Previous studies have shown that nonnative listeners have poorer speech recognition compared to native listeners, particularly under difficult listening conditions. While native English speaking listeners can tolerate large amounts of distortion to speech, listeners to whom English is a second language (L2) may require a much longer period of learning to develop equally robust central speech patterns.; The present study looked at speech perception by nonnative listeners under conditions of noise (SNR of l5dB, 10dB, 5dB, 0dB) and reduced spectral information (2, 4, 6, 8 and 16 frequency bands). Subjects were tested using speech processed to simulate the listening condition typically encountered by cochlear implant listeners. Normal hearing listeners whose first language was Spanish were tested with American English phonemes, words and sentences. Results were compared with results for native English listeners tested under the same conditions.; Speech perception depends on phonetic, lexical and linguistic knowledge. We wanted to determine the relative contributions of phonemic and lexical processing to speech recognition as a function of language experience by varying both the spectral resolution and the linguistic complexity of the materials. Contrary to what was initially expected, results determined that the main difficulty nonnative listeners encountered in L2 perception is in vowel perception, not in linguistic integration. Context factors were determined using the Boothroyd and Nittrouer model (1988), in which a parameter ‘k’ reflects the degree of predictability in the sentence material. Results suggest that early L2 learners are able to use linguistic information when they are forced to do so.; Plomp's SRT (Speech Reception Threshold) model (1986) was applied to the data. Plomp defines a parameter ‘D’ that measures the distortion in perception of the speech stimuli. The difficulties in speech perception of the L2 by nonnative listeners are similar to difficulties encountered by cochlear implant patients. We do not suggest that lack of experience with a L2 is the same as suffering from a hearing loss or hearing impairment, but results show that this simple model can account for both factors.
展开▼
机译：先前的研究表明，与本地收听者相比，非本地收听者的语音识别能力较差，尤其是在困难的收听条件下。虽然以英语为母语的听众可以忍受语音的大量失真，但英语为第二语言（L2）的听众可能需要更长的学习时间才能发展出同样强大的中心语音模式。本研究着眼于在噪声（SNR为15dB，10dB，5dB，0dB）和频谱信息减少（2、4、6、8和16频段）的情况下非本地听众的语音感知。使用语音处理测试对象，以模拟人工耳蜗植入听众通常遇到的听觉状况。母语为西班牙语的普通听力人员接受了美国英语音素，单词和句子的测试。将结果与在相同条件下测试的英语母语收听者的结果进行比较。语音感知取决于语音，词汇和语言知识。我们想通过改变材料的频谱分辨率和语言复杂性，确定语音和词汇处理对语音识别的相对贡献，作为语言经验的函数。与最初预期相反，结果确定了在L2感知中遇到的非母语听者的主要困难在于元音感知，而不是语言整合。使用Boothroyd和Nittrouer模型（1988）确定上下文因素，其中参数“ k ”反映了句子材料中可预测性的程度。结果表明，初二语言学习者能够在被迫使用语言信息的情况下使用它们。将Plomp的 SRT （语音接收阈值）模型（1986）应用于数据。 Plomp定义了一个参数“ D ”，用于测量语音刺激的感知失真。非本地听众在听L2语音时遇到的困难与人工耳蜗植入患者所遇到的困难相似。我们不建议缺乏L2经验与遭受听力损失或听力障碍相同，但结果表明，此简单模型可以解释这两个因素。
展开▼
