Past Event

“The structure of variability in spontaneous speech: evidence from voice onset time”

Dr. Morgan Sonderegger
(McGill University)

How variable is speech? Decades of work have shown that in spontaneous speech, the acoustic cues (such as voice onset time (VOT) or vowel formants) used to realize sounds are subject to massive variation as a function of many factors, such as coarticulation, speaking rate, and even time. At the same time, significant enough variability would endanger the reliability of these phonetic cues for perceiving the phonological contrasts they signal. This tension suggests the broad questions: how much variability is there, and how is it structured (with respect to contrasts), in spontaneous speech? We still know relatively little about this, in part because of the difficulty of acquiring enough data to tease apart the effects of the many sources of variability in spontaneous speech. I will present work bearing on this question for the case of VOT, which serves as the primary cue to (phonological) voicing in English. A semi-automatic method for highly accurate VOT measurement (AutoVOT) is used to obtain two large datasets of VOT for voiced and voiceless stops in spontaneous British English speech, one from individuals over three months (Big Brother corpus: Sonderegger, 2012), and one from the same community at two points in time (1970s, 2000s:

/Sounds of the City/ corpus: Stuart-Smith et al., in press). These datasets allow us to model VOT as a function of both synchronic factors (such as place of articulation and speaking rate) and diachronic factors.

Synchronically, the general finding is that a variety of factors strongly affect VOT for voiced and voiceless stops, resulting in decreased contrast (less difference between voiced and voiceless stops) for higher speaking rate, higher frequency words, etc.; that is, when speakers are engaged in hypospeech (Lindblom, 1990). However, even for extremely casual speech, the VOT difference never decreases enough to endanger the contrast. Diachronically, variability in the VOT contrast is ubiquitous: both within individual speakers from day to day, and in a community over 30 years. However, in neither case is the /amount/ of variation in VOT enough to endanger its role as a cue to stop voicing.

The general picture that emerges is that variability in spontaneous speech (at least for VOT) is ubiquitous, but bounded, given the importance of maintaining contrasts.

(Different parts of this work are in collaboration with Joseph Keshet, Thea Knowles, Jane Stuart-Smith, Rachel MacDonald, and Tamara Rathcke.)

About the Presenter

Dr. Morgan Sondereggers received his PhD from the University of Chicago in 2012, with a dissertation on the phonetics of reality television. His research focuses on sound change in progress, phonetic accommodation, and laboratory phonology.