A functional principal component analysis of acoustic changes under articulatory perturbation

Abstract

The extent and time course of compensation under articulatory perturbation provide a critical test of speech motor control models. This study investigates acoustic adaptation to a bite block using functional principal component analysis (FPCA) of formant trajectories. Fifty-four native Dutch speakers produced pseudowords containing /ɪ/ and /ɛ/ during a pretest and a perturbation phase, with articulatory condition (bite block vs. control) and auditory feedback (normal vs. masking noise) manipulated. FPCA was applied to trial-by-trial trajectories of F1 and F2. The analysis showed no differentiation between bite-block and control conditions in the F1 FPCA space, indicating rapid and near-complete adaptation for vowel height. In contrast, F2 trajectories were affected by the articulatory perturbation: the first principal component separated bite block conditions and revealed gradual, incomplete adaptation across trials. No effects of auditory feedback were observed. These findings suggest dimension-specific compensation, with stable task-level control of F1 and slower adjustment of F2, consistent with models that incorporate both motor equivalence and feedback-based adaptation. FPCA provides a sensitive approach to capturing non-linear dynamics in speech production under perturbation.