元音感知是第二语言（L2）学习的难点，在理论研究与教育实践中都得到很多关注。已有研究发现，学习者在二语元音感知上存在很大的个体差异。深入理解这一个体差异的实质，有助于推进对二语语音学习机制的认识，对提高二语语音感知的学习成效也有启示。

已有研究采用多个经验相关的因素来解释二语元音感知的个体差异，如二语学习起始年龄、熟练度、使用时间等。然而，在这些因素比较接近的情况下，二语学习者仍然在二语元音感知上存在显著的个体差异。为解释这样的差异，基于语音刺激的正确感知取决于语言表征的完善程度，有些研究尝试用音位类别的建立程度来解释音位感知水平。其中，最具代表性的是范畴感知研究。这类研究用辅音范畴化来理解辅音感知的难易。

然而，与辅音间的区分特点不同，元音间因声学属性第一共振峰（First Formant; F1）和第二共振峰（Second Formant; F2）重叠较大而无法范畴化。因此，已有研究用“相似性”来定义二语元音的类别建立，认为跨语言语音间越相似，二语语音越容易被同化到母语元音类别和被识别，反之，则不易识别。然而，根据相关文献，除跨语言语音相似性之外，二语对比元音类别间的语音差异也显著影响二语元音感知。

为更好地解释二语元音感知的个体间差异，本论文提出“二语元音类别建立的独特性”的概念，将“跨语言语音类别间的分化程度”和“二语对立语音类别间的分化程度”整合起来，并将“二语元音类别建立的独特性水平”作为二语元音类别建立的测量指标。在此理论基础上，本论文以大学阶段的母语为汉语的英语学习者为研究对象，结合与英语母语者的组别对比、汉语母语者的个体差异对比、训练追踪等设计，在三项研究中系统探讨汉语母语者英语元音类别独特性的建立及其与英语元音识别能力的关系，进一步完善二语语音感知的理论模型，探讨二语元音类别建立的独特性与二语元音感知的联系。

研究1提出二语元音类别建立独特性的概念及其测量方法，进而对汉语母语者，探讨其英语元音类别建立的独特性。研究1a基于跨语言（汉语母语者46名、英语母语者9名）英语元音识别表现的比较，和对汉语母语者英语元音混淆矩阵的分析，考察汉语母语者英语元音类别建立的特点。结果表明，汉语母语者英语元音的识别成绩显著地低于英语母语者。汉语母语者对英语元音/i/和/I/的识别率较低、混淆率较高，混淆模式具有不对称性（将/i/混成/I/，而未将/I/混成/i/）。基于这些结果，将英语元音/i/-/I/确立为后续汉语母语者英语元音类别建立独特性研究的分析对象。

在研究1b中，首先建构了步长为0.2 个单位等效矩形带宽值（Equivalent Rectangular Bandwidth, ERB）的F1 x F2空间中的元音变体刺激矩阵，测量了9名英语母语者和31名汉语母语者的英语学习者，在英语元音感知分类任务和等级评定任务上的表现。根据被试的英语元音感知分类成绩，绘制了英语元音感知分类图，并根据英语元音感知分类成绩和等级评定成绩确定了最有代表性的元音变体刺激项目，继而确定了英语元音原型（最具元音类别代表性的范例），再以英语元音和汉语相似元音间的原型距离、以及英语对比元音间的原型距离量化了汉语母语者的英语元音类别建立的独特性水平。然后，基于跨语言英语元音原型位置分布比较和英语元音感知分类模式的个体差异分析，考察了汉语母语者的英语元音类别建立的个体分化特点。结果发现，汉语母语者的英语元音类别建立存在相当大的个体差异，表现为区分良好、混淆重叠、/i/主导和/I/主导四种感知分类模式。这提示，与母语者的元音类别建立不同，二语者对二语元音类别的建立是高度个性化的。

研究2结合汉语母语者英语元音类别建立的独特性、以及多个英语经验特征与英语元音识别的关系，考察相似教育背景汉语母语者的英语元音感知的影响因素。通过分析汉语母语者在不同感知分类模式下的英语元音识别表现，还考察了汉语母语者对英语元音/i/-/I/类别建立与其元音识别能力的关系。结果表明：（1）汉语母语者的英语元音类别建立的独特性水平与其英语元音识别显著相关。英语元音/i/和/I/间的原型距离、英语元音/i/和汉语元音/i/之间的原型距离都与汉语母语者的英语元音平均识别表现显著相关（r = 0.350, p = 0.02; r = - 0.380, p = 0.01）。在进一步控制英语元音/i/和汉语元音/i/间的原型距离后，英语元音/i/和/I/间的原型距离仍然独立解释了英语元音识别13.8%的变异。（2）汉语母语者的英语元音识别水平与其所处的英语元音感知分布模式存在紧密联系，具体表现为：区分良好模式下的汉语母语者对英语元音类别/i-I/的分类最好，其英语识别成绩最高（46%）；重叠混淆模式下的汉语母语者对英语元音类别/i-I/的分类最差，其英语识别成绩最低（35%）；对英语元音对立/i-I/中一个元音的类别建立占优势的/i/主导和/I/主导两种模式，其英语识别成绩处于中等水平（41%, 38%）。（3）汉语母语者的英语学习起始年龄、每周英语使用时长、英语听力理解水平、英语语言熟练水平与其英语元音识别表现相关不显著（- 0.236 < r < 0.339, p > 0.05）。这可能与研究被试为教育背景相似的大学生，英语学习经验和熟练性水平变异较小有关。以上结果表明，即使汉语母语者的二语学习经验相似，二语元音类别建立的独特性仍能显著地解释二语元音感知的个体差异。这就提示二语元音类别建立的独特性，是衡量二语元音感知水平的一个有效指标。

为进一步探讨针对性语音经验对二语元音类别建立独特性的影响，研究3开展了高强度、外显的实验室训练，对14名汉语为母语的英语学习者实施了连续6天，每天1小时的英语元音感知训练，考察了英语元音感知训练对被试英语元音类别建立独特性水平的影响。研究采用随机化对照设计，把被试随机分至训练组（n = 14）和控制组（n = 11）。训练组接受对于英语元音/i/-/I/的识别训练，控制组则观看同等时长的无字幕美剧。两组被试训练前在生理年龄、英语学习起始年龄、每周英语使用时间、英语听力理解、英语语言熟练度、以及英语元音识别水平和英语元音类别建立独特性水平上均没有显著差异。测试共三次，包含前测，无延迟后测，以及一个月后的延迟后测。测试任务包括英语元音识别任务、英语元音感知分类任务及典型性评价任务、以及汉语元音感知分类任务及典型性评价任务。结果显示：（1）二语元音感知训练显著增加了学习者英语元音/i/与/I/间的原型距离，对汉语元音/i/与英语元音/i/间的原型距离没有显著影响。（2）训练后，训练组被试对英语元音/i/-/I/的感知分类模式分布比例有显著变化。其中混淆重叠模式人数占比从14.3%减少到7.1%，区分良好模式人数占比从0%显著增加到7.1%，/i/主导模式人数占比从64.3%显著增加到78.7%，/I/主导模式人数占比从21.4%显著下降到7.1%。（3）英语元音/i/和/I/的识别成绩都在训练后显著提升，并保持到一个月后。（4）未训练过、但与英语元音/I/高度混淆的英语元音/e/的元音识别成绩，在训练后没有显著提升，提示特定类别元音感知训练的迁移效应不显著。以上结果表明，二语元音训练显著地提高了汉语母语者对英语元音对比/i-I/的类别表征间的区分，改善了汉语母语者对英语元音对比/i-I/的感知分类模式，并提高了对这两个英语元音的识别水平。对于英语元音/i/与汉语元音/i/间的区分没有明显改善，可能针对性英语语音的训练还不够充分。

It is well known that second language (L2) learners have great difficulties and large individual differences in L2 vowel perception. Previous studies have revealed the effect of age of L2 acquisition, L2 proficiency, and frequency of L2 use on L2 vowel perception, however, even when the above experiences are very close for L2 learners, there are still significant individual differences in L2 vowel perception. In this dissertation, the concept of distinctiveness of English vowel category establishment was introduced to reveal its relationship with English vowel perception of Chinese college students in three experiments. One main research focus is whether and how the establishment of English vowel category is associated with English vowel perception for Chinese college students, a L2 group with similar education and L2 experiences.

Experiment 1 measured how distinctively L2 vowel category was established, and explored the characteristics of the establishment of English vowel categories for Chinese native speakers. In Experiment 1a, English vowel identification were evaluated for Chinese- (CN) and English-native (EN) speakers. As expected, English vowel identification performance of CN speakers was significantly lower than that of EN speakers. Meanwhile, among all English vowels, Chinese native speakers had lower identification accuracy and higher confusion rate for English vowel contrast /i-I/, and the confusion pattern is asymmetric (misperceive /i/ as /I/, but not /I/ as /i/). These results suggested that the English vowel categories /i/-/I/ may serve as the candidates for analyzing the establishment of English vowel categories of CN speakers. In Experiment 1b, English vowel category establishment was measured. The purpose of this experiment was to quantify the establishment of L2 vowel category by measuring the perceptual distance between vowel categories in the F1 x F2 matrix. It involved three steps: 1. establishing the F1 x F2 matrix, which ensures that the vowel perception distribution can be measured; 2. establishing the vowel perception map according to listener’s English vowel categorization scores. Besides, in this map, the most representative item score was measured to determine whether the L2 vowel category had been established; 3. For L2 learners who have established the English vowel category, the most representative item (MRI) is the vowel prototype. In this experiment, the prototype distance between vowel categories in the F1 x F2 matrix is used to quantify the establishment of L2 vowel categories. Secondly, based on the cross-linguistic comparisons, results showed that the English vowel perception maps with large individual variability were classified into four patterns: Distinctive Two, Confusion Overlap, /i/ dominant and /I/ dominant. Only 6.4% of Chinese native speakers reached the pattern of Distinctive Two, a native-like pattern, while most of Chinese native speakers (93.6%) were located in the other three patterns. In short, experiment 1 put forward the concept and measurement of distinctiveness of establishment of L2 vowel categories, and pointed out the characteristics of Chinese native speakers in the establishment of English vowels /i/-/I/.

Experiment 2 conducted group comparison analysis on English vowel identification between 46 native Chinese speakers and 9 native English speakers, and examined whether establishment of English vowel categories can explain the perception of English vowels by native Chinese speakers. This experiment measured English vowel identification, English vowel categorization and typicality rating, Chinese vowel categorization and typicality rating for Chinese native speakers. The main findings are as follows: 1) the prototype distance between Chinese /i/ and English /i/ accounted for 14.9% variability of English vowel identification and the prototype distance between English /i/ and English /I/ accounted additional 13.8% variability of English vowel identification for CN speakers. These results suggest that CN speakers' perception of English vowels relied on their capacity to establish English vowel categories. On the other hand, CN speakers' perception of English vowels was not significantly correlated with their age of English acquisition, use of English per week, English listening comprehension and English proficiency;

In Experiment 3, the impact of L2 vowel training on the distinctiveness of English vowel category establishment for CN speakers was further explored. Twenty-five native Chinese speakers were trained in identification of English vowel /i-I/ for six consecutive days. Three test sessions were conducted: pre- and post-training, and one month after training. Before and after the training, all subjects completed tasks of English vowel identification, English vowel categorization and typicality rating, Chinese synthetic vowel categorization and typicality rating. Results showed that: 1) For the prototype distance between English vowels /i/ and /I/, the training groups improved significantly right after training and retained after one month. But the training effect was not found for the prototype distance between English /i/ and Chinese /i/; 2) The training effect of the categorization patterns of English vowels /i-I/ showed that the number of subjects in the "Distinctive Two" pattern increased significantly after training, and the number of subjects in the "Confusion Overlap" pattern decreases significantly with the number of subjects in the “/i/ dominant” pattern increased significantly and the number of subjects in the “/I/ dominant” pattern decreased significantly after training; 3) The identification of vowel /i/ and /I/ were significantly improved after training, and remained after one month. These results revealed that L2 vowel training with high intensity significantly improved the distinctiveness between English vowels /i/ and /I/, and enhanced the perception of the two English vowels.