| This study investigated English speakers' perception of Mandarin sounds, aiming to answer three major questions. First, does the perceived similarity between Mandarin and English sounds affect learners' acquisition of Mandarin sounds? Second, do the groups with different amounts of Mandarin experience differ in their perception of Mandarin sounds? Third, is there a difference between the perception of consonants, vowels, and tones? To answer these questions, three experiments were conducted. Experiment 1 assessed participants' perceived similarity of Mandarin sounds to English. Experiment 2 evaluated their discrimination of Mandarin syllables that differed by different parts of speech. Experiment 3 tested the learner groups on their categorization of L2 sounds with L2 symbols. The participants included three groups of English speakers: 17 experienced learners of Mandarin (Ex group), 18 inexperienced learners (Inex group), and 18 participants with no Mandarin experience (Noex group).;The results showed that the Noex group tended to differ extensively from the learner groups in the L2-L1 assimilation, suggesting that Mandarin training may alter learners' perception of L2 sounds in reference to their L1. As for the L2 discrimination, the Noex group performed worse than the learner groups on some contrasts, while the two learner groups never differed in their accuracy. This suggests that the Inex learners could distinguish most Mandarin contrasts as well as the Ex learners. However, all three groups performed poorly on the Tone 2-Tone 3 contrast, showing that this tonal distinction may remain difficult even after years of learning. The L2-L1 assimilation patterns failed to predict L2 discrimination accuracy in most cases, however, the assimilation seemed to be a good indicator of a learning effect. Finally, Experiment 3 showed that the Ex group categorized L2 sounds more accurately than the Inex group, particularly with tonal targets, suggesting that the mastery of tones may require more extensive L2 exposure. The results also largely conformed to the predictions of the Speech Learning Model (SLM), that only the L2 tones perceived to be dissimilar from the L1 showed a learning effect. Such findings support the extension of SLM to the acquisition of suprasegmentals. |
