The Interplay Between Segmental Timing and Coarticulation: Acoustic, Perceptual, and Phonological Investigations

在国家科学基金会的支持下，帕特里斯·贝多博士将对语音的可变性进行为期三年的语言学研究。这个项目调查了一种特殊的可变性如何最终成为一种语言的系统属性，从而导致语言的变化。重点是相邻或邻近声音的重叠发音，称为“协同发音”。在某些声音变化中，触发协同发音的声音会随着时间的推移而消失，只留下协同发音效果。例如，在法语、印地语和数百种其他语言中，元音中的协同发音鼻音最初只由紧随其后的鼻辅音触发，如“m”或“n”。但随着时间的推移，鼻辅音消失了，留下了元音鼻音的系统性，现在是这个词的必备属性。Beddor博士假设，失去一个协同发音触发因素并将其特性与另一个声音联系起来，涉及到一个阶段，在这个阶段，触发因素持续时间的缩短被更广泛的协同发音影响所抵消(例如，鼻辅音随着附近元音变得更加鼻音而缩短)。她进一步假设，听者通过对这种时序关系中的变化不敏感来促进这些变化。贝多尔博士的研究通过研究英语、泰语、希腊语和伊卡兰加语四种不同计时模式的语言，模拟了一种语言变化的各个阶段。将对这些语言进行声学分析，以核实初步研究中出现的时间模式。在这些分析之后，将进行知觉测试和进一步的音系学研究。这一跨言语科学影响领域的多语言研究具有理论和实践意义。首先，它将拓宽声音变化、言语产生和言语感知理论所依据的经验数据库。其次，更好地理解语音变异和语音变化揭示了一些方言差异是如何产生的。更好地理解方言的出现可能会减少对某些语言变体的社会污名。第三，探索听者对语音变化的敏感度的知觉测试对婴儿如何在他们正在学习的语言中提取语音规则的理论具有启示意义。最后，这些结果将与语音技术有关。计算机语音识别系统比人类犯的错误更多。这在一定程度上是因为计算机识别器对协同发音变化表现出更高的敏感度。对人类听者敏感的语音变异性的研究可能有助于解决语音工程中的这一实际挑战。

ABSTRACTBeddor 0118684With National Science Foundation support, Dr. Patrice Beddor will conduct three years of linguistic research on the variability of speech sounds. This project investigates how a particular kind of variability can eventually become a systematic property of a language, leading to language change. The focus is the overlapping articulation of adjacent or nearby sounds known as "coarticulation". In some sound changes, the sound that triggered the coarticulation is lost over time, leaving behind only its coarticulatory effects. For example, in French, Hindi, and hundreds of other languages, coarticulatory nasality in vowels was originally triggered only by a following nasal consonant, such as "m" or "n". But the nasal consonant dropped out over time, leaving vowel nasality as a systematic and now required property of the word. Dr. Beddor hypothesizes that the loss of a coarticulatory trigger and the linking of its properties to another sound involve a stage in which shortened duration of the trigger is offset by more extensive coarticulatory influences (e.g., nasal consonants shorten as nearby vowels become more nasal). She further hypothesizes that listeners facilitate these changes by being perceptually insensitive to variation in such timing relationships. Dr. Beddor's research mimics the stages of a language's change by studying four languages that differ in the relevant timing patterns: English, Thai, Greek, and Ikalanga. Acoustic analyses of these languages will be conducted to verify the timing patterns that emerged in preliminary studies. These analyses will be followed by perceptual testing and further phonological investigations.This multi-language research has theoretical and practical significance across domains affected by speech science. First, it will broaden the empirical database on which theories of sound change, speech production, and speech perception are grounded. Second, better understanding of speech variation and speech sound change reveals how some dialect differences emerge. Better understanding of dialect emergence may reduce the social stigmatization attached to some speech varieties. Third, the perceptual tests that explore listener sensitivity to speech variation have implications for theories of how infants extract phonetic regularities in the languages they are learning. Finally, these results will bear on speech technology. Computer speech recognition systems make more errors than humans do. This is partly because computer recognizers display greater sensitivity to coarticulatory variation. Study of the kinds of speech variability that human listeners are sensitive to may help solve this practical challenge in speech engineering.