PhD dissertation: The dynamical foundation of articulation in stuttered speech

Stuttering is a fluency disorder characterized by repetition, prolongation, and blocking of speech sounds, especially at syllable-initial consonants. While these characteristics are perceptual features of stuttering, the articulatory behaviors that give rise to them have been poorly understood. Without knowledge of the articulatory manifestation of stuttering, the foundational nature of the behaviors underlying perceived disfluencies remains elusive, and theories of stuttering are prevented from being empirically and comprehensively grounded. This dissertation addresses this opportunity for rigor in the articulatory domain by using real-time MRI during naturalistic speech production tasks to dynamically image vocal tract movements of seven adults who stutter. Leveraging the theoretical frameworks of Articulatory Phonology and Task Dynamics, the kinematic properties of syllable onset gestures and their coordination with syllable nucleus gestures during stuttering disfluencies were analyzed and modeled. Results reveal that percepts of stuttering disfluencies are underpinned by the oscillation or fixation of the primary oral articulator for the syllable onset constriction. These atypical movements do not arise, as some theories have proposed, due to the syllable’s nucleus not being prepared in time. The articulatory gesture for the nucleus is in fact already appropriately initiated during stuttering moments, though a solely perceptual record cannot reveal this. We argue that oscillation and fixation may not be coping strategies but are instead integral to the core processes that cause stuttering. Specifically, we develop a novel dynamical model of stuttering events as resulting from a qualitative shift in the behavior of an underlying nonlinear dynamical system. This shift occurs when the system’s control parameter governing the temporal properties of gestures falls into a specific infelicitous region. We speculate that such dynamics, via a variation in control parameter(s), may in fact drive both disfluent and typical speech production. We conclude that real-time articulatory data combined with dynamical systems modeling can provide insight to advance the scientific understanding of stuttering and general speech fluency.

For the full dissertation, please reach out here.

Publications (ongoing):