Slide Slam R1
Voystick: a vocal joystick for vowel production training
Caroline Niziolek1; 1University of Wisconsin–Madison
Auditory and somatosensory feedback help guide speech movements, but learning novel production targets, such as non-native vowels, can still be a challenge for adult speakers. The current study investigates speakers’ ability to use visual feedback to refine vocal movements during vowel production. Sixteen participants took part in a single-session speech motor learning paradigm in which their spoken acoustics were mapped to a real-time cursor display. Participants used this “voice joystick” (Voystick) to move the cursor to visual targets defined in formant frequency (F1-F2) space. Two targets corresponded to native English vowels (/ɛ/, /ow/), with target formants defined by speaker-specific baseline productions; two targets were novel vowels (/ø/, /ɯ/), defined as linear combinations of formants from the baseline productions. Participants completed 45 5-second trials for each target in a pseudorandom order. To achieve a “hit” for a given trial, they had to vocalize while maintaining vowel formants within 50 mels of the target for at least 300 ms. The experiment instructions encouraged participants to vary their vowels continuously, demonstrating a gradual continuum between /i/ and /u/, and specified that changes to voice pitch would not affect the cursor movement. Formant tracks were visually inspected, and mistracked trials or segments of trials were excluded from further analysis. Over the course of the experiment, speakers showed evidence of motor learning, achieving vowel target “hits” more often and more quickly, and spending proportionally more time in the vicinity of the targets, as defined by the average distance over the course of the trial (all p < 0.005). Native and novel vowels did not differ in the percentage of targets reached or in the average time to target, but native vowels achieved a closer average distance to the target by the end of the experiment than novel vowels (p = 0.002). This work establishes the capacity of the Voystick system to promote speech motor learning, and has the potential to investigate the formation of vocal motor programs in the absence of auditory feedback.