The effects of delayed auditory feedback on speech sequencing: acoustics, physiology, and computational modeling

Trustees Of Boston University, Boston

Investigators

Abstract

Healthy individuals continuously monitor their sound output while speaking, allowing them to perceive and correct errors quickly. When the content or timing of this auditory feedback signal is experimentally altered, most speakers rapidly modify their speech. This project will primarily address how the timing of auditory feedback impacts the speech production system and will assess individual differences in the use of auditory feedback. Results will guide development of a computational model of speech planning and production that has the potential to accelerate the design of new therapeutic approaches to developmental and acquired disorders of speech. Additional benefits to society will accrue from a series of teaching modules that will be developed that explore the role of auditory feedback in normal and disordered speech. The project team will also develop educational materials that bridge neuroscience and communication sciences and that provide formative experiences for undergraduate and K-12 students, including those from disadvantaged backgrounds and with disabilities, to attract them to future study in this area. All project materials will be made available as a public database. The investigators will use a combination of experimental and computational studies. The first research goal is to determine the role of delayed auditory feedback (DAF) on speech sequencing using multimodal experimental techniques. Acoustic data will be collected and analyzed to determine the impact of DAF on the production of non-word syllable sequences and sentences. These data will be used to assess the types of errors elicited by DAF and the relationship between these effects and responses to rapid shifts in the frequency content, rather than timing, of the feedback signal. The collected data and perceptual evaluations, obtained via a crowdsourcing platform, will enable the study of individual speaker variability. Data will also be made available as a large public database of speech under altered feedback conditions. The second research goal is to develop a computational model, informed also by kinematic measures of speech from a subset of participants, that can capture the interaction of auditory feedback with speech sequence production. The model will simulate a series of hypotheses about the interactions among speech perception and planning and production components and assess their ability to account for the patterns of errors observed in the experimental data. The model will also assess the sources of individual variation in susceptibility to the effects of DAF, as observed experimentally.