We are not concerned here with discrimination using anthroposcopic or anthropometric characters. In passing, however, it should be noted that such studies, particularly those for teeth and fingerprints, can be very informative. Our own surveys have been restricted to traits detectable in samples of blood under simple genetic control by loci on many different chromosomes. The enzyme and other protein systems used, and their distributions in Pacific populations, have been reviewed recently by Kirk (1989), and in more detail for Papua New Guinea by Kirk (1992).
The data can be analysed in two ways. Some genetic differences are unique to certain populations and their patterns of distribution suggest common ancestry. In addition, variations in frequency of genetic factors can be subjected to multivariate analysis to give “genetic” distances between populations. These distances can be used to construct evolutionary trees by a number of methods including cluster and maximum likelihood analysis, or by principal component analysis, to obtain the distribution patterns of populations.
Kirk (1992), reviewing previous studies of the distribution of unique alleles in the western Pacific, recognized three patterns relevant to understanding the relationships between linguistic and genetic differentiation. The first of these patterns, the “Australoid”, is associated with the transferrin allele Tf*D1 and the GC*1A1 allele of the vitamin D-binding protein system. The second, or “Proto-Papuan”, is characterized by alleles such as PGM1*3, PGM2*9, PGM2*10, PGK*4 and MDH*3. None of these alleles is found in Australia, suggesting that they were brought to, or originated in, New Guinea after the separation of New Guinea and Australia at the end of the Pleistocene, 8-10,000 years ago. These “Proto-Papuan” alleles all have relatively high frequencies in the Papua New Guinea Highlands and in parts of Irian Jaya, with lower frequencies in New Guinea coastal areas and even lower frequencies in the Solomons, Banks Islands and Polynesian Outliers.
The third pattern is “Austronesian”. Alleles in this group are not found in Australia and rarely in the Papua New Guinea Highlands. They have their highest frequencies in the Solomons, Polynesian Outliers, Banks Islands, some coastal areas in the north and east of Papua New Guinea, the western Carolines and Fiji. These alleles include PGM1*7, PGK*2, probably HB*Tongariki, Albumin*NG, GPT*3 and GPT*6.
Previous studies, reviewed by Kirk (1986, 1989, 1992), have shown discrimination between Waskia (NAn) and Takia (An) on Karkar Island. However, for 17 other populations in the north coastal regions of Papua New Guinea the An speakers are not clearly differentiated from NAn speakers. In these cases more detailed analysis shows that geographic location is more important than linguistic division (Serjeantson et al. 1983).
Nevertheless, consideration of populations over a wider geographic area, including many from the Highlands of Papua New Guinea and others from coastal areas and other parts of the western Pacific, show that language is an important discriminant, with the exception of the Mailu in southeast Papua. The exceptional position of the Mailu is due probably to the incorporation of An genetic components from neighbouring populations into a group which continues to speak a NAn language (see Kirk 1992 for further details).
In a detailed comparison of genetic distances between An-speaking Indonesian and other western Pacific populations, Sofro (1982) has shown that the Indonesian populations, including Ternatens and Galelarese from Halmahera whose languages are NAn, cluster with the An-speaking populations of New Guinea and elsewhere in the Pacific but are distinctive from the NAn-speaking populations both in Papua New Guinea and Irian Jaya.
To examine further the question of linguistic and genetic relationships in the Pacific area we have used more recently developed statistical procedures to re-analyse some of our previous data, and have included some populations for which new genetic marker information is now available. Multi-locus allele frequency data were used to estimate the phylogeny of two population groups, using a partial maximum-likelihood method (Felsenstein 1981).
This method has been shown by Kim and Burgman (1988) to be more accurate than the more commonly used unweighted pair-group arithmetic average clustering (UPGAA) method of estimating phylogeny from allele frequency data, particularly when a relatively small number of loci are analysed and where rates of evolution may vary among populations. The maximum-likelihood approach results in an estimate of the evolutionary history of a group of populations in the form of a maximum-likelihood network (or tree) connecting them. The reliability of the estimate can be tested by comparing the “likelihood” of the maximum-likelihood network with that of other networks connecting the same populations through different patterns of branching.
The first group of populations analysed consisted of the same 17 populations referred to above, investigated by Serjeantson et al. (1983) and located in the Bogia District and Gogol Valley in northern coastal Papua New Guinea, and on the adjacent Manam, Karkar and Siassi Islands. These populations include both An and NAn speakers. The second group consists of An and NAn speakers from various localities on New Guinea and from throughout Indonesia, Island Melanesia, Micronesia and Polynesia (Map 1).
