Gradiency vs categoricity in the production of prosodic boundaries as reflected in different kinematic measures

Abstract

Two questions are examined in an EMA experiment: (1) are speech gestures at prosodic boundaries produced in a categorical or in a gradient manner (extending work by Krivokapic, Mooshammer, and Tiede 2013), and (2) how do different articulatory measures reflect boundary strength. Forty-eight sentences were constructed, each containing one, two, or three prosodic boundaries, for a total of 56 boundaries per sentence set. The range of predicted boundary types varied from a weak clitic boundary to a strong sentence boundary. Each boundary fell between the words “column and.” Seven subjects read six to eight repetitions of these sentences. Various temporal properties of boundaries have been examined, including the duration of the lip opening movement for [m], which is the movement closest to the boundary (and therefore most strongly reflects the boundary properties), and the duration of the jaw movement from the opening movement for the first vowel in “column” to the opening movement for the postboundary vowel (a variable which spans the boundary). These measures have been subjected to Gaussian mixture model analysis. Results show evidence for a more fine-grained structure than can be predicted by the three prosodic levels of traditional models. [Work supported by NIH DC003172-16, DC008780, DC002717].