Neutralization and Homophony in Phonological Learning

语言理论的中心任务之一是识别和形式化语言变化和语音模式变化背后的驱动力。先前的研究一直围绕着这样的争论，即这种力量是否可能来自学习压力，例如，更难学习的模式可能是跨语言罕见的，因为它们不太可能代代相传（例如，Wilson, 2006）；或从环境选择中，例如，从功能上更有效地进行交流的模式（例如，Wedel, 2006）。我的论文将通过提供一个更完整的视角来为这场辩论做出贡献——虽然我们对学习更容易的模式的偏好，作为一种普遍的约束，在塑造我们的语言中起着至关重要的作用，但可学习性偏好本身在某些情况下可以追溯到我们的功能需求，例如，我们倾向于避免同音异义词，以防止模棱两可的语言。从类型学的角度来看，来自先前未被充分研究的语言的新证据表明，由于避免同音谐音等因素，我们的语音模式和词汇可能彼此紧密相连（例如，Silverman， 2010年关于韩语；Kaplan & Muratani， 2015年关于日语）。然而，目前的音韵学理论，其早期发展深深植根于西方语言的音韵学，仍然与这些新的跨语言发现脱节。因此，我的论文试图通过使用类型学方法和形式分析，将研究不足的语言纳入理论建设，从而弥合这一差距。这项研究采用了一种更新颖的方法，从学习的角度来解释类型学数据，通过学习实验和计算模型来测试某些语音模式的可学习性是否会受到词义特性和我们避免同音异义词的需要的影响。最终，这项研究有助于我们理解语言是如何在形式和功能压力下进化成有效的交流系统的。

One of the central tasks of linguistic theories is to identify and formalize the driving forces behind language change and variation in speech patterns. Previous research has centered around the debate on whether such forces could arise from learning pressures, e.g., patterns that are harder to learn may be cross-linguistically rare since they are less likely to be transmitted over generations (e.g., Wilson, 2006); or from environmental selections, e.g., for patterns that are functionally more effective for communication (e.g., Wedel, 2006). My thesis will contribute to this debate by offering a more integrated perspective-while our preference for learning easier patterns, as a universal constraint, plays a vital role in shaping our speech, the learnability preference itself could in some cases trace its roots back to our functional needs, e.g., our tendency to avoid homophones for preventing ambiguous speech. Typologically, new pieces of evidence from previously understudied languages have suggested that our speech patterns and lexicon could be tightly linked to each other, due to factors such as homophony avoidance (e.g., Silverman, 2010 on Korean; Kaplan & Muratani, 2015 on Japanese). However, current phonological theories, with their early development deeply rooted in phonology of Western languages, are still disconnected with these new cross-linguistic findings. Thus, my thesis seeks to bridge this gap by incorporating the under-researched languages into theory building, using typological methods and formal analysis. The study takes an even more novel approach to explain the typological data from a learning perspective, by testing whether the learnability of certain speech patterns could be affected by the properties of word meaning and our need to avoid homophones, through learning experiments and computational modeling. Ultimately, the study contributes to our understanding of how languages evolve towards an effective communication system under both formal and functional pressures.