Effects of Native Language and Linguistic Exposure on Non-Native Listeners' Use of Prosodic Cues in Speech Segmentation

语音是连续不断的声音流，没有单一的设备明确地标记单词的边界。对于语言学习者来说，一个关键的挑战是，不同语言的单词边界信号是不同的；因此，在试图将第二语言分割成单词时，说话者使用母语的经验可能会产生误导。明确成人语言学习者如何识别连续言语中的单词是非常重要的，首先是因为它可以帮助解决关于成人语言学习和语言理解如何发生的关键理论争论。更准确地说，它可以揭示成年人的大脑是否（如果是，在多大程度上）具有足够的可塑性，以发展对新的分割线索的敏感性，以及调节这种学习是否可能的因素。因此，这项研究对认知科学和语言科学具有重要意义。此外，语音分割对于多语使用者之间的成功沟通至关重要。随着我们的社会越来越多地使用多种语言，语言已经成为K-12和大学课程的固有组成部分，计算语言学家和计算机科学家一直在努力开发能够在多语言使用者之间进行有效交流的技术。因此，调查影响成年语言学习者如何将第二语言分割成单词的因素，对语言教学和通信技术（例如，自动语音识别）的发展具有重要意义。本研究的重点是母语和最近的语言接触对成年语言学习者在语音切分中使用韵律线索（特别是音高）的影响。其主要目的是确定母语和第二语言之间的异同如何影响成人语言学习者在语音切分中使用韵律线索的能力。本文测试了两个假设：(i)韵律同化假设：与母语韵律系统相似但又不同的第二语言韵律系统被同化，因此在语音分割中比与母语韵律系统差异很大的第二语言韵律系统更难学习；线索权重转移假说：母语分词线索的功能负荷会被转移到第二语言中。本研究的第二个目的是评估最近的语言暴露对语音分割的影响。第三个假设进行了测试：（iii）概率语音分割假设：语音处理系统使用一套单一的语音分割策略，这些策略反映了跨母语和第二语言的单词边界线索的概率。为了验证这些假设，本研究将重点放在母语为法语、韩语、英语和荷兰语的听众身上，这些听众是否会法语或韩语作为第二语言。听众将完成视觉世界眼球追踪和人工语言分割实验。

Speech is a continuous flow of sounds where no single device explicitly marks word boundaries. A crucial challenge for language learners is that the cues that signal word boundaries differ across languages; thus, a speaker's experience with her native language may be misleading when attempting to segment a second language into words. Specifying how adult language learners recognize words in continuous speech is very important, first because it can help resolve key theoretical debates about how language learning and language comprehension take place in adults. More precisely, it can shed light on whether (and if so, the extent to which) the adult brain is sufficiently plastic to develop sensitivity to new segmentation cues and the factors that modulate whether or not this learning is possible. This research can therefore have important implications for cognitive and linguistic sciences. Furthermore, speech segmentation is essential to successful communication among multilingual speakers. As our society is becoming increasingly multilingual, languages have become intrinsic components of K-12 and college curricula, and computational linguists and computer scientists have striven to develop technologies that could afford efficient communication among multilingual speakers. Investigating the factors that influence how adult language learners segment a second language into words can thus have important implications for the teaching of languages and for the development of communication technologies (e.g., automatic speech recognition).This research focuses on the influence of the native language and of recent linguistic exposure on adult language learners' use of prosodic cues, specifically pitch, in speech segmentation. Its primary aim is to determine how the similarities and differences between the native language and second language affect adult language learners' ability to use prosodic cues in speech segmentation. Two hypotheses are tested: (i) the Prosodic Assimilation Hypothesis: Second-language prosodic systems that are similar to, yet different from, native-language prosodic systems are assimilated and thus more difficult to learn in speech segmentation than second-language prosodic systems that are very different from native-language prosodic systems; and (ii) the Cue-Weighting Transfer Hypothesis: The functional load of segmentation cues in the native language is carried over to the second language. A secondary aim of this research is to assess the effect of recent linguistic exposure on speech segmentation. A third hypothesis is tested: (iii) Probabilistic Speech Segmentation Hypothesis: The speech processing system uses a single set of speech segmentation strategies that reflect the probabilities of cues to word boundaries across the native language and second language. To test these hypotheses, this research focuses on native French, Korean, English, and Dutch listeners with or without knowledge of French or Korean as second languages. Listeners will complete visual-world eye-tracking and artificial-language segmentation experiments.