Brain processes of loud speech and faking an accent as a window on motor speech planning/programming

Numerous factors interfere with the successful transmission of messages during verbal conversations leading speakers to use different speech modes, i.e. specific prototypes of speech with unique phonatory and articulatory characteristics. Despite their omnipresence in verbal exchanges, no theoretical model in the speech production literature has provided a mechanistical account of the encoding processes underpinning speech in different modes. The present study thus aims at proposing how speech modes are planned/programmed relative to standard speech by using the high temporal resolution provided by electroencephalography (EEG). Participants uttered pseudowords in three different conditions - standard speech, speaking louder than usual and faking an English accent in French - during a delayed production task. Event related potential (ERP) of standard speech was contrasted separately with the two non-standard speech modes. Results indicate that speaking by adopting speech modes varying in articulatory and phonatory properties entails increased neural activity of the brain networks that are already involved in standard speech production. Especially, electrophysiological signatures of loud speech and faking an accent were both associated to differences in ERP responses relative to standard speech in a time period covering the last 200 ms preceding the vocal onset. This observation was coherent across waveform analysis (with more extended differences in time and space for faking an accent), topographical dissimilarity analysis and microstates analysis (mode extended for loud speech). The findings highlight that different speech modes are encoded in the last 200 ms preceding their vocal production, possibly in a mode-specific way which will need further investigation.