Neuroimaging can reveal large-scale information flows in the cortex. The "intrinsic" information flow pattern, emerging in resting-state paradigms, is perturbed by several pathologies. By applying covariance-based Granger causality, we revealed several global changes in intra- and inter-hemispheric information flows in stroke patients. However, such changes can be due to different causes : changes in excitability/input to each area, changes in effective connectivity, or hemodynamic anomalies. In this seminar, we report on our findings and discuss how the trilemma of causes could be solved by applying effective connectivity models.