Understanding Perceptual Compensation in Sound Change

语言学家长期以来一直假设，当听者误解了说话者并将这种误解融入到他/她自己的语法中时，就会发生声音变化——一种语言随着时间的推移而改变发音。然而，人们对这种调整发生的机制知之甚少，事实上，越来越多的证据表明，听众善于补偿语言感知和产生的变化。由此可见，与事实相反，声音变化的发生是微乎其微的。本研究探讨了感知补偿（PC）无法纠正语音变化的情况，以及这种失败如何导致声音变化。研究小组将开发PC（以及扩展声音变化）的正式表示，以使用贝叶斯推理（一种通常用于认知科学研究的统计工具）来证实实验结果。该模型预测，当调整模型中的某些参数时，PC的影响可以最小化甚至克服；也就是说，补偿的相对大小是由语言的特性和说话者先前的语言经验所调节的。为了验证PC理论及其与声音变化的联系，研究小组将进行一系列实验室实验，测试PC效应的极限和可变性。实验范式包括通过改变(1)所涉及声音的频率，(2)语音信号发生的背景，以及(3)语音信号本身的性质来测试听者的补偿反应。这项研究的结果将拓宽音系学、声音变化、语音产生和感知理论的实证数据库，并将为理论研究开辟新的领域。本研究从跨语言的角度对说话者对各种语言输入的评价反应进行了详细而深入的实验和理论研究，填补了语音变化研究和语言学研究的空白。对声音变化的更好理解将为语言变化、语言习得和人类交流的本质提供一个独特的窗口，并可能为第一/第二语言习得、语言教学和语音技术的发展提供见解。

Linguists have long hypothesized that sound change--changes in a language over time that alter pronunciation--occurs when a listener misinterprets a speaker and incorporates the misperception into his/her own grammar. However, little is known about the mechanism through which such adjustments occur and mounting evidence is in fact suggesting that listeners are adept at compensating for variation in speech perception and production. From this, it should follow that sound change would occur minimally, contrary to fact. This research investigates the circumstances under which perceptual compensation (PC) fails to correct for variation in speech and how such failures lead to sound change. The research team will develop a formal representation of PC (and by extension sound change) to corroborate the experimental findings using Bayesian inference, a statistical tool commonly employed in cognitive science research. This model predicts that the effects of PC may be minimized or even overcome when certain parameters in the model are adjusted; that is, the relative magnitude of compensation is mediated by the properties of a language and the speaker's prior experience with it. To test this theory of PC and its connection to sound change, the research team will conduct a number of laboratory experiments testing the limit and variability of PC effects. The experimental paradigm involves testing listeners' compensatory responses by varying (1) the frequency of the sounds involved, (2) the context in which the speech signal occurs, and (3) the nature of the speech signal itself.The results of this research will broaden the empirical database on which theories of phonology, sound change, and speech production and perception are grounded and will advance novel venues for theoretical investigations. This research will fill a gap in work on sound change and in linguistics in general by providing detailed and in-depth experimental and theoretical investigations on speakers' evaluative responses to various linguistic inputs from a cross-linguistic perspective. A better understanding of sound change will offer a unique window into the nature of language change, language acquisition, and human communication and may provide insight into first/second language acquisition, language-teaching, and the development of speech technology.