第二语言学习者对目标语言语音的感知往往会遇到困难，这在日常生活中大量存在的噪音环境中更为突出。以往研究主要集中于句子和单词识别，较少涉及音位识别层面，尤其较少系统探讨噪音条件下第二语言的元音识别及其机制。本研究结合对不同语言经验背景群体比较和短期语音知觉训练，探讨以汉语为母语的个体对英语元音感知的特点，并从噪音掩蔽机制入手，探讨汉语母语者在噪音条件下英语元音感知困难的机制。研究一以比较不同语言经验背景的群体为主进行研究设计，考察了29名没有英语国家生活经验的汉语母语者（Chinese-native listeners in China，CNC）和12名在英语国家短暂生活（1-2年）的汉语母语者（Chinese-native listeners in the US.，CNU）以及12名英语母语者（English-native listeners，EN）在安静条件和不同类型噪音条件下英语元音识别的特点。结果表明，两组汉语母语者在所有条件下的元音识别平均成绩都显著差于英语母语者，尤其是对英语后元音的识别。CNU在安静条件和长时程语谱噪音条件下的表现与CNC没有显著差异，但在英语多人语音噪音条件下的表现显著好于CNC。这说明，1-2年的英语国家生活经验特异性的促进了汉语母语者在多人语音噪音条件的英语元音感知。研究一还第一次揭示，没有英语国家生活经验的汉语母语者在多人语音噪音中的掩蔽释放（相对于长时程语谱噪音）表现出负向的反转，表明其在多人语音噪音条件下的元音感知受到了信息掩蔽的影响。研究二仍然采用对不同语言经验背景群体比较的设计，在汉语和英语两种语音噪音条件下，进一步考察没有英语国家生活经验的汉语母语者在多人语音噪音条件下的英语元音感知特点与噪音信息掩蔽特点。通过比较10名CNC、12名CNU和11名EN的英语元音识别受英语多人语音噪音和汉语多人语音噪音信息掩蔽的情况，研究二进一步证实，CNC的英语元音识别受到英语多人语音噪音的信息掩蔽显著大于CNU，但两组人受汉语多人语音噪音的信息掩蔽没有显著差异。这说明，汉语母语者的抗信息掩蔽能力确实由于1-2年的英语国家生活经验获得了提高，但这种提高特异于英语多人语音噪音条件。为了进一步考察信息掩蔽与噪音条件下英语语音感知经验的关系，研究三对三组没有英语国家生活经验的汉语母语者分别进行了总量为6个小时的三类语音知觉训练，包括（1）安静条件下英语元音识别（安静训练组）；（2）多人语音噪音条件下英语元音识别（噪音训练组）；（3）无字幕美剧观看（看美剧组）。两个元音识别训练组训练的元音为美国英语中的5个后元音。三组被试英语熟练水平、英语学习时间与起始年龄、前测各项任务表现匹配良好。研究结果表明，训练前被试安静条件下的元音识别成绩与信息掩蔽大小没有显著相关，但与动态噪音掩蔽的释放显著相关；训练后，结果显示，噪音训练组被试受到信息掩蔽的影响在训练之后显著降低，安静训练组被试受到的信息掩蔽没有显著变化，两组被试动态噪音条件下英语元音识别的掩蔽释放没有显著区别，说明噪音条件下的训练提高了汉语母语者英语元音感知过程中抗信息掩蔽的能力。综合这些结果可以认为，信息掩蔽的大小可能主要取决于处理噪音的能力，而动态噪音掩蔽的释放与音位加工能力有关。对训练前数据的分析表明，对元音共振峰辨别（VFD）的阈限与多人语音噪音条件下的元音识别及信息掩蔽有关；训练后被试的VFD阈限显著降低，对元音音长线索的依赖也相应减弱。可见，短期的知觉训练可以有效改变二语学习者语音知觉加工的线索利用水平。综上，本研究表明，在缺乏英语宏观语言环境经验的情况下，汉语为母语个体对英语元音的感知存在困难，尤其表现为对后元音的感知困难、在噪音条件下的英语元音感知困难。对元音音长的依赖和共振峰感知的敏感性不足显著影响汉语为母语个体英语元音感知的准确性。特异于英语的信息掩蔽干扰是造成汉语为母语个体在语音噪音下英语元音识别问题的重要原因。目标语（英语）宏观语言环境经验和非目标语宏观语言环境下的密集目标语噪音处理经验都可以显著提高二语学习者元音感知的抗信息掩蔽干扰能力。

Speech perception of the second language is difficult for the non-native listeners, especially in the natural noisy condition. While the existing studies mainly focused on word and sentence recognition, fewer studies examined the issue on the phoneme level, systematically study into the characteristic and mechanism of vowel perception by the second language listeners was especially scarce. This dissertation examined the characteristic of non-native’s vowel identification through comparing listeners with different language background and short-term speech perceptual training, furthermore, explored the difficulty of non-natives’ vowel perception in noise from the point of masking mechanism.In study 1, based on the design of comparing listeners with different language background, we examined the English vowel identification scores under babble noise, long-term speech-shaped noise (LTSSN) and quiet conditions between the 29 Chinese-native listeners in China (CNC), 12 Chinese-native listeners in the US (CNU) and 12 English-native listeners (EN). Results showed that CNC and CNU performed worse than EN in all conditions, especially for the back vowels. Furthermore, CNU outperformed their contemporaries in China (CNC) in the babble noise condition but not in the LTSSN and quiet conditions. It was interpreted that the non-natives’ ability of coping noise were improved by the experience of 1-2 years’ English country residence, while the improvement was restricted to the babble noise condition. In addition，we discovered for the first time that the CNC turned out to be a negative masking release from babble (compare with long-term speech-shaped noise), which indicated an informational masking effect. Study 2 examined the characteristic of vowel identification and informational masking by comparing the vowel identification between10 CNC, 12 CNU and 11 EN, under English babble noise and Chinese babble noise. Results showed that CNU listeners were more affected by the informational masking of English babble than CNC, because of their 1-2 years English native country residence, but no significant difference for Chinese babble between the two Chinese groups. The results suggested that the Chinese-natives’ suffering from informational masking was released by the 1-2 years’ English country residence, which was specific to the English babble noise.We employed short-term speech perceptual training in study 3 to further examine the characteristic of vowel identification and informational masking of Chinese-native listeners’ English vowel perception. Three CNC groups accepted different training task for six hours in total: (1) vowel identification in quiet, (2) vowel identification in noise, and (3) American TV series watching without subtitles. The vowels used in vowel identification training task for group 1 and group 2 were five American back vowels. The analysis of the pretest data showed that the vowel identification accuracy under quiet condition bear no significant correlation with the amount of informational masking, but significant with the amount of masking release from fluctuating noise. Results of speech perceptual training showed that the Chinese-native listeners’ informational masking was significantly reduced by English vowel identification training in noise, but not by training in quiet, which suggested that the Chinese-native listeners’ suffering from informational masking of babble in vowel identification were relieved by speech perceptual training in noisy background. Based on these results above, it can be thought that the amount of informational masking mainly depended on the ability of noise coping, while the amount of masking release from fluctuating noise was about phonemic processing. The results of pretest analysis showed a significant correlation between the amount of informational masking and the threshold of vowel formant discrimination (VFD), the threshold of VFD and duration cue usage significant decreased after training, which suggested that usage of acoustic/phonemic cue can be change by short-term speech perceptual training.Our studies suggested that the Chinese-native listeners have obvious difficulty in English vowel perception，especially for the back vowels, in the case of lacking target speech (English) environment. The accuracy of English vowel perception by the Chinese-native listeners is significantly affected by the perceptual weighting of duration cue and the sensibility to the change of vowel formants. Informational masking is the key mechanism contributing to the difficulty of English vowel perception in babble noise of the Chinese-native listeners. Both the experience under target speech (English) environment and the intensive processing experience of target speech noise under non-target speech environment can boost the ability of resistant to informational masking in the process of English vowel perception by listeners who learn English as their second language.