Differentiating language-specific from language-related network dynamics using dynamic functional connectivity analysis

Language processing is a highly complex dynamic neural activity that requires the recruitment of both regions critical for language processing and large-scale cognitive networks. Because language networks are interconnected and dynamically reconfigure their connections depending on task demand and complexity, it is challenging to differentiate language-specific networks form language-related cognitive networks during both resting-state fMRI (rs-fMRI) and task-based fMRI (tb-fMRI). Prior evidence demonstrated state-dependent differences in dynamic functional connectivity (FC) within identical networks across rest- and task-conditions. Based on this evidence, we hypothesized that examining dynamic FC patterns could help differentiate the functional specialization of large-scale neural networks in the human brain. Thus, we examined whether temporal patterns of dynamic FC could functionally differentiate networks engaged during a language task. Using a data-driven dynamic FC framework, we tested whether rest–task state transitions preferentially modulate subsets of large-scale networks. Specifically, we examined whether monotonic temporal trends in dynamic FC, used as a first-order approximation of sustained task engagement, were more prominent in networks showing task-sensitive dynamics, while others exhibited resting-state-like dynamics. We analyzed rs-fMRI and tb-fMRI data from 172 healthy participants (86 females, 27.6 ± 3.8 years old) from the Human Connectome Project. Dynamic FC analyses revealed significant state-dependent differences in correlation variability between rs-fMRI and tb-fMRI (Bonferroni-corrected p < 0.001). A subset of networks exhibited task-sensitive temporal modulation, whereas other networks showed dynamic patterns closely paralleled resting-like dynamics without reaching statistical significance. These findings suggest that dynamic FC provides a useful framework for functionally distinguishing language-specific networks from language-related cognitive networks.