We analyzed 375 base pairs (bp) of the first hypervariable region (HVS-I) of the mitochondrial DNA (mtDNA) control region and intergenic COII/tRNALys 9-bp deletion from 47 Karkar Islanders (north coast of Papua New Guinea) belonging to the Waskia Papuan language group. To address questions concerning the origin and evolution of this population we compared the Karkar mtDNA haplotypes and haplogroups to those of neighbouring East Asians and Oceanic populations. The results of the phylogeographic analysis show grouping in three different clusters of the Karkar Islander mtDNA lineages: one group of lineages derives from the first Pleistocene settlers of New Guinea-Island Melanesia, a second set derives from more recent arrivals of Austronesian speaking populations, and the third contains lineages specific to the Karkar Islanders, but still rooted to Austronesian and New Guinea-Island Melanesia populations. Our results suggest (i) the absence of a strong association between language and mtDNA variation and, (ii) reveal that the mtDNA haplogroups F1a1, M7b1 and E1a, which probably originated in Island Southeast Asia and may be considered signatures of Austronesian population movements, are preserved in the Karkar Islanders but absent in other New Guinea-Island Melanesian populations. These findings indicate that the Karkar Papuan speakers retained a certain degree of their own genetic uniqueness and a high genetic diversity. We present a hypothesis based on archaeological, linguistic and environmental datasets to argue for a succession of (partial) depopulation and repopulation and expansion events, under conditions of structured interaction, which may explain the variability expressed in the Karkar mtDNA.