抄録
The glottal stop in the Amis language has largely been treated as a phoneme. However, in this paper, I argue that it should not be considered a phoneme for the following two reasons. First, the insertion rules predict the occurrence of glottal stops at root boundaries or between two vowels. A hiatus breaker can appear, which can be [w], [j], or a glottal stop, wherever there is a hiatus. There are two types of insertion rules for a hiatus breaker, depending on whether the hiatus position is in root-internal or root-external position. Second, glottal stops that are adjacent to a consonant can be analyzed as allophones of another phoneme, the epiglotto-pharyngeal stop phoneme. The epiglotto-pharyngeal stop phoneme used to be confused with the glottal stop, but they have conclusively been shown to have different sound qualities. Based on an analysis of the distribution of glottal stops and epiglotto-pharyngeal stops, I argue that glottal stops adjacent to a consonant and epiglotto-pharyngeal stops elsewhere are in complementary distribution and thus are allophones of the same phoneme. Reconstructed forms of certain words also support my analysis; some glottal stops are reconstructed as *q, which in Amis is basically reflected as the epiglotto-pharyngeal stop. This means that some glottal stops could originally have been epiglotto-pharyngeal stops. Previous studies have attested several examples that are not in line with the present analysis. However, based on the data gathered during fieldwork, they do not contradict my analysis. Two words with epiglotto-pharyngeal stops in the position adjacent to a consonant were identified, that pose a problem with my analysis. Further data and investigation are needed to show what makes them exceptional.
