Functional, Effective and Directed Connectivity

Functional connectivity denotes any statistical dependence between two signals and is symmetric and undirected: coherence, phase-locking value, and correlation all measure that two regions co-vary without saying which drives which. Effective connectivity denotes a directed causal influence of one region on another, estimated by models that ask whether one signal helps predict or explain another (Granger causality, transfer entropy, partial directed coherence, dynamic causal modelling). Directed connectivity is the umbrella for these asymmetric measures.

The graduate-level caution is that none of these estimate anatomical connections, and directedness in a statistic is not physiological causation. A common unobserved input driving both regions with different delays will masquerade as a directed link; volume conduction and a shared reference create spurious zero-lag coupling; and downsampling or additive sensor noise can even reverse the inferred direction of Granger causality. Reported connectivity is therefore a statement about a model fitted to particular signals, not a wiring diagram.