英文摘要：

This thesis investigates how Taiwan Mandarin speakers perceive nasal codas. TM nasal codas, /n/ and /ŋ/ are reported to be undergoing merging in non-low vowel contexts (/ə/N and /i/N), as opposed to the low vowel context (i.e., /a/N). Specifically, in the /i/N context, /n/ is generally merged into [ŋ]; yet, there is a sub-pattern merging into [n] in Southern Taiwan. In the /ə/N context, however, northern and southern TM speakers overall merge /ŋ/ into [n] (Fon et al., 2011). While the contrast in the place of articulation fades away, previous studies have shown that TM speakers maintain the contrast by the degree of nasalization of the pre-nasal vowels and by the nasal coda duration (Chiu et al., 2020; Chuang, 2017).

TM participants’ reliance on formant transitions in perception was tested in Experiment 1. Results showed that (i) participants reached a ceiling effect (M= 97.2%) in accuracy with the shortest response time (M= 1038.9 ms) in the /a/N context, that (ii) a higher accuracy for [in] compared to [iŋ] and a higher accuracy for [əŋ] compared to [ən] were observed, and that (iii) a lower accuracy for [in] compared to [iŋ] for some participants from Southern Taiwan.

Experiment 2 was conducted to examine the relevance of nasal coda duration in nasal perception by manipulating the rime/nasal duration. Results suggested that, in merging contexts [ən] and [iŋ], participants’ nasal coda identification was indeed biased to different degrees by different rime/nasal durations, while in non-merging contexts [əŋ] and [in], their perception relied mainly on the vowel transitions.