A Systematic Characterization of Causal Interactions Between Human Visual Areas

Abstract

Human visual cortex comprises three interacting streams, but anatomy and correlated activity cannot determine the direction or reliability of interareal influence. Characterizing directional interactions among these streams can reveal the architecture through which activity can propagate across the visual system. Here, we used single-pulse electrical stimulation during intracranial EEG recordings in 23 patients to map directed effective connectivity among 22 atlas-defined visual cortical areas. The resulting effective connectivity matrix revealed a selective and asymmetric architecture of interareal influence. Feedforward influences from early visual areas to the dorsal and lateral streams were more reliable and more prevalent than the corresponding feedback influences, whereas interactions between early visual and ventral temporal areas were comparatively balanced. Cross-stream interactions favored temporal-to-parietal over parietal-to-temporal influence, with higher response reliability and a greater proportion of significant responses among sampled connections. Network-level profiles were consistent with source-like organization in early visual areas and the ventral stream, and integrator-like organization in the dorsal and lateral streams. Visual-task data collected in three of the same patients illustrated how stimulation-derived connectivity may relate to functional responses across connected visual regions. These findings reveal a directional architecture for activity propagation across the human visual cortex and provide empirical constraints for biologically grounded models of visual processing.