Higher-order intrinsic routes support flexible task-evoked communication
Flexible task-evoked communication requires the brain to transiently establish functional interactions that are not evident during rest, yet the network mechanisms supporting these newly formed connections remain unresolved. Using whole-brain functional MRI from 92 participants performing six cognitive tasks, we show that task-evoked communication is implemented within a constrained intrinsic network architecture rather than through large-scale rewiring. Across tasks, we found that approximately 70% of functional connections were preserved from rest, forming a dominant stable core, while task engagement selectively reconfigured a smaller subset of connections, primarily at the between-network level. While existing activity-flow models explain task-evoked information transfer primarily through direct stable core connections, we demonstrate that newly formed connections between regions lacking direct resting-state coupling were supported by selectively strengthened indirect higher-order resting-state routes embedded within the intrinsic scaffold. Network-level analyses revealed a multiscale organization of information transfer: we observed that direct stable routes dominated within-network communication, most prominently in sensory-motor and default mode systems, whereas indirect higher-order routes preferentially supported between-network integration through control and attentional systems. By extending activity-flow modelling to incorporate these higher-order intrinsic routes, we show a significant improvement in the prediction of task-evoked activity at the global, between-network scale, confirming their functional recruitment during task performance. Together, our findings identify higher-order intrinsic routes as a key mechanism enabling flexible task-evoked communication within a stable large-scale network architecture.