Untangling Leti infixation (Austronesian languages, morphology)

Leti(nese) is an Austronesian language spoken on the island of Leti, just e ast of Timor. Descriptions of Leti include Jonker (1932) and van Engelenhov en (1995a, 1996). In this paper, I focus on Leti infixation, a little-studi ed aspect of Leti morphology. In Leti, infixation yields nouns from verb ro ots. There are eight distinct phonological forms of the nominalization affi x: the three infixes -ni-, -n-, -i-; the three prefixes ni-, i-, nia-; the parafix i- + -i-; and a zero allomorph. Leti nominalizing infixation poses two serious problems of analysis. The first challenge is to properly predic t the distribution and shape of the eight allomorphs. A second problem is a ccounting for the fact that some of the sound patterns that result from inf ixation are exactly the opposite of those predicted by Optimality approache s like those of Prince and Smolensky (1993). In this paper I demonstrate ho w the eight allomorphs via phonological rules, with allomorph selection rel ated to verb class. There appears to be no phonological motivation for the treatment of /ni-/ as a prefix that has been shifted to infixal position du e to dominant phonological constraints. The positioning of /-ni-/ must be m orphologically specified, either in terms of an infixation rule or some con straint-based equivalent.