Poster C60, Wednesday, August 21, 2019, 10:45 am – 12:30 pm, Restaurant Hall
SimpleDIVA: A 3-Parameter Model for Examining Adaptation in Speech and Voice Production
Elaine Kearney1, Alfonso Nieto-Castañón1, Ayoub Daliri2, Frank H Guenther1,3,4;1Boston University, 2Arizona State University, 3Massachusetts Institute of Technology, 4Massachusetts General Hospital
Sensorimotor adaptation paradigms have become an important experimental technique in examining the neural mechanisms of motor control, including speech and voice production. In a typical adaptation paradigm, participants produce speech while they receive perturbed auditory feedback (e.g., shifts in formants or fundamental frequency). When the perturbations sustain over several trials, participants gradually learn to adjust their movements to compensate for the perturbation (i.e., participants adapt). This process relies on an interplay between feedback control (in detecting and correcting errors within a trial) and feedforward control (in updating the motor command for the following trial). However, it is challenging to determine the relative contribution of each system based on behavioral data alone. Here, we describe a simple 3-parameter computational model (SimpleDIVA) that estimates the relative contribution of feedback and feedforward control mechanisms to sensorimotor adaptation. The model is based on the DIVA model of speech production (Guenther, 2006; 2016) and the three parameters reflect the three subsystems underlying speech motor control, specifically, auditory feedback control, somatosensory feedback control, and feedforward control. The model is tested through computer simulations that identify optimal model fits to six existing datasets (Abur et al., 2018; Ballard et al., 2018; Chao & Daliri, unpublished data; Daliri et al., 2018; Haenchen et al., 2017; Heller-Murray, 2019). Through the simulations, we show how SimpleDIVA can be used in the interpretation of adaptation experiments involving first and second formants and fundamental frequency. The results highlight the model’s sensitivity to changes in experimental protocols, for example, when using masking noise (to eliminate the effect of auditory feedback) and when perturbing more than one auditory dimension at the same time (e.g., first and second formants). The model also captures the differential role of feedback control and feedforward control early and late in the production of a trial. The final property of the model revealed through the simulations is its power in predicting average group responses from one experimental condition to another. To illustrate this, we model data from an adaptation paradigm with a gradual onset of the perturbation and use the resulting parameters to predict performance in a paradigm with a sudden perturbation onset (from the same groups of participants). Across all simulations, the model fits were excellent and showed strong positive correlations with the experimental data (range of Pearson correlation coefficients = .86 – .97). SimpleDIVA offers new insights into speech and voice motor control by providing a mechanistic explanation for the behavioral responses to the adaptation paradigm that are not readily interpretable from the behavioral data alone. In future work, the model can be used to develop clear, testable hypotheses that can be evaluated empirically, ultimately advancing our understanding of speech motor control and informing future directions of rehabilitation research for individuals with communication disorders. Compiled SimpleDIVA code, including an easy-to-use graphical user interface, is available to facilitate the use of the model by other groups in future studies (http://sites.bu.edu/guentherlab/software/simplediva-app).
Themes: Speech Motor Control, Computational Approaches
Method: Computational Modeling