Long-distance amplitude correlations in the high γ band reveal segregation and integration within the reading network.

Reading sentences involves a distributed network of brain regions acting in concert surrounding the left sylvian fissure. The mechanisms of neural communication underlying the extraction and integration of verbal information across subcomponents of this reading network are still largely unknown. We recorded intracranial EEG activity in 12 epileptic human patients performing natural sentence reading and analyzed long-range corticocortical interactions between local neural activations. During a simple task contrasting semantic, phonological, and purely visual processes, we found process-specific neural activity elicited at the single-trial level, characterized by energy increases in a broad gamma band (40-150 Hz). Correlation analysis between task-induced gamma-band activations revealed a selective fragmentation of the network into specialized subnetworks supporting sentence-level semantic analysis and phonological processing. We extend the implications of our results beyond reading, to propose that gamma-band amplitude correlations might constitute a fundamental mechanism for large-scale neural integration during high-level cognition.