CAREER: Representations of phonetic reduction and dialect variation in speech production and perception

人类的言语本质上是高度多变的：同一个说话者发出的同一个词的两个发音在声学上是不相同的。这种可变性的来源包括个别单词的属性，如它们在口语中出现的频率；单词发出的语境属性，如给出先前单词的单词的可预测性；对话环境的属性，如形式；以及说话者的属性，如性别或地区背景。以前在这一领域的大多数研究都集中在一个或两个变异来源上。然而，统计分析的最新进展使得在这项研究中同时检验许多变量成为可能。要检验的两个具体假设是，由于语言和社会因素造成的变化是高度系统化的，这种系统性可以促进语音处理。因此，这个项目的第一个目标是确定这些语言和社会因素在多大程度上可以解释言语的可变性。为了实现这一目标，以美国英语为母语的人所产生的言语将被记录下来，并在一系列生产实验中进行分析，这些实验操纵这些可变性来源的不同组合。该项目的第二个目标是确定这些可变性来源如何影响对言语所承载的语言和社会信息的感知和表达。为了实现这一目标，将使用在生产实验中收集的录音进行单词识别、单词处理和方言识别任务。除了对我们理解语音产生和感知的变异有很大的帮助外，这个项目的结果还有可能为语言习得、自闭症谱系障碍的语言处理以及包括自动语音识别在内的语音技术的研究提供信息。

Human speech is inherently highly variable: no two utterances of the same word produced by the same talker are acoustically identical. The sources of this variability include properties of individual words, such as how often they occur in spoken language; properties of the context in which a word is uttered, such as the predictability of a word given the previous word; properties of the conversational setting, such as formality; and properties of the talker, such as gender or regional background. Most previous research in this area has focused on one or two sources of variation at a time. However, recent advances in statistical analysis make possible the examination in this study of many variables simultaneously. The two specific hypotheses that will be tested are that the variation due to linguistic and social factors is highly systematic and that this systematicity can facilitate speech processing. Thus, the first goal of this project is to determine how much of the variability in speech can be accounted for by these kinds of linguistic and social factors. To achieve this goal, speech produced by native speakers of American English will be recorded and analyzed in a series of production experiments that manipulate different combinations of these sources of variability. The second goal of the project is to determine how these sources of variability affect the perception and representation of the linguistic and social information that is carried by speech. To achieve this goal, word recognition, word processing, and dialect identification tasks using the recordings collected in the production experiments will be conducted. In addition to substantially contributing to our understanding of variation in speech production and perception, the results of this project have the potential to inform research on language acquisition, language processing in autism spectrum disorders, and speech technology, including automatic speech recognition.