Estimating the prevalence of creaky voice: a fundamental frequency-based approach

Abstract

Anecdotal claims of increasing prevalence of creaky voice in varieties of English, particularly among younger female speakers, have piqued the interest of sociophonetic researchers, speech pathologists, and public commentators alike. However, studies quantifying creaky voice prevalence are few in number and modest in scale, possibly because manual annotation of creaky voice – the method most often used for its detection – is time-intensive. Since low F0 characterizes most manifestations of creaky voice, it is conceivable that it can be detected, with a high degree of approximation, using an automated F0-based method. This paper describes such an approach, drawing on previous work by Dorreen [7], and explores its application and validity across male and female speakers of Australian English and across speaking tasks. Our findings suggest that our approach is an effective means of estimating creaky voice prevalence, with potential for generating new insights in an area where a reliable evidence base is much-needed.