Files
Abstract
This dissertation deals broadly with subphonemic variation in English stops /p t k b d g/. Each chapter deals with a different aspect of these stops: Chapter 1 examines long-distance coarticulatory properties of these stops, and how there is a relationship between the patterns of voicing and the acoustic segments surrounding them (making distinctions beyond the phonological voiced/voiceless, including 5 different patterns of voicing based on acoustics). Using machine learning to classify the voicing pattern from the long-distance acoustics shows that there is a relationship between these patterns and the acoustics of the surrounding segments. Chapter 2 examines voice onset time in these stops in the Digital Archive of Southern Speech (Kretzschmar et al. 2013, Kretzschmar et al. 2019) and how this subphonemic variable is affected by several linguistic and sociolinguistic variables. Results show linguistic variables to be significant, both in line with and clarifying previous research. Finally, Chapter 3 improves the alignment results of the Montreal Forced Aligner by adding pronunciation variants featuring t/d deletion, and this technique will prove to be a useful tool in future sociolinguistic research. Additionally, this chapter examines the subphonemic properties of /t d/ in an effort to understand the acoustic correlates that influence MFA's determination. For /d/, the burst was found to influence MFA's decision, and for /t/, the closure, burst, and other variants were found to influence MFA's decision.