Rapid Auditory Feedback Control of Speech is Based on the Current Production, Not an Idealized Target

Purpose: The concept of an auditory target that is monitored by the sensory-motor system during speech is central to most models of speech production, yet little is known about the nature of these auditory targets. The goal of this study was to determine whether the auditory feedback controller for speech uses “idealized” acoustic targets or simply attempts to correct deviations from the motor commands to the speech articulators.Method: We investigated the stability over time of the auditory targets used for online auditory feedback control (AFC) of fundamental frequency (fo) and first formant frequency (F1) using a combination of auditory perturbation experiments and computational modeling. Twenty-three young adults participated in reflexive fo and F1 perturbation experiments across four different days: two mornings and two afternoons. Results: Analysis of unperturbed productions across participants indicated that baseline fo varied with session number, and baseline fo and F1 varied with time of day. The pattern of the participants’ compensatory responses to upward and downward fo and F1 perturbations across sessions with low and high baseline fo/F1 values strongly indicated that the auditory target used for AFC differed between low and high baseline sessions. Computer simulations of the SimpleDIVA model were used to quantitatively compare this variable-target model (in which AFC utilizes a different target in different sessions) to a single-target model; these simulations verified the superiority of the variable-target model.Conclusions: Our findings indicate that the auditory targets for AFC arise from predictions generated from the ongoing motor command, i.e. the efference copy, rather than an idealized and optimal auditory target, and they have important implications for models of speech motor control.