The Dispersal and Diversification of Oceanic Languages

Subgrouping: the sequence of genetic splits in Oceanic

Kinds of subgroups.

Innovations occur in two basic patterns across languages, enabling linguists to identify subgroups. In the first pattern, all member languages of a subgroup exclusively share a common set of innovations. In the second pattern, innovations form an overlapping pattern, such that, for example, languages A, B and C reflect one bunch of innovations, languages C, D, and E another bunch, languages D, E, F, and G yet another, and languages G and H still a different bunch of innovations. A subgroup reflecting the first pattern may be termed an “innovation-defined” subgroup. One reflecting the second pattern may be called an “innovation-linked” subgroup.[8]

The two patterns allow linguists to make certain inferences about how languages have diverged. [9]

In the case of an innovation-defined subgroup, we infer that all member languages of the subgroup exclusively share a set of innovations because these innovations occurred in the language ancestral to them. That is, the subgroup has been formed because a community speaking a single language has become separated geographically and/or socially into two or more communities; after separation, changes have occurred in the speech of each of the new communities until what was one language has become two or more.

The inference that can be made about an innovation-linked subgroup is of a different order. Here, it can be inferred that the overlapping pattern of innovations reflects the fact that the languages of the subgroup once formed a network of related dialects. During this phase, innovations occurred at various places in the network, spreading from their dialect of origin into neighbouring dialects, but without affecting the whole network. Over time the dialects have diverged until they have become mutually unintelligible for practical purposes, but they continue to reflect the innovation pattern of the former network.[10]

Innovation-defined and innovation-linked subgroups are not mutually exclusive. A subgroup may be both: innovation-defined in that there are innovations shared by all members of the subgroup, and innovation-linked because there is also a pattern of overlapping innovations. In such a case, both sets of inferences apply. The subgroup is innovation-defined because its member languages are descended from a single interstage language, and innovation-linked because that interstage language diversified into a dialect network out of which the present-day languages have evolved.

With the small size and geographical isolation of political entities in Oceania, dialect networks have readily become innovation-linked subgroups of discrete languages. The result is that innovation-linked subgroups are important in Oceanic historical linguistics. But they are also problematic. The internal structure of an innovation-linked subgroup is known to us through its patterning of innovations, but, unless the subgroup is also innovation-defined, we have no direct evidence as to its origin: it may be descended from a discrete interstage language which happens not to have undergone an identifiable set of innovations, or it may simply be descended from a detached section of an earlier dialect network.

In the 1950s and 1960s, comparative linguists writing in the Austronesian field saw their primary task as genetic classification and reconstruction, and applied a simple-minded family tree model in which all language splits are assumed to be sudden and complete. But a family-tree model is often unsatisfactory for making sense of and for representing historical relationships among languages. One reason is that it forces those linguistic relationships produced by dialect differentiation (and subsequent break-up of the network) into a distorted scenario, that of the sharp separation model — and in early Oceanic, dialect differentiation and network-breaking were the rule rather than the exception. Since the early 1970s Austronesianists have given more attention to the problem of dialectal diffusion in prehistoric languages. In the Oceanic field the most incisive study of this kind is Geraghty’s (1983) treatment of the relationship of the Fijian speech traditions to each other and to Polynesian. Other works that tackle dialect diffusion in particular regions include Pawley (1975) and Ross (1994b) on Central Papua; Ross (1988) on various regions of western Melanesia; Lichtenberk (1988) on Malaita; Clark (1985) and Tryon (1976) on Vanuatu; Pawley and Sayaba (1971) on Fiji; Pawley (1979) on Fiji and Rotuma; Biggs (1980), Dyen (1981) and Rensch (1987) on western Polynesia; and Jackson (1983) and Marck (1986) on Nuclear Micronesia.

The Comparative Method is not equipped to investigate certain kinds of historical change, such as structural convergence resulting from bilingualism, or from “drift” (pressures inherent in certain structural types that allegedly favour parallel changes). Recent work has begun to address these phenomena using a range of methods (e.g. Thurston 1987; Ross 1987; Ross forthcoming; and a number of the papers in Dutton and Tryon 1994).

High-order subgroups.

The following high-order subgroups of Oceanic seem fairly well supported:[11]

  1. Admiralty Islands
  2. St Matthias Islands
  3. Western Oceanic
  4. Southeast Solomonic
  5. North/Central Vanuatu
  6. South Vanuatu
  7. Southern Oceanic
  8. Central Pacific
  9. Nuclear Micronesian

The regions covered by these groups are shown in Map 2. We are not seeking to claim that POc underwent a primary nine-way split, but simply that in our view persuasive arguments have yet to be presented for higher-order groupings.[12] It should be noted that Groups 3, 5 and 8 are innovation-linked subgroups: there is no innovation which all their members share relative to POc.

The Admiralties group is clearly established as an innovation-defined subgroup (Blust 1978b; Ross 1988:ch. 9). The St Matthias group is also innovation-defined, with the possibility that it is associated in some way with the Admiralties; it certainly is not a close relative of its southern neighbours, the languages of New Ireland (Ross 1988:ch. 9).

The Western Oceanic grouping was proposed as an innovation-linked subgroup by Ross (1988:386-389). Except in a few small relic areas the languages of the region merge POc *r and *R, share reflexes of the innovative disjunctive pronoun *idri[a] ‘they’ (in contrast to POc *(k)ira), and sporadically reflect other features which imply that the languages of the region are descended from a dialect network which has a history separate from that of other Oceanic groups.

The Southeast Solomonic group is a fairly well innovation-defined and apparently very conservative subgroup, which divides in its turn into the Guadalcanal/Gela and Malaita/Makira subgroups (Pawley 1972; Levy 1979, 1980; Tryon and Hackman 1983; Lichtenberk 1988).

North/Central Vanuatu is an innovation-linked subgroup which has been documented by Pawley (1972), Tryon (1976) and Clark (1985). Clark views the subgroup as the outcome of a dialect network which gradually differentiated out into local languages. He recognizes a major division between North and Central Vanuatu, “with the boundary running between Santo and Malekula and between Raga and the remainder of Pentecost” (1985:221).

The South Vanuatu group embraces the languages of the islands of Erromanga, Tanna and Aneityum, and is an innovation-defined subgroup documented by Lynch (1978).

Geraghty (1989) has proposed a New Caledonian group (he calls it “Southern Oceanic”) which includes all the languages of New Caledonia and the Loyalty Islands. He also suggests that New Caledonian may subgroup with South Vanuatu, and implies that this grouping may form an innovation-linked subgroup with North/Central Vanuatu.[13] However, the evidence for this is limited and in our view not yet persuasive.

The Central Pacific grouping has been well studied (Grace 1959; Pawley 1972; Geraghty 1983, 1986, 1990). Proto-Central Pacific was evidently located in the Fijian archipelago, where it became differentiated into a dialect network with considerable variation from east to west. Speakers of an eastern dialect separated, probably by moving to Tonga and other islands in western Polynesia, where Proto-Polynesian developed (Geraghty 1983; Green 1981). Speakers of a western dialect moved to Rotuma, where a distinct language developed (Pawley 1979). Subsequently, the stay-at-home dialects in Fiji continued to interact, so that they now form an innovation-linked group, within which an eastern and a western subgroup are clearly distinguishable (Pawley and Sayaba 1971; Geraghty and Pawley 1981; Geraghty 1983). Proto-Polynesian has diversified into the languages of the Polynesian group, a very well defined subgroup of Oceanic (Pawley 1966, 1967; Biggs 1971).

Map 2: High-order subgroups of Oceanic Austronesian languages.

Map 2: High-order subgroups of Oceanic Austronesian languages.

Nuclear Micronesian includes all languages of Micronesia except Belau, Chamorro and probably Yapese (Bender 1971; Jackson 1983; Bender and Wang 1985).[14]

Subgroups of Western Oceanic.

Ross (1988) proposes that the Western Oceanic group is made up of three subgroups or “clusters”: Meso-Melanesian, Papuan Tip and North New Guinea (see Map 3).[15] The first two of these are innovation-defined subgroups, and in both cases the interstage proto-language or an early daughter interstage evidently differentiated into an extensive network. The North New Guinea grouping, however, is an innovation-linked subgroup, in that there are no innovations shared by all languages of the cluster relative to POc (Ross 1988:120). Instead, the cluster consists of three innovation-defined groups (Ngero/Vitiaz, Huon Gulf and Schouten), each defined by a set of shared innovations and each including lower-order subgroups (see Map 4). One or more subgroups within each innovation-defined group also has some innovations which it shares with one or more subgroups of the other two innovation-defined groups, joining the three in a network relationship.[16]

Map 3: Oceanic subgroups in western Melanesia.

Map 3: Oceanic subgroups in western Melanesia.

Map 4: The North New Guinea grouping and its subgroups.

Map 4: The North New Guinea grouping and its subgroups.

Map 5: The Meso-Melanesian grouping and its subgroups.

Map 5: The Meso-Melanesian grouping and its subgroups.

The internal structure of the Meso-Melanesian subgroup (Map 5) is very complex. A primary three-way split into the Bali-Vitu language, the Willaumez chain, and the New Ireland/Northwest Solomonic linkage (stretching from New Hanover to Santa Isabel) is recognizable. The latter in its turn seems to comprise the Madak, Tabar and Lavongai/Nalik groups which occupy central and northern New Ireland and the far-flung South New Ireland/Northwest Solomonic linkage. However, since the departure from southern New Ireland (we assume) of those whose descendants became the speakers of Northwest Solomonic languages, the languages the whole length of New Ireland have continued to interact so that they form an innovation-linked group. The New Ireland communalects have thus behaved similarly to those of Fiji.

Map 6: The Papuan Tip grouping and its major subgroups.

Map 6: The Papuan Tip grouping and its major subgroups.

The structure of the Papuan Tip subgroup (Map 6) is less complex. There is a possible (but not particularly well supported) initial split into “Nuclear” and “Peripheral” Papuan Tip linkages. The languages of the Nuclear network split into the Suauic chain (located mostly on the southeastern-most portion of the mainland) and the North Mainland/D’Entrecasteaux network, which forms a rough arc stretching from Normanby Island, through Fergusson and Goodenough Islands to the mainland coast at Collingwood Bay, then southeastwards along and around the coast as far as Milne Bay (Ross 1988:191, 1992).

Melanesia”. The reader will notice that nowhere do we speak of “Melanesian” languages or a “Melanesian” group. In describing the location of the Oceanic languages, one cannot avoid “Melanesia” as a geographic term. Nowadays, however, the comparative linguist referring to the genetic grouping of languages recognizes no such adjective as “Melanesian”. Melanesia is occupied, as we have noted, by speakers of both Papuan and Austronesian languages, but even among Austronesian languages there is no “Melanesian” group. Instead, some languages spoken in Melanesia are more closely related to Polynesian languages than to other languages of Melanesia.

Locating the POc language community and directions of dispersal

One method for locating the probable dispersal centre (or “homeland”) of a family of languages is distributional. The view that the most likely dispersal centre for a family of languages is in the area of their greatest genetic diversity was stated by Sapir (1916) some years before Vavilov applied a parallel idea to cultivated plants. More recently, its linguistic application has been substantially refined by Dyen (1956). Underlying it is the principle of parsimony, which requires that one posits the fewest and shortest moves needed to account for the geographic distribution of the subgroups. It can be seen that this method depends on subgrouping, because genetic diversity is defined in terms of subgroups.

One needs to look at both the external relationships and the internal subgrouping of the languages in question. We accept Blust’s (1978a, 1983-84, 1993) conclusions that Oceanic belongs to an Eastern Malayo-Polynesian group whose other members are found around Cenderawasih Bay in northwest New Guinea and in South Halmahera, and that the closest relatives of Eastern Malayo-Polynesian also lie in Indonesia. Given this, the method requires us to conclude that Proto-Eastern Malayo-Polynesian was probably located around Cenderawasih Bay and that the separation of the Oceanic branch began with a movement of Eastern Malayo-Polynesian speakers from Cenderawasih Bay eastwards along the north coast of New Guinea.

But where did this pre-Oceanic movement end up? That is to say, what was the dispersal centre of POc itself? The internal subgrouping outlined in the section on subgrouping allows for two main possibilities, both located in northwest Melanesia. (Map 2 gives the location of high-order subgroups of Oceanic.) One possibility is that pre-Oceanic speakers moved only as far as the Sarmi Coast or Jayapura, because here we find two small and little studied Oceanic groups which may be first-order branches of Oceanic.

But if, as the limited available evidence suggests, these groups belong to Western Oceanic, then POc developed further east, in the Bismarck Archipelago, where several fairly well-established high-order subgroups meet. In the Bismarck Archipelago we find two first-order groups of Oceanic, Admiralties and Western Oceanic, and perhaps a third, St Matthias. Western Oceanic in turn divides into three large second-order groups, two of them represented in New Britain: the North New Guinea group and the Meso-Melanesian group.

Another procedure for locating linguistic dispersal centres is often called the Wörter und Sachen (“words and things”) method, because it depends on reconstructing words for things associated with culture and physical environment. For POc, for example, it is necessary to reconstruct terms for a number of animals — cassowary, possum and bandicoot — that are found in New Guinea and at locations in the Bismarck and Solomon archipelagoes, but which are absent from southern Melanesia, Micronesia and Polynesia. These reconstructions are required not only because the cognate sets are represented in more than one primary subgroup of Oceanic but also because cognates are found in non-Oceanic Austronesian languages in eastern Indonesia (Blust 1982). It is also necessary to reconstruct words for many marine animals and other features associated with the sea. However, this line of evidence hardly allows any conclusion more specific than that POc speakers lived near the sea somewhere in eastern Indonesia, New Guinea or the Bismarck Archipelago.

Let us turn now to the break-up of POc. Primarily on the basis of distribution of subgroups, we have placed the likely dispersal centre in the Bismarck Archipelago, without favouring any particular part of this region. The distribution of first-order subgroups indicates a more or less contemporaneous movement of POc speakers in several directions over a considerable area of Melanesia. Assuming that POc was not located in the Admiralties (an assumption which needs further investigation) there was an early movement to that region, contemporaneous with the break-up of POc. The same can be said of the St Matthias group and New Britain, each of which contains a high-order subgroup of Oceanic. At about the same time there were one or more movements into the Solomons and into Remote Oceania, leading to the divergence of the Southeast Solomonic, North/Central Vanuatu, Southern Vanuatu, Southern Oceanic, Central Pacific and Nuclear Micronesian groups.

New Britain is the likely dispersal centre for the Western Oceanic group (Ross 1989). Western Oceanic is an innovation-linked subgroup, but the two innovations noted earlier are shared by the vast majority of its member languages, suggesting that it is descended from a dialect network which once occupied quite a limited area (otherwise the innovations would not have spread through almost all the network). The available evidence indicates that this area was along the north coast of New Britain, perhaps straddling the Willaumez Peninsula, as this is the seam of two primary divisions of Western Oceanic, the Meso-Melanesian and North New Guinea groups. Furthermore, the area of greatest diversity of the Meso-Melanesian group is found in this area. Two of its three primary subgroups (Bali-Vitu and the Willaumez chain) are situated around the Willaumez Peninsula.

If the Western Oceanic homeland was on the north coast of New Britain, Western Oceanic speakers expanded in two main directions. There was progressive occupation of the New Britain coast, of New Ireland (apparently from south to north) and of the Northwest Solomonic area to form the Meso-Melanesian group.[17] The second direction of expansion was across the Vitiaz Strait to the mainland coast of New Guinea. This last movement was soon followed by a split between the North New Guinea group, occupying parts of the north coast of New Guinea and offshore islands in the Morobe, Madang and Sepik Provinces, and the Papuan Tip group, comprising the Milne Bay and Central Province languages. Within a short time the ancestral North New Guinea network split into the Schouten, Ngero/Vitiaz and Huon Gulf groups. Sequences of reconstructible innovations particularly support the south to north settlement of New Ireland and the east to west settlement of the Schouten Islands.

The weight of the subgrouping evidence, then, points to a quite rapid diaspora of Austronesian speakers across Melanesia and into the central Pacific. We have noted distributional evidence for an approximately coeval occupation by Oceanic speakers of New Britain, St Matthias, the Admiralties, the Southeast Solomons and several regions of “Remote Oceania” (the widely separated island groups and isolated islands of the Pacific beyond the Solomons chain).

Can this dispersal be accurately dated? An association between the spread of Oceanic languages and Lapita pottery and associated artefacts in the central Pacific has been argued by a number of scholars since the 1960s, but Shutler and Marck (1975) were perhaps the first to argue that the break-up of POc itself should be associated with the movements of Lapita bearers, a view that has gained increasing acceptance (Pawley and Green 1984; Spriggs 1984, 1993; Kirch 1984, 1992; Bellwood 1989; Green 1994). Lapita pottery appears in the Bismarck Archipelago around 1600 BC (Kirch and Hunt 1988; Spriggs 1989, 1990, this volume) and in several regions of Remote Oceania soon after. Archaeological dates for the first Lapita assemblages in New Caledonia, Vanuatu, Fiji, Tonga and Samoa are all between 1300 and 1000 BC. Current archaeological dates indicate that settlement of the Papuan Tip area may not have occured until somewhat later, reaching Central Papua about 2000 years ago (Vanderwal 1973; Allen 1977a, 1977b; Bulmer 1982). At present the earliest dates which we can associate with confidence with North New Guinea speakers (on the north New Britain coast west of the Willaumez Peninsula, and on the islands and the mainland coast of the Vitiaz Strait) are only about 1500 years ago (Lilley 1988, 1990).

Rapidity of settlement bears on the conundrum of how Polynesians and Micronesians, with their Southern Mongoloid physical type, could have come out of Melanesia, where the so-called “Melanesian” physical type, marked superficially by tightly frizzy hair, dark skin and relatively large teeth, is allegedly dominant. The answer appears to be that they did not “come out of” Melanesia: some Oceanic speakers moved through Melanesia into the central Pacific, and they moved through rapidly enough to retain their Southern Mongoloid phenotype. Today’s Austronesian speakers in Melanesia have acquired Melanesian characteristics in varying degrees by intermarriage in the intervening millennia and by gene flow after the initial dispersal of Oceanic speakers.[18] However, even in western Melanesia, pockets of Austronesian speakers’ superficial Southern Mongoloid features are still found, e.g. in much of southeast Papua, and on Wuvulu and Aua. Recent work in biological anthropology (referred to in Green 1994:38) indicates that other Austronesian-speaking populations in Melanesia also exhibit genetic markers linking them with Southern Mongoloid peoples of Southeast Asia and Polynesia.

Modes of Oceanic dispersal

The fact that so many Oceanic subgroups as far east as Fiji form linkages rather than families suggests that within each major island group the main mode of settlement was a continuous expansion into neighbouring territory that resulted again and again in the formation of dialect chains and networks. Each new village maintained contact with its parent village, and only gradually did village speech communities differentiate out into separate languages. The main variation in this pattern was caused by factors which hindered contact, whether geographical (a large ocean gap or a land barrier such as a mountain range) or social (the intervention of a Papuan-speaking group). The Northwest Solomonic case is instructive. Choiseul, the New Georgia group, and Santa Ysabel are each occupied by linkages within which there are no radical breaks, yet the breaks between each network, corresponding with the intervening sea, are fairly clear. On the island of Bougainville, on the other hand, the Austronesian languages show considerable differences among themselves, a product of their separation by Papuan-speakers.

There are some, but not many, notable exceptions to this pattern. For example, within the North New Guinea group the Schouten chain is sufficiently different from its neighbours to the east, and its easternmost members are sufficiently conservative in their phonology and verbal morphology (Ross 1988: 122-132) to imply that Proto-Schouten was an early departure from the North New Guinea dialect network and became completely isolated from it.