I describe and analyse data from Amarasi, a language with morphological consonant-vowel metathesis. Depending on the phonotactic structure of the stem to which it applies, metathesis is associated with a number of other phonological processes including: vowel deletion, consonant deletion and two kinds of vowel assimilation. By proposing that Amarasi has an obligatory CVCVC foot in which C-slots can be empty all these phonological processes can be derived from a single process of metathesis and one associated morphemically conditioned process. I consider analyses other than the rule-based one adopted in this paper and show that they cannot account for all the data in a consistent, plausible way.

Acknowledgements vii Abbreviations and symbols ix Speech varieties listed in the dictionary xi † 'Meto' is used when a form is taken from Jonker (1908) who does not specify which variety of Meto his data is from. ROTE-METO COMPARATIVE DICTIONARY xii Speech varieties (with at least four occurrences) in out-comparisons Lect Region Sub-region ISO Glottocode

I make a bottom-up reconstruction of the historical phonology and a portion of the lexicon of the Rote-Meto languages of western Timor. The regular sound correspondences of these languages necessitate reconstruction to Proto-Rote-Meto of a large amount of material that cannot be fully explained by Austronesian inheritance. The nature of this material indicates that it is substrate retention from pre-Austronesian languages of the region. This substrate can be detected by application of the comparative method alone. The bottom-up reconstruction also provides evidence for subgrouping Meto with both West Rote and East Rote and I propose that this is because Meto has shared a period of common development with both groups that involves a synthesis of the tree model and wave model of historical linguistics. 5. Jonker (1908) does not record the phonetic differences in the realization of voiced plosives. No variety of Rote is known to contrast plain voiced stops and imploded voiced stops.

Two parallel sets of sound correspondences are attested in the historical phonology of the Uab Meto (also known as Dawan[ese], Timorese, Atoni) language/dialect cluster. A top-down approach to the data reveals one set of regular sound correspondences in reflexes of Proto-Malayo-Polynesian lexemes, while a bottom-up approach to the data reveals another set of regular correspondences in lexemes for which no Malayo-Polynesian origin has yet been found. I examine each set of sound correspondences in detail and propose a framework for addressing the apparently contradictory data.

Questions have been raised about the precise genetic affiliation of the Enggano language of the Barrier Islands, Sumatra. Such questions have been largely based on Enggano's lexicon, which shows little trace of an Austronesian heritage. In this paper, I examine a wider range of evidence and show that Enggano is clearly an Austronesian language of the Malayo-Polynesian (MP) subgroup. This is achieved through the establishment of regular sound correspondences between Enggano and Proto-Malayo-Polynesian reconstructions in both the bound morphology and lexicon. I conclude by examining the possible relations of Enggano within MP and show that there is no good evidence of innovations shared between Enggano and any other MP language or subgroup. In the absence of such shared innovations, Enggano should be considered one of several primary branches of MP.