Cortico-subcortical neural timescale alignment underlies speech recognition in noise

Speech recognition recruits the auditory-language brain system for accurate segregation and integration of complex linguistic information that consists of various acoustic signals with various timescales. However, whether and how such multi-timescale auditory information processing is linked with the neural timescale of the auditory-language system remains unclear. To answer this, we compared behavioural performance on a speech-in-noise recognition test with the intrinsic neural timescales (INTs) of the auditory-language system. We found that the speech recognition ability was not correlated with the INT of the entire auditory-language network but was enhanced when the INT of the frontotemporal sub-system in the auditory-language network was close to that of the inferior colliculus, an auditory midbrain region. Moreover, this cortico-subcortical INT similarity was associated with the task-related neural activity in the frontotemporal sub-system. These findings indicate that the better cortico-subcortical INT alignment facilitates the activation of the auditory-language cortical system, resulting in accurate speech processing.