What Is Functional Connectivity?

Functional connectivity is the statistical dependence between neurophysiological signals recorded from spatially separated regions of the brain. It is an observational quantity: two regions are said to be functionally connected when their measured activity covaries reliably over time, without any claim about the anatomical pathway that links them or about which region drives the other. That makes it distinct from structural connectivity, which describes physical fiber pathways reconstructed from diffusion imaging, and from effective connectivity, which models directed causal influence within an explicit generative framework.

The concept became central to systems neuroscience after 1995, when correlated low-frequency fluctuations in the blood oxygen level dependent signal were observed between left and right motor cortex in subjects lying still and performing no task. Those fluctuations, concentrated below about 0.1 hertz, turned out to organize into reproducible spatial patterns now referred to as resting-state networks, including the default mode, salience, dorsal attention, frontoparietal control, sensorimotor, and visual networks.

Measurement Modalities

Resting-state functional MRI is the dominant method, and it was one of the two primary modalities acquired by the Human Connectome Project, which collected long, high-resolution scans in more than a thousand healthy adults and released them for open reuse. The protocol paper describing that acquisition sets out the multiband accelerated sequences and preprocessing choices that have since become common practice. Electrophysiological methods measure the same underlying phenomenon at much finer temporal resolution: electroencephalography and magnetoencephalography support coherence, phase locking, and amplitude envelope correlation across frequency bands, and invasive electrocorticography provides the same with far better spatial specificity in surgical patients. Functional near-infrared spectroscopy offers a portable hemodynamic alternative for infants and for naturalistic settings. Animal work extends the approach across species, and resting-state connectivity mapping in animal models allows the hemodynamic measurement to be validated against simultaneous electrophysiology.

Analysis Methods and Network Models

The simplest estimator is a Pearson correlation between the time series of a seed region and every other voxel, producing a connectivity map for that seed. Data-driven decompositions such as independent component analysis recover multiple networks at once without requiring a seed, and dual regression projects group components back onto individual subjects for comparison. Parcellating the brain into a few hundred regions and computing all pairwise correlations yields a connectivity matrix that can be treated as a weighted graph, opening the way to measures of modularity, path length, clustering, and hub centrality. Partial correlation and regularized inverse covariance estimation address the problem that marginal correlations conflate direct with indirect coupling. Sliding-window and state-based methods relax the assumption that connectivity is stationary across a scan, treating it instead as a sequence of recurring configurations.

Reliability and Confounds

Connectivity estimates are sensitive to nuisance signals that resemble neural coupling. Head motion introduces spurious distance-dependent correlation, and cardiac and respiratory cycles inject structured variance that aliases into the sampled signal. Standard mitigations include motion parameter regression, censoring of corrupted frames, physiological noise modeling, and component-based nuisance removal, although global signal regression remains contested because it shifts the correlation distribution. Reliability also depends strongly on how much data is collected, and analyses of scan length and shrinkage in Human Connectome Project data show that individual-level estimates stabilize only with substantially more acquisition time than early studies used.

Applications

Functional connectivity analysis has applications in a range of fields, including:

• Presurgical mapping of eloquent cortex in epilepsy and tumor patients
• Candidate biomarkers for Alzheimer disease, depression, and schizophrenia
• Assessment of consciousness in unresponsive patients
• Developmental and aging studies of network maturation and decline
• Target selection for neuromodulation and deep brain stimulation
• Brain-computer interface design and neurofeedback