Doctoral Dissertation Research: Cross-language transfer in voice onset time: A window into perceptual adaptation in brain and behavior

在作为第二语言获得的所有语言中，英语拥有最多的用户(9.78亿)和最广泛的地理覆盖范围。在美国，大约22%的人口同时会说英语和另一种语言。因此，带有“外国口音”的英语变体很常见。外国口音品种与“本土口音”品种有几个不同之处。这篇博士论文主要研究将英语作为第一语言或第二语言的不同说话者如何发音。此前的研究表明，听者能够在一分钟内适应外国口音演讲的声音。然而，听众如何获得这种快速适应仍然是未知的。了解听者如何在语音感知中实现这种程度的灵活性，不仅对于为语音感知的语言学理论提供信息，也对于为广受欢迎的语音助手Alexa和Siri等自动语音识别系统的技术进步提供信息至关重要。本论文通过一系列的行为学和电生理学实验来研究第二语言言语产生和第一语言言语知觉之间的关系。博士论文的研究以三个研究问题为指导：(1)西班牙口音英语的经历如何促进后来对西班牙口音英语言语的理解？(2)这种经验需要有多具体？(3)语言加工的时间敏感型神经认知测量，如脑电--头皮电活动的测量--如何解释第二语言口音言语知觉？理想的顺应性框架为研究第二语言(L2)言语产出和第一语言(L1)言语知觉之间的相互作用提供了坚实的理论基础。在这个框架下，听者对声音线索和语音类别之间的概率分布很敏感。听者可以学习特定的线索-类别映射，这些映射表征了一组特定的说话者，以实现稳健的感知。本文使用语音起始时间(Vot)与清音辅音/p/、/t/、/k/之间的线索-范畴映射作为这一假设的检验案例。由于带有西班牙口音的英语的跨语言转换，这些无声停顿和它们的有声停顿之间的知觉模糊提供了一个了解适应过程的窗口。本研究采用跨通道启动的方法，通过脑电(EEG/ERP)对西班牙口音/p-tk/的行为适应和神经认知适应进行了测量。研究人员实施了一种创新的实验设计，以比较(A)具有相同口音的多个说话者暴露于可变性的因素，以及(B)暴露阶段遇到的重音特征与测试阶段遇到的重音特征之间的相似程度。到目前为止，这些因素还没有被直接比较，这限制了我们对二语口音的顺应的理解。此外，研究人员使用不同暴露条件下的一组说话者来控制每个说话者的体验类型和数量。最后，本研究首次用脑电研究了二语重音语音的知觉适应。使用这一细粒度、对时间敏感的衡量标准，揭示了语音、语音和语义信息在解决二语口音演讲中感知歧义方面的相对贡献。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Of all languages acquired as a second language, English has both the largest number of users (978 million) and the broadest geographic reach. In the U.S., approximately 22% of the population speaks both English and another language. Consequently, “foreign-accented” varieties of English are common. Foreign-accented varieties differ from “native-accented" varieties in several ways. This doctoral dissertation focuses on how speech sounds are pronounced by different speakers who speak English as a first or second language. Previous research has shown that listeners are able to adapt to the speech sounds of foreign-accented speech within a minute of exposure. However, how listeners acheive this rapid adaptation is still unknown. Understanding how listeners achieve this level of flexibility in speech perception is critical for informing not only linguistic theories of speech perception but also technological advances in automated speech recognition systems like the popular voice assistants Alexa and Siri. The present dissertation investigates the interface between second language speech production and first language speech perception in a series of behavioral and electrophysiological experiments.The doctoral dissertation research is guided by three research questions: (1) How does experience with Spanish-accented English facilitate later comprehension of Spanish-accented English speech? (2) How specific does this experience need to be? (3) What can time-sensitive neurocognitive measures of language processing such as electroencephalography--the measure of electrical activity at the scalp--illuminate about second language-accented speech perception? The ideal adapter framework provides a strong theoretical foundation from which to investigate the interface between second language (L2) speech production and first language (L1) speech perception. Under this framework, listeners are sensitive to the probability distributions between acoustic cues and phonetic categories. Listeners can learn the specific cue-category mappings that characterize a particular group of talkers in order to achieve robust perception. The present dissertation uses the cue-category mapping between voice onset time (VOT) and the voiceless stop consonants /p/, /t/, and /k/ as a test case for this hypothesis. Perceptual ambiguity between these voiceless stops and their voiced counterparts /b/, /d/, and /g/, respectively, due to cross-language transfer in Spanish-accented English provides a window into the adaptation process. The present dissertation uses cross-modal priming to measure adaptation to Spanish-accented /p t k/ both behaviorally and neurocognitively with electroencephalography (EEG/ERP). The researchers implement an innovative experimental design to compare the factors of (A) exposure to variability across multiple talkers with the same accent and (B) the level of similarity between the accented features encountered during the exposure phase and those encountered during the test phase. To date, these factors have not been compared directly, limiting our understanding of adaptation to L2-accented speech. In addition, the researchers use a single group of talkers across exposure conditions to control the type and amount of experience with each individual talker. Finally, this study is the first to investigate perceptual adaptation to L2-accented speech with EEG. Using this fine-grained, time-sensitive measure reveals the relative contributions of phonetic, phonological, and semantic information in resolving perceptual ambiguity in L2-accented speech.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.