This dissertation is a description of the grammar of Belep [yly], an Austronesian language variety spoken by about 1600 people in and around the Belep Isles in New Caledonia. The grammar begins with a summary of the cultural and linguistic background of Belep speakers, followed by chapters on Belep phonology and phonetics, morphology and word formation, nouns and the noun phrase, verbs and the verb group, basic clause structure, and clause combining. The phonemic inventory of Belep consists of 18 consonants and 10 vowels and is considerably smaller than that of the surrounding languages. This is due to the fact that Belep consonants do not contrast in aspiration and Belep vowels do not contrast in length, unlike in Belep’s closest relative Balade Nyelâyu. However, like-vowel hiatuses—sequences of heterosyllabic like vowels—are common in Belep, where the stress correlates of vowel length, intensity, and pitch do not generally coincide. Belep morphology is exclusively suffixing and fairly synthetic; it is characterized by a large disconnect between the phonological and the grammatical word and the existence of a number of proclitics and enclitics. Belep nouns fall into four noun classes, which are defined by their compatibility with the two available (alienable and inalienable) possessive constructions. Belep transitive verbs are divided into bound and free roots, while intransitive verbs are divided between those which require a nominative argument and those which require an absolutive argument. While the surrounding languages have a split-ergative argument structure, Belep has an unusual split-intransitive nominative-absolutive system, with the further complication that transitive subjects may be marked as genitive depending on the specificity of the absolutive argument. Belep case marking is accomplished through the use of cross-linguistically unusual ditropic clitics; clitics marking the function of a Belep noun phrase are phonologically bound to whatever element precedes the noun phrase. In general, Belep lacks true complementation, instead making use of coordinate structures with unique linkers as a complementation strategy.