Whole-Brain Millisecond-Scale Effective Connectivity of Auditory and Visual Naming

Clinical mapping studies in epilepsy and brain tumor surgery have informed neurobiological models of speech. However, a comprehensive model requires clarification of when, for how long, and in what directions cortical regions transmit information to support processes ranging from perception and comprehension to lexical retrieval and articulation. We aimed to characterize whole-brain patterns of directional neural information flow during auditory and visual naming.

We studied 127 patients who performed auditory and picture naming tasks during intracranial EEG recordings, analyzing 9,526 artifact-free nonepileptic recording sites. We applied transfer entropy-based effective connectivity analysis to estimate information flow between brain regions. Information flow was classified as accelerating when increases in group-level high-gamma amplitude in one region predicted subsequent increases in another region connected by white matter, and decelerating when decreases predicted subsequent decreases in the downstream region. The functional relevance of these information flow patterns was evaluated using electrical stimulation mapping and behavioral response times.

Following stimulus onset, accelerating flows emerged from modality-specific sensory cortices and propagated to higher-order regions before becoming bidirectional. These transient accelerating flows were typically followed by decelerating phases, during which new accelerating flows arose. Stronger accelerating flows at specific time points predicted a higher probability of stimulation-induced symptoms (Spearman's rho: 0.54-0.81; p<0.00001). During auditory naming, faster responses were associated with stronger accelerating flows within left perisylvian pathways, whereas during visual naming faster responses were associated with stronger accelerating flows within bilateral basal temporal pathways.

Strong accelerating flows, followed rapidly by decelerating flows, predicted stimulation-induced symptoms and faster behavioral responses, indicating that information transfer through white matter supports sequential transitions from perception to lexical retrieval and articulation. By resolving the direction of neural propagation through white matter pathways, this study provides a whole-brain model of temporally precise, anatomically specific, and directional network interactions underlying human speech.