Doctoral Dissertation Research: Effects of phonological contrast on phonetic variation

University Of Massachusetts Amherst, Amherst MA

Investigators

Abstract

Every time someone speaks, they produce complex sound waves. Understanding the acoustic details of these sound waves can help improve voice recognition technologies and provide insight on language differences which can potentially aid in language learning. A large body of work in linguistics and related disciplines has shown that speakers do not always pronounce sounds in the same way. For example, the sound "p" in English will be pronounced differently in different contexts (beginning of a word vs. end of word) and at different times simply because talkers are humans. People do not usually hear these differences, but they are present in the acoustics and can be measured. Variation in speech is natural and happens in every language. However, less is known about exactly how much variation occurs in different languages. This research seeks to address this gap by investigating whether speakers of some languages have more variable pronunciation than speakers of other languages and what might explain the difference. One frequently assumed hypothesis says that speakers of languages with more sounds will show less variation (be more precise) when they pronounce those sounds. The researchers will gather data from different languages to test and revise this hypothesis. They will also develop concrete methods for measuring acoustic variability in language which can then be applied to future work. The experiments compare stop sounds in Hindi ("b", "bh", "p" and "ph") and English ("b" and "p"), and sibilant sounds in Polish ("s", "ś", and "sz") and French ("s" and "sh"). Both experiments examine multiple acoustic cues to the contrasts in each language. The second portion of the project will use the conclusions from these experiments to investigate whether speakers imitate the variability of other speakers that they hear. It is well documented that speakers change their pronunciation based on speech they have recently heard, but it is not known whether speakers will change the amount of variability in their speech based on what they have heard. To address this question, the researchers will carry out an imitation study where speakers listen to speech before being prompted to speak to see if the variability in their speech changes after listening. The results of these studies will increase understanding of variability in pronunciation across different languages and contexts. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.