From voicing to register: the evolution of a sound change in Southeast Asia

毫无疑问，语言会随着时间的推移而变化，但为什么特定的变化会在何时何地发生，这个问题——即所谓的“驱动问题”——已被证明是极难回答的。在单个声音的层面上尤其如此：发音的变化，几乎从定义上来说，在发生时是不被察觉的。然而，为了更好地理解声音的变化，我们需要在变化的早期阶段研究语言，以确定在某些情况下导致变化而在其他情况下不会导致变化的特性。在这个项目中，我们通过关注所谓的“寄存器语言”的进化来解决驱动问题。在语域语言中，单词之间的区别通常是基于一系列声学特性，如语音质量、元音质量和音高。例如，在泰国所说的孟语中，音节/klan/有两种不同的发音：一种是“高音域”版本，意思是“舔”，发音清晰，音调略高，元音略低；还有“低音域”版本，意思是“淘气的”，用呼吸更急促的声音发出，音调略低，元音略像弱读音。我们知道，许多，或者可能是所有的语域语言曾经都很像英语，主要通过声带振动的存在或不存在来区分这些单词对，这种特性被称为“发声”。就像英语根据发音来区分“bin”和“pin”一样，说蒙古语的人也曾经把“naughty”这个词发成/glan/，以/g/开头，以区分它和/klan/。我们在这个项目中的目标是理解语言是如何以及为什么从像英语这样的“发声语言”转变为像mon这样的“注册语言”。这个问题的部分答案涉及到声带振动并不是发声的唯一线索。例如，我们知道，跟在不发音辅音后面的元音，比如/p t k/，比跟在浊音后面的元音，比如/b d g/，开头的音调要高。在像汉语这样的语言中，这种音高差异被夸大了，并发挥了主要作用，把曾经基于声音的对比变成了以音高为标志的对比——语言学家称之为声调语言。但音高并不是伴随发声的唯一线索：有证据表明，发声的质量（呼吸声或嘎吱声）和元音质量（或音色）也会受到影响。然而，在真实语言中，还没有对这些发声的次要线索进行系统的声学研究；我们也不明白为什么语域系统经常演变成简单的音调系统或具有许多元音特征的语言。为了更好地理解导致语域系统形成的条件，我们将对一系列东南亚语言进行声学和感知研究，包括那些保留发音对比的语言，以及在发展语域道路上的不同阶段的语言。我们把重点放在东南亚，因为在这个地区发现了丰富多样的发声语言和注册语言。我们的发现将用于建立发声进化的计算模型，旨在解决以下问题：为什么有些语言发展了语域，而另一些语言保留了原始的发声对比？为什么不同的语域系统似乎强调不同的语音次要属性，比如音高或音质？为什么一些改变的途径更普遍，而另一些却很少得到证实？这些模型将允许我们测试诸如感知显著性、词形和接触等因素如何影响发声和语域进化的假设。研究结果不仅有助于我们理解注册系统是如何产生的，而且还将对声音变化的生命周期提供独特的见解。

While there is no doubt that language changes over time, the question of why particular changes happen when and where they do - the so-called 'actuation problem' - has proven extremely difficult to answer. This is especially true at the level of individual sounds: changes in pronunciation are, almost by definition, unnoticeable while they are happening. In order to gain a better understanding of sound change, however, we need to be studying languages at these early stages of change, in order to identify the properties which lead to change in some cases but not others.In this project, we address the actuation problem by focusing on the evolution so-called 'register languages'. In a register language, words are often distinguished from one another based on a 'bundle' of acoustic properties such as voice quality, vowel quality, and pitch. For example, in the Mon language as spoken in Thailand, the syllable /klan/ has two distinct pronunciations: a 'high register' version, with the meaning 'lick', produced with a clear voice, at a slightly higher pitch, and with a slightly lower-sounding vowel; and 'low register' version, with the meaning 'naughty', produced with a breathier voice, at a slightly lower pitch, and with a slightly more schwa-like vowel.We know that many, or perhaps all, register languages were once much like English in distinguishing such word pairs chiefly through presence or absence of vocal fold vibration, a property known as 'voicing'. Just like English distinguishes the word 'bin' from 'pin' on the basis of voicing, so too did Mon speakers once pronounce the word for 'naughty' as /glan/, with an initial /g/ sound, to distinguish it from /klan/. Our goal in this project is to understand how and why languages shift from being 'voicing languages' like English, to being 'register languages' like Mon.Part of the answer to this question involves the fact that vocal fold vibration is not the only cue to voicing. For example, we know that vowels following voiceless consonants, like /p t k/, start with a higher pitch than those following voiced consonants, like /b d g/. In languages like Chinese, this pitch difference was exaggerated and took on a primary role, turning a contrast once based on voicing into a contrast signaled by pitch - what linguists call a tone language. But pitch is not the only cue that accompanies voicing: there is evidence that both voice quality (breathiness or creakiness) and vowel quality (or timbre) are affected by voicing as well. However, no systematic acoustic study of these secondary cues to voicing has yet been conducted in real languages; nor do we understand why register systems often evolve into simple tone systems or languages with many vowel qualities. In order to better understand the conditions which lead to register systems, we will conduct acoustic and perceptual studies of a range of Southeast Asian languages, including those which retain a voicing contrasts as well languages at various stages along the path to developing register. We focus on Southeast Asia because of the rich diversity of voicing and register languages that are found in this part of the world. Our findings will be used to build computational models of the evolution of voicing designed to address questions such as: Why do some languages develop register, while others preserve the original voicing contrast? Why do different register systems seem to emphasize different secondary properties of voicing, like pitch or voice quality? Why are some paths of change more common, while others are more rarely attested? These models will allow us to test hypotheses about how factors such as perceptual salience, word shape, and contact impact the evolution of voicing and register. The results will contribute not only to our understanding of how register systems emerge, but will also give a unique insight into the life cycle of a sound change.