Speech motor control variables in the production of voicing contrasts and emphatic accent
Mills, Timothy Ian Pandachuck
This dissertation looks at motor control in speech production. Two specific questions emerging from the speech motor control literature are studied: the question of articulatory versus acoustic motor control targets, and the question of whether prosodic linguistic variables are controlled in the same way as segmental linguistic variables. In the first study, I test the utility of whispered speech as a tool for addressing the question of articulatory or acoustic motor control targets. Research has been done probing both sides of this question. The case for articulatory specifications is developed in depth in the Articulatory Phonology framework of Haskins researchers (eg. Browman & Goldstein 2000), based on the task-dynamic model of control presented by Saltzman & Kelso (1987). The case for acoustic specifications is developed in the work of Perkell and others (eg Perkell, Matthies, Svirsky & Jordan 1993, Guenther, Espy-Wilson, Boyce, Matthies, Zandipour & Perkell 1999, Perkell, Guenther, Lane, Matthies, Perrier, Vick,Wilhelms-Tricarico & Zandipour 2000). It has also been suggested that some productions are governed by articulatory targets while others are governed by acoustic targets (Ladefoged 2005). This study involves two experiments. In the first, I make endoscopic video recordings of the larynx during the production of phonological voicing contrasts in normal and whispered speech. I discovered that the glottal aperture differences between voiced obstruents (ie, /d) and voiceless obstruents (ie, /t) in normal speech was preserved in whispered speech. Of particular interest was the observation that phonologically voiced obstruents tended to exhibit a narrower glottal aperture in whisper than vowels, which are also phonologically voiced. This suggests that the motor control targets of voicing is different for vowels than for voiced obstruents. A perceptual experiment on the speech material elicited in the endoscopic recordings elicited judgements to see whether listeners could discriminate phonological voicing in whisper, in the absence of non-laryngeal cues such as duration. I found that perceptual discrimination in whisper, while lower than that for normal speech, was significantly above chance. Together, the perceptual and the production data suggest that whispered speech removes neither the acoustic nor the articulatory distinction between phonologically voiced and voiceless segments. Whisper is therefore not a useful tool for probing the question of articulatory versus acoustic motor control targets. In the second study, I look at the multiple parameters contributing to relative prominence, to see whether they are controlled in a qualitatively similar way to the parameters observed in bite block studies to contribute to labial closure or vowel height. I vary prominence by eliciting nuclear accents with a contrastive and a non-contrastive reading. Prominence in this manipulation is found to be signalled by f0 peak, accented syllable duration, and peak amplitude, but not by vowel de-centralization or spectral tilt. I manipulate the contribution of f0 in two ways. The first is by eliciting the contrastive and non-contrastive readings in questions rather than statements. This reduces the f0 difference between the two readings. The second is by eliciting the contrastive and non-contrastive readings in whispered speech, thus removing the acoustic f0 information entirely. In the first manipulation, I find that the contributions of both duration and amplitude to signalling contrast are reduced in parallel with the f0 contribution. This is a qualitatively different behaviour from all other motor control studies; generally, when one variable is manipulated, others either act to compensate or do not react at all. It would seem, then, that this prosodic variable is controlled in a different manner from other speech motor targets that have been examined. In the whisper manipulation, I find no response in duration or amplitude to the manipulation of f0. This result suggests that, like in the endoscopy study, perhaps whisper is not an effective means of perturbing laryngeal articulations.