Abstract
This thesis investigates echo vowels in Budai Rukai (an Austronesian language spoken in Taiwan) from an Optimality Theory perspective. An echo vowel is an epenthetic element which is identical to the vowel in the preceding syllable. In my observation, Budai Rukai has CVC in underlying forms. However, due to the prohibition that consonants can not appear in the coda position in surface form, an echo vowel must be inserted after the final coda consonant of a word. Two phenomena of echo vowels are discussed in this study: (i) in terms of form, an echo vowel is in general the same as the vowel in the preceding syllable. However, when the vowel in the preceding syllable is /a/, the echo vowel becomes /ə/; (ii) echo vowels can not be parsed in foot. Through a cross-linguistic comparison of echo vowels, this thesis assumes that echo epenthesis is formed by the interaction between phonological system and perception. So this thesis follows the framework of autosegmental constraints (Uffmann 2001, 2004, 2006) and P-map (Steriade 2008) to explain the behavior of echo vowels. First, the autosegmental constraints explain that the phonological system of Budai Rukai uses the vowel harmony strategy to insert echo vowels. However, when the preceding vowel is /a/, it will switch to the default vowel insertion strategy. Second, P-map explains that perception might play a role in the choice of the default vowel /ə/ when Budai Rukai using the default vowel insertion strategy. As for the phenomenon that echo vowels can not be parsed in foot, we can assume the constraint HEADDEP (i.e. any insertion vowel can not exist in foot) to predict the right form of stress assignment, glide formation, and whether long vowels are independent phonemes in words that have echo vowels.