994. World linguistic diversity, Scientific American 268(7): 104-10.

ROBERTS, J.M., C.C. MOORE & A.K. ROMNEY. 1995. Predicting similarity in material culture among New Guinea villages from propinquity and language, Current Anthropology 36: 769-88.

Ross, M. 1989. Early Oceanic linguistic prehistory, Journal of Pacific History 24: 135-49.

1991. Redefining Guy's sociolinguistic types of language change, Diachronica 8: 119-29.

ROWLANDS, M. 1994. Childe and the archaeology of freedom, in D. Harris (ed.), The archaeology of V. Gordon Childe: 35-54. London: UCL Press.

RUHLEN, M. 1987. A guide to the world's languages. Stanford (CA): Stanford University Press.

RUIZ-LINARES, A., E. MINCH, D. MEYER & L.L. CAVALLI-SFORZA. 1995. Analysis of classical DNA markers for reconstructing human population history, in S. Brenner & K. Hanihara (ed.), The origin and past of modern humans as viewed from DNA: 123-48. Singapore: World Scientific.

SCHOOLING, S. 1990. Language maintenance in New Caledonia. Dallas (TX): Summer Institute of Linguistics.

TERRELL, J. 1986. Prehistory in the Pacific Islands. Cambridge: Cambridge University Press.

1988. History as a family tree, history as an entangled bank, Antiquity 62: 642-57.

TERRELL, J., GOSDEN, C. & T. HUNT. In press. The dimensions of social life in the Pacific, Current Anthropology 38(2).

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

TILLOTSON, D. 1994. Who invented the Dayaks? Unpublished Ph.D thesis, Australian National University, Canberra.

TRUBETZKOY, N.S. 1939. Gedanken uber das Indogermanen-problem, Acta Linguistica (Hafniensia) 1: 81-9.

UPHAM, S. 1994. Nomads of the Desert West: a shifting continuum in prehistory, Journal of World Prehistory 8: 113-68.

WARDHAUGH, R. 1987. Languages in competition. Oxford: Blackwell.

WELSCH, R.L., J. TERRELL & J.A. NADOLSKI. 1992. Language and culture on the North Coast of New Guinea, American Anthropologist 94: 568-600.

WELSCH, R. & J. TERRELL. 1994. Reply to Moore & Romney, American Anthropologist 96: 392-6.

WURM, S.A. 1991. Language death and disappearance, in R.H. Robins & E.M. Uhlenbeck (ed.), Endangered languages: 1-12. Oxford: Berg.

YOFFEE, N. 1990. Before Babel, Proceedings of the Prehistoric Society 56: 299-314.

ZEITLIN, R.N. 1994. Accounting for the prehistoric long-distance movement of goods with a measure of style. World Archaeology 26: 208-34.