Creaky voice in Australian English

The term creaky voice, or creak, describes a complex voice quality used by speakers to convey linguistic, paralinguistic and social functions across several languages including English. Although creak is attested in Australian English (AusE), very little is known about its use in this variety. This thesis examines aspects of creaky voice in AusE speakers, specifically those from Sydney. Firstly, we address the methodological question of how to annotate creaky voice. Previously, creak has generally been identified manually; however, this is a time-consuming and labour-intensive process which is subject to annotator-bias. Considering this, we turn to automatic creaky voice detection methods. We use a corpus of AusE recordings to explore the best automated method, or combination of methods, for detecting creaky voice. Two methods used in combination, the Creak Detector algorithm and the AntiMode method, when used with sonorant segments only, yield significantly better creak detection than each tool used independently. We call this combined method the Union method. We then investigate how well listeners can identify creak in different voices. Studies have shown that creak is generally produced with low fundamental frequency (f0) and stereotypically associated with women’s speech. These findings have led previous research to ask whether pitch contrast between modal voice and creak or, alternatively, social biases have a stronger influence on listeners’ creak identification. Results, however, have been equivocal. We use a set of highly controlled stimuli to address this question and find that the strongest influence on creak identification by AusE-speaking listeners is their expectation about typical pitch range according to speaker gender, rather than pitch or social biases alone. The remaining chapters in this thesis explore a large data set of conversations from 132 teenagers recorded around Sydney, using the Union method to identify creaky voice. Firstly, we examine the effects of prosodic context on creak prevalence. Previous research has identified creak as a marker of finality in British English, American English and other languages such as Spanish and Finnish. Our results suggest that creak also marks finality in AusE. Additionally, creak use decreases throughout a conversation in our data, which we suggest could be a task effect, and that creak appears to be more prevalent in male than female speech. We then examined whether speakers converge in their creak use across a conversation. Findings confirmed that convergence was present for these speakers and that rates of creaky voice use were more similar among females than among males. Finally, we explore creak prevalence among Sydney teenagers and discuss findings with regard to ethnic diversity and gender. We see that the stratification of creaky voice use across macro social categories in AusE is complex and more nuanced in how it relates to speaker gender than results suggested in our analysis of prosodic context. We also present a qualitative analysis in which we investigate how/whether speakers use creak to convey social meaning. This thesis provides methodological contributions for large-scale creaky voice research and advances our understanding of the perception and production of creaky voice in AusE.