A Note on Higher-Order Subgroups in Oceanic

Abstract

The Oceanic (Oc) subgroup is remarkable for at least two reasons. First, it contains nearly half of the more than I,ooo Austronesian languages, making it the largest well-defined linguistic subgroup in the world, with the possible exception of Broad Bantu (Ruhlen 1987).' Second, it combines great lexical and typological diversity with particularly clear evidence of exclusively shared innovations. There is, however, a third way in which the Oc subgroup is perhaps equally remarkable. Despite the relatively clear evidence for Oc itself, which has been known at least since Dempwolff (1927), the higher-level subgroups of Oc have been extremely elusive. Pawley and Green (1984:136) present a tree diagram of Oceanic languages that has a rake-like structure rather than a hierarchical nesting structure: there are 29 primary branches, in many cases with little internal ramification. They stress that a tree of this shape implies the rapid spread of a colonizing population, and observe that a similar inference is independently reached from the radiocarbon dating of Lapita pottery, which first appears in western Melanesia about 3600 BP, and in Fiji and western Polynesia only three or four centuries later. In a landmark publication, Ross (1988) recognizes several large subgroups of Oc languages in western Melanesia. He calls these the North New Guinea Cluster, the Papuan Tip Cluster, the Meso-Melanesian Cluster, and the Admiralties Cluster, and he notes that the languages of the St. Matthias Islands (Mussau-Emira and Tenis) stand apart from these large groups, but may be more closely connected with the Admiralty Cluster than with the others. At the conclusion of this work he then proposes a set of exclusively shared innovations that link the first three groups into a supergroup exclusive of the Admiralty Cluster.