Background: Speech perception relies on a wide range of neural time windows to integrate and segregate diverse neural inputs derived from auditory signals. Such a capacity to process temporal information is represented by the intrinsic neural timescale (INT), which quantifies neural dynamics in local brain regions. While previous studies have shown that multi-timescale processes in the language-related cortex underpin speech perception, how the processes in subcortical areas, particularly the inferior colliculus (IC) and medial geniculate body (MGB), affect speech perception, and whether mismatches of the timescales across cortical and subcortical areas disturb it, are unknown.

Rationale: Investigating the relationship between INT of language-related cortical areas and subcortical areas may clarify how these regions coordinate to support speech perception, especially in noisy environments.

Methods: To explore the relationship, we collected behavioural and functional magnetic resonance imaging (fMRI) data from 31 healthy right-handed participants, all of whom provided written informed consent in accordance with protocols approved by the University of Tokyo. We first recorded fMRI signals from participants during a sentence recognition task with varying levels of background noise (from −12 dB to 12 dB SNR); then, by applying hierarchical clustering analysis to noise-dependent brain activity patterns, we determined three primary cortical clusters: auditory-language, somatosensory, and fronto-parietal cortical networks. Along with this experiment, we calculated the INTs of these clusters and the bilateral IC and MGB (subcortical components) by examining the autocorrelation of resting-state fMRI signals recorded from the same participants. Speech recognition performance in the behavioural task was measured using the thresholds of a psychometric function.

Results: Partial correlation analysis indicated that the INT of the auditory-language cluster did not correlate with the MGB, but it did correlate with the other cortical clusters, the IC, and speech recognition performance. This finding underscores the auditory-language cluster's role as a hub that orchestrates information processing across the cortical and midbrain network. Furthermore, individuals with higher speech recognition performance had shorter INTs in the auditory-language cluster but longer INTs in the IC. In fact, individuals with a larger INT mismatch between the auditory-language cluster and the IC were likely to exhibit poor speech recognition performance.

Conclusions: This study emphasizes that a balance between cortical and subcortical INTs is essential for accurate speech processing, pointing to the auditory-language cluster as a hub facilitating efficient information transfer between cortical and subcortical areas.