This dissertation presents a new subgrouping of South Halmahera-West New Guinea (SHWNG) languages. The 38 SHWNG languages form a small, poorly known branch of Austronesian. The Austronesian family originated in Taiwan and later spread into Indonesia, across New Guinea, and to the remote Pacific. In New Guinea, approximately 3500 years ago, Austronesian speakers first came into contact with so-called Papuan languages -- the non-Austronesian languages indigenous to New Guinea, comprising more than 20 families. The Austronesian languages still extant from this initial spread into New Guinea fall into two branches: SHWNG and Oceanic. In great contrast to Oceanic, only a few SHWNG languages are well-described, and almost nothing has been reconstructed at the level of Proto-SHWNG. Contact with Papuan languages has given the SHWNG languages a typological profile quite different from their linguistic forebears.

Chapter 1 puts the SHWNG languages in context, describing their significance for Austronesian and their broader relevance to historical linguistics. It outlines the theoretical framework of the work, covering models of language diversification, diagnostic features for subgrouping, and language contact. A scale is proposed for ranking innovation types from most to least diagnostic for subgrouping. Morphological innovations are ranked above phonological innovations in this scale.

Chapter 2 gives an overview of the Austronesian family, focusing on the aspects most crucial to understanding the rest of the work: an outline of Proto-Austronesian phonology and the history of the branches ancestral to Proto-SHWNG.

Chapter 3 summarizes previous work on SHWNG languages, covering language membership, environmental and social characteristics, descriptive sources, shared innovations, subgrouping, reconstruction, and contact-induced change.

Chapters 4-6 are the main empirical contribution. Chapter 4 covers segmental sound change in 25 SHWNG languages and dialects. Chapter 5 covers tonogenesis in the Raja Ampat languages Ma'ya and Magey Matbat and the Cenderawasih Bay languages Moor, Yaur, and Yerisiam. Chapter 6 covers subject agreement and inalienable possessive morphology in 37 SHWNG languages and dialects. In these chapters, the goal is to identify shared innovations and determine their usefulness for establishing subgrouping relationships among SHWNG languages. Morphological innovations are found to be more diagnostic than phonological innovations, confirming the scale proposed in chapter 1.

Chapter 7 proposes a new subgrouping for SHWNG languages, synthesizing the results of chapters 4-6. The homelands of Proto-SHWNG and its branches are also discussed. The homeland of Proto-SHWNG is located in southern Cenderawasih Bay.

Chapter 8 concludes by considering the contributions of SHWNG languages to models of language diversification and change, and laying out questions for future research.

The Appendix contains the complete database of SHWNG cognate sets from which the analysis in chapters 4 and 5 is drawn.