Speech tracking in auditory and motor regions reflects distinct linguistic features

Abstract

AbstractDuring online speech processing, our brain tracks the acoustic fluctuations in speech at different time-scales. Previous research has focussed on generic time-scales (for example, delta or theta bands) that are assumed to map onto linguistic features such as prosody or syllables. However, given the high inter-subject variability in speaking patterns, such a generic association between the time-scales of brain activity and speech properties can be ambiguous. Here, we analyse speech tracking in source-localised magnetoencephalographic data by directly focusing on time-scales extracted from statistical regularities in the speech material. This revealed widespread tracking at the time-scales of phrases (0.6 – 1.3 Hz), words (1.8 – 3 Hz), syllables (2.8 – 4.8 Hz), and phonemes (8 – 12.4 Hz). Importantly, when examining the relevance for single-trial comprehension, we found stronger tracking for correctly comprehended trials in the left premotor cortex at the phrasal scale, and in left middle temporal cortex at the word scale. Control analyses using generic bands confirmed that these effects were specific to the stimulus-tailored speech regularities. Furthermore, we found that the phase at the phrasal time-scale coupled to beta-power in motor areas. This cross-frequency coupling likely mediates the comprehension effect in the motor system, and implies top-down temporal prediction in speech perception. Together, our results reveal specific functional and perceptually relevant roles of distinct entrainment processes along the auditory-motor pathway. These processes act concurrently at time-scales within the traditional delta band and highlight the role of neural tracking mechanisms that reflect the temporal characteristics of speech.