This study investigates articulatory dynamics in Mandarin sibilant-vowel sequences (/si/, /zi/, /ri/) using Generalized Additive Mixed Models (GAMMs) applied to ultrasound tongue imaging data. Polar coordinates from tongue surface tracings were converted into spatiotemporal representations, enabling dynamic analysis of radial distance differences across normalized production intervals and angular positions along the tongue contour. Results reveal complementary articulatory patterns with distinct spatiotemporal organization. For /si/, positive differences in the tongue root region during early production indicate retraction for vowel resonance, while negative differences at the tongue tip reflect post-fricative lowering. /zi/ exhibits similar but attenuated patterns. Retroflex /ri/ shows pronounced radial extension near the tongue tip during early phases, while vowel influence emerges at higher angles during later phases. These findings quantitatively capture fluid sub-segmental coarticulatory transitions. This methodology advances articulatory phonetics by revealing dynamic patterns fundamental to Mandarin phonology, providing empirical evidence for tongue root-tip coordination defining sibilant and retroflex natural classes. Results have broader applications for cross-linguistic analyses, clinical speech assessment, and computational speech models.
Feng-fan Hsieh (Wed,) studied this question.