Motor-based prediction mediates implicit vocal imitation

Abstract

Phonetic convergence—the unconscious adaptation of one's speech to resemble that of an interlocutor—is thought to arise from predictive mechanisms. Two types of predictions have been proposed to modulate others’ speech: memory-based predictions, which reduce sensitivity to acoustic features reflecting a speaker's vocal identity, and motor-based predictions, which are grounded in the listener's own vocal characteristics. Compared to a relatively well-established role of memory-based predictions, whether motor-based predictions suppress or enhance sensitivity to listener-matched predicted features and how they contribute to phonetic convergence remain unclear. In the present study, we examined these processes using a novel speaking oddball task in which participants were randomly prompted to repeat words they heard. Auditory mismatch negativity served as a neural index of mismatch detection. Prior to the oddball task, participants were divided into a shadow group—engaging in an additional shadowing task to promote vocal convergence—and a non-shadow group that did not receive such exposure. EEG analyses revealed that motor-based predictions enhance sensitivity to listener-matched predicted features following convergence behaviour, with this enhancement correlating with greater vocal convergence. Our novel oddball design provided an efficient method for revealing the dynamic interplay between internal predictive signals and external inputs that mediates phonetic convergence. These findings challenge the view that motor-based predictions only suppress neural responses to predicted features, and instead highlight their potential role in enhancing perceptual learning and guiding vocal adjustments. Motor-based predictions orchestrate sensorimotor interaction and memory-based operations to mediate implicit learning behaviour in a social context.