As already noted, the hypothesis that gamma-globulin (Gm) allotypes are associated with differential resistance or susceptibility to malaria is unsupported by any evidence. Is there then a clear dichotomy between non-Austronesian and Austronesian-speaking peoples in lowland New Guinea with Gm and other genetic markers? The answer, as given in detail by Serjeantson et al. (1983), is that there is not. For Gm, as for other genetic markers, geographic distance is a much stronger determinant of genetic affinity than is language, and this is true also for Markham Valley populations where Gm allotypes were originally proposed as sensitive indicators of Austronesian ancestry.

The human leukocyte antigen (HLA) profiles for populations in Melanesia support the notion that there is genetic diversity within non-Austronesian-speaking peoples that must have predated the overlaying of Austronesian elements in some coastal and island non-Austronesian-speaking peoples. A simple test is whether lowland coastal non-Austronesian-speaking people represent a hybrid pool of New Guinea Highlander and Polynesian genes, or whether they have their own separate genetic profile. The HLA-DR system shows that hybridity is not the case. For instance, DRB1^*1408 has a frequency of 18 per cent in New Guinea Highlanders but is absent from lowland non-Austronesian-speakers. The latter have some unusual genes, DRB1^*1104 and 1602, that are found neither in Highlanders nor Polynesians. The Polynesian gene pool, characterized by high frequencies of DRB1^*1201 and 0901 can be drawn from the east Asian but not the Melanesian gene pool.

In a perspective on evolution and population genetics in Papua New Guinea, Serjeantson et al. (1992) showed that populations cluster into four main groups on the basis of ABO blood group and other genetic markers. These are: 1. Austronesian-speaking Melanesians from the New Guinea Islands and Papuan south coast, 2. An and NAn-speaking populations from Madang and Sepik Provinces, 3. speakers of the New Guinea Highlands NAn languages, and 4. non-Austronesian-speakers from New Britain and Papua. The diversity of colonizers in Melanesia is confirmed by the study of mitochondrial DNA, which has shown that at least 18 surviving maternal lineages colonized Papua New Guinea, all deriving from Southeast Asia.

Austronesian settlements have rarely penetrated the hyperendemic malarial hinterland of New Guinea, but rather, are found on offshore islands and in coastal regions. It seems that early Austronesians may have arrived in Melanesia to find a malarious region inhabited by peoples comparatively well adapted to the environment. For many early Austronesians, it may have been prudent to continue east. It is no coincidence that the majority of Lapita sites are found on small islands.

In the remainder of this paper, we examine the closely-linked, highly polymorphic human leukocyte antigen (HLA) class II genes, HLA-DR and -DQ, for evidence of in situ evolution in some Austronesian-speaking populations of Oceania. We define the evolutionary forces of founder effect, genetic drift, mutation, migration and selection and look for evidence that these evolutionary forces shaped the genetic profiles seen in contemporary populations of Austronesian speakers.

The analyses are based on new data generated by using the polymerase chain reaction (PCR) technique to amplify the HLA-DRB1, DRB3, DRB5, DQA1 and DQB1 genes, followed by hybridization with sequence-specific oligonucleotides (SSOs). The PCR-based HLA typing protocol is rapid and sensitive, and looks directly at the gene of interest rather than at flanking regions of DNA as is often the case in restriction fragment length polymorphism (RFLP) analysis. Using the PCR approach, it has been possible to examine many more individuals in many more populations than was possible with traditional serological techniques or RFLP analysis. We have examined more than 2,600 chromosomes from Asia, Oceania and Australia and have identified 82 DR,DQ haplotypes (there are additional haplotypes in Caucasoids); this compares with less than ten DR haplotypes identified in early serological studies (Serjeantson et al. 1982) and with 32 DR,DQ haplotypes detected in RFLP studies (Kohonen-Corish et al. 1988). That is, subtle variation in HLA class II alleles, sometimes a single DNA base change, was not detected in less-sensitive serological and RFLP analyses. In the following, we examine these data for evidence of the impact of evolutionary forces.