| This dissertation examines the issue of tone-to-segment alignment in Mandarin Chinese and aims to provide a unified account of all four Mandarin tones by making reference to the control structure of speech, namely gestural coordination, within the framework of Articulatory Phonology. The alignment patterns reported in previous studies display great diversity within and across languages, and it has often proved problematic to provide a simple, unified account for a wide range of patterns.;A theoretical model that treats tones as combinations of (tone) gestures is proposed in this dissertation. Two invariant tone gestures, targeted at H(igh) and L(ow) pitch respectively, are used to represent the four tones in Mandarin. T(one) gestures, like C(onsonant) gestures and V(owel) gestures, are viewed as dynamic events that couple (coordinate) with other gestures in various coupling modes.;Simultaneous kinematic and acoustic data of various Mandarin syllables were collected using EMMA (Electromagnetic Midsagittal Articulometer) in the production experiments. The onset-to-onset lags (i.e. the temporal distance between the onset of the T gesture and the onset of the C or V gesture) were computed and used to define the tone-to-segment alignment patterns in the target syllables. Two different alignment patterns among the C, V and T gestures emerged among the four lexical tones. However, we argue that both patterns can be accounted for with reference to the so-called C-center effect (Brownian & Goldstein, 1988). In Mandarin syllables, the T gesture behaves like an additional onset C gesture: the C and T gestures are coordinated sequentially with one another and are both coupled synchronously with the V gesture. The two different patterns emerge from differences among the four tones regarding their gestural composition. The alignment patterns across a range of syllables and the results from statistical tests provide evidence for the C-center coupling model for Mandarin. In addition, Mandarin syllables were modeled by means of TADA implementations based on the proposed coupling relations among the C, V and T gestures. The outputs are shown to be highly comparable to real speech data, which lends further support to the proposed coupling model. |
