Megan Thompson, Program in Bioengineering, UCSF, San Francisco, CA and UC Berkeley, Berkeley, CA
Sensory feedback, both auditory and sensorimotor, plays a crucial role in speech production in both healthy individuals and in individuals with speech disorders. However, the vast majority of speech production feedback studies focus on auditory feedback, while the role of the vocal tract is less well-characterized. We investigate the role of the vocal tract during speech production by examining speech production in its absence by training subjects to use a touchscreen-based speech production platform. Contact with the screen produces different vowels depending on the touch location, with every possible vowel within a wide formant range producible. As subjects performed the experiment, both accuracy (distance between the target and response) and consistency (distance between each response to the same target) rapidly improved within 30 trials. Further, post-training, subjects were able to respond to novel targets with significantly greater accuracy than pre-training target responses, indicating that the training generalized to other speech sounds. When the map was subjected to a post-training 150Hz F2 shift, subjects shifted their response to certain targets to compensate, just as they generally compensated to formant alterations when they performed an analogue of the experiment vocally. Stable, rapid increases in both precision and accuracy, generalization to novel targets, and adaptation to changes imply the development of an internal sensorimotor map, allowing subjects to predict the auditory consequences of each touch and to develop the stereotyped responses necessary to achieve the desired feedback. This provides evidence that healthy adults are capable of rapidly learning a new platform of speech production without vocal tract feedback that bears similarities to vocal speech. The neural similarities of the task to vocal speech are currently being further examined through neuroimaging performed concurrently with the task.