Subgrouping (cladistics or family tree construction) plays a central role in the comparative method, especially in lexical reconstruction and the chronology of changes. Although the sound system of the ancestral language of a hetero- geneous family can, in principle, be reconstructed without reference to sub- grouping assumptions, this is not the case for the lexicon. Each phoneme of the proto-language has a determinable outcome or reflex in every daughter lan- guage in those morphemes that are retained from the ancestral stock. (Even if a phoneme is lost, that loss is detectable as a zero reflex, i.e. nothing, in the position where other languages have a positive reflex. For example, Wayan and Tongan zero correspond to Tagalog g, Malay r, and Sa'a 1 in the compari- sons given earlier.) Unlike the phonemes, not every morpheme or word of the proto-language is continued in each daughter language. Inferences about the antiquity of a word shared by two members of a family (a cognate set) depend on how closely the two languages are thought to be related, i.e. about the structure of the family tree.
Subgrouping also provides a relative chronology for changes in a speech tradition-whether these are generated internally or borrowed from other tra- ditions. Finally, by indicating the geographic centers of genetic diversity, a family tree may give strong clues about directions of language dispersal. Dyen (81) has developed a set of procedures for inferring the most probable disper- sal centers or "homelands" of language families or subgroups whose internal relationships are known, based on the principle of fewest moves.
At least five radically divergent hypotheses about the high order subgroups of AN have been proposed. Until the 1930s views on AN high order subgroups were ill-defined but most commentators (without providing any sound justifi- cation) spoke of four main branches-Indonesian, Melanesian, Micronesian, and Polynesian. The last three groups corresponded to the familiar geographic regions, and "Indonesian" (better labeled "Western AN") encompassed all the languages not in the other groups, including the languages of Formosa, the Philippines, and Madagascar. However, as we shall see, the status of the Melanesian languages was particularly controversial.