The lexicostatistical study of Dyen (1965) revealed significantly more diversity in the An languages spoken in Melanesia than among those spoken further west in Indonesia and Malaysia. It is now recognized that this diversity reflects heavy borrowing from the NAn languages which were significantly diversified at the time of Austronesian settlement.
Most of the An borrowing appears to have been from the smaller NAn phyla, with little influence from the two major NAn phyla, Trans New Guinea and Sepik-Ramu. The geographical distributions of these two phyla only overlap with the Austronesian speakers on mainland New Guinea, and that also marginally. Besides, the speakers of languages belonging to these phyla have only recently expanded into their present areas of distribution. The Highlands migration of the Trans New Guinea Phylum languages is considered to have begun around 5,000 to 2,000 years ago. The occupation, west to east, of the coastal areas of Sepik and Madang provinces by Sepik-Ramu speakers, who are essentially a riverine people, is much more recent a phenomenon. Investigation of the An/NAn dichotomy in Melanesia therefore must take into consideration this diversity of NAn languages and the extent to which it has influenced the An substratum.
To evaluate the relationships among An and NAn speakers on a wider scale we selected for analysis representatives of three different NAn phyla, namely, the North Halmahera Stock of West Papuan Phylum (Ternatens and Galelarese), the Iatmul of the Sepik-Ramu Phylum, and speakers of languages in the Trans New Guinea Phylum. In view of the extensive diversification of Trans New Guinea Phylum languages, we selected one population each from five different regions of New Guinea: the north coast (Pila), the northern Highlands-fringe (Gainj), the Highlands proper (South Fore), the southern Highlands-fringe (Pawaia) and the south coast (Asmat). In addition, two populations were added from the islands off the coast of New Guinea (Waskia and Kovai). We also included widely distributed An speakers (Figure 1).
Figure 2. Maximum likelihood network connecting representative Austronesian and non-Austronesian-speaking populations from Indonesia, Melanesia, Micronesia and Polynesia. Branch lengths are drawn in proportion to genetic distance.
The linguistic diversity among NAn speakers is reflected by their genetic diversity (Figure 2). Differences in the branching order between Figures 1 and 2 for those populations represented in both are due to slight differences in the data used in the two analyses.
The Ternatens and Galelarese from Indonesia (both belong to the West Papuan Phylum) are closely related to the majority of An speakers. They are quite distinct from the other NAn groups, which are themselves loosely clustered with no hierarchical structure to their relationships. It appears that the Trans New Guinea Phylum speakers failed to homogenize the genetic diversity underlying the linguistic substructure already in place at the time of their arrival.
Among the An populations, Ham clusters well within the NAn populations indicating that they have acquired an Austronesian language from outside. With the exception of Tolai and Buka (see below), the remaining An populations are relatively tightly clustered. This indicates close genetic affinities, despite the populations being widely distributed geographically. The pattern is consistent with these populations having spread rapidly and recently to occupy their present location.
It is now accepted that the Bismarck Archipelago was home to the progenitors of Lapita cultures, although an opposing view suggests that these cultures arrived fully formed into the region. Supporters of both these views, however, agree that the islands of the Archipelago were central to the spread of Lapita people further east into the Pacific. Unfortunately, the genetic data on the contemporary Bismarck populations are scanty and the above argument is unlikely to be resolved without some information from New Ireland and New Britain. Genetic data on populations surrounding the Bismarck Archipelago are available, and it may be argued that the Bismarck populations are unlikely to be very different from these neighbouring populations. We argue, on the one hand, that the populations settled on both sides of St George’s Channel, between New Britain and New Ireland, would more likely to have been influenced by the Lapita movement than those surrounding the Vitiaz Strait, between New Britain and New Guinea, if the colonizers of Polynesia originated in Southeast Asia and largely bypassed Melanesia. On the other hand, if the Lapita populations did develop entirely in the Bismarck Archipelago then one might expect greater genetic homogeneity among populations in the region of the Bismarck Sea.
To test these two opposing hypotheses we have included in the analyses populations bordering St George’s Channel and the Vitiaz Strait. The Tolai occupy the western end of St George’s Channel, whereas the Buka are located further east. For the Vitiaz Strait we have selected two NAn-speaking populations, the Kovai from Umboi Island and Waskia from Karkar Island.
The analysis reveals that the four Papua New Guinean island populations do not join either the group composed of Polynesians, Micronesians and Indonesians, or the remaining NAn populations (Figure 2) occupying an intermediate position in the network between the two. Waskia and Kovai share a common branch in Figure 2. Tolai and Buka branch quite separately and distinctly, but are closer to the remaining An populations. There is thus no homogeneity among the populations surrounding the Bismarck Sea and greater affinity between the main group of An populations and those adjacent to St George’s Channel consistent with the hypothesis of a movement of Lapita culture through the region.