Motor Learning of Articulator Trajectories in the Production of Novel Utterances

In this study EMA has been used to observe speech articulator movements during successive productions of a variety of novel polysyllabic nonsense words conforming to English phonotactics. Analysis of sensor trajectories comparing initial and final repetitions shows in general reduction of overall duration and distance travelled, lower variability, and fewer acceleration peaks. Comparison of consonant closure timings delimited using velocity extrema suggests that as fluency increases, the overlap between adjacent consonantal gestures also increases.