Functional connectivity is a measure of how synchronized different brain regions are over time. Rather than tracing physical wires between areas, it captures statistical patterns: when one region becomes active, does another region reliably become active at the same time? Two regions with highly correlated activity are considered functionally connected, even if no direct fiber bundle links them. This concept has become central to understanding how the brain organizes itself into networks and how those networks change in conditions like Alzheimer’s disease, ADHD, and depression.
How It Differs From Structural Connectivity
The brain has two kinds of connectivity that researchers care about. Structural connectivity refers to the physical architecture: the bundles of nerve fibers that wire one region to another, like cables running between buildings. Functional connectivity, by contrast, refers to the statistical relationship between activity patterns in separate regions. Two areas can be functionally connected without a direct structural link between them, because signals can travel through intermediate regions or emerge from shared inputs.
Structural wiring does shape functional connectivity, but not in a one-to-one way. Brain regions with strong physical connections tend to show correlated activity, yet the degree of overlap varies substantially. Think of it this way: the roads between two cities (structure) influence how much trade flows between them (function), but trade patterns also depend on economics, demand, and relationships that the roads alone don’t explain.
How Functional Connectivity Is Measured
The most common tool is functional MRI, or fMRI. It doesn’t measure brain cell activity directly. Instead, it detects changes in blood oxygenation. When neurons in a region fire, nearby blood vessels dilate and deliver extra oxygenated blood, a process called neurovascular coupling. Oxygenated and deoxygenated blood have different magnetic properties, and fMRI picks up that difference. The resulting signal is called BOLD, for blood-oxygen-level-dependent.
To calculate functional connectivity, researchers record these BOLD signals from many brain regions simultaneously, then check how closely the signals rise and fall together over time. The standard metric is the Pearson correlation coefficient, which quantifies how linearly related two signals are. A correlation near +1 means two regions activate and deactivate in lockstep; near zero means their activity patterns are unrelated. More advanced methods exist to capture nonlinear or time-varying relationships, but Pearson’s correlation remains the workhorse of the field.
fMRI isn’t the only option. EEG (electroencephalography) measures electrical activity at the scalp and can assess connectivity through a metric called coherence, which captures how synchronized brain waves are between electrode sites. MEG (magnetoencephalography) does something similar using magnetic fields. These techniques pick up brain activity on a millisecond timescale, far faster than fMRI, but they’re less precise about where in the brain that activity originates. Each method emphasizes different spatial scales and source orientations, so researchers sometimes combine them.
The Brain’s Major Functional Networks
When researchers scan people lying quietly in a scanner with no particular task, distinct groups of brain regions consistently show synchronized activity. These are called resting-state networks, and several have been mapped reliably across thousands of people.
The default mode network (DMN) is the most studied. It includes the medial prefrontal cortex, posterior cingulate cortex, and angular gyri on both sides of the brain. This network is most active during internally directed thought: daydreaming, remembering the past, imagining the future, thinking about yourself. When you shift attention to an external task, the DMN typically quiets down.
The salience network (sometimes called the ventral attention network) does roughly the opposite. It includes the anterior insula and anterior cingulate cortex and becomes active when you detect something important in your environment, whether that’s a loud noise, a surprising face, or a task that demands your attention. It acts as a switch, helping the brain transition between internal focus and external engagement.
Other well-established networks include the frontoparietal network (involved in executive control and decision-making), the dorsal attention network (sustained focus on external stimuli), and several sensory-motor networks handling vision, hearing, and movement. These networks don’t operate in isolation. The connections between them, called inter-network connectivity, matter as much as the connections within them.
What Changes in Disease
Disrupted functional connectivity patterns have been identified in a wide range of neurological and psychiatric conditions, making them potential biomarkers for disease.
In Alzheimer’s disease, the default mode network shows some of the earliest changes. Within-network connectivity tends to weaken, meaning regions that normally synchronize together start falling out of step. At the same time, connections between the DMN and other networks, particularly sensory-motor and visual networks, can paradoxically increase. A similar pattern appears in the frontoparietal network: stronger connections to outside networks are associated with higher levels of Alzheimer’s-related proteins in the brain, while intact internal connectivity is associated with lower pathology. Higher dementia severity has been linked to this combination of increased between-network connectivity and reduced within-network connectivity.
In ADHD, the interaction between the default mode network and the salience network is thought to play a central role in inattention symptoms. When the DMN doesn’t quiet down properly during tasks requiring external focus, or when the salience network fails to flag important stimuli, attention suffers. Abnormal resting-state connectivity between these two networks has been documented in adults with ADHD.
Static vs. Dynamic Connectivity
Traditional functional connectivity analysis is “static.” It takes the entire recording session, often 5 to 15 minutes, and calculates a single average correlation between each pair of brain regions. This gives a useful summary but assumes that connections stay constant throughout the scan.
Dynamic functional connectivity relaxes that assumption. Using a technique called sliding-window analysis, researchers calculate correlations over short segments (often around 30 to 60 seconds), then slide the window forward in time and recalculate. This reveals how the brain’s network organization shifts moment to moment. Some region pairs that look moderately connected on average may actually alternate between periods of strong synchrony and periods of independence. Studies in autism spectrum disorder, for example, have used 50-second sliding windows to uncover connectivity dynamics that static measures miss entirely.
The distinction between static and dynamic connectivity can feel somewhat arbitrary, since even “static” measures are derived from a time series that’s always changing. But dynamic approaches capture a layer of information that averages wash away, and they’ve become increasingly popular over the past decade.
Why Head Motion Is a Problem
One of the biggest technical challenges in functional connectivity research is head motion. Even tiny movements, fractions of a millimeter, can create artifacts that look like real neural signals. Since 2012, when three independent research groups published findings on the issue nearly simultaneously, the field has recognized that motion can systematically bias connectivity measurements.
The bias works in a specific way: motion inflates connectivity estimates between nearby regions while reducing apparent connectivity between distant regions. This is especially problematic in studies comparing groups that move differently in the scanner, such as children versus adults, or patients with a neurological condition versus healthy controls. If one group moves more, the connectivity differences you measure may reflect motion, not the brain. Some earlier published findings have been reevaluated because of this issue.
Researchers now use sophisticated correction strategies, including statistical models that remove motion-related signals and “scrubbing” methods that throw out the most motion-contaminated time points. Models using 36 parameters combined with scrubbing perform well across multiple benchmarks, though no method completely eliminates motion-related noise. These corrections have become standard practice, but they’re a reminder that functional connectivity findings need to be interpreted carefully.
The Human Connectome Project
The largest effort to map the brain’s functional architecture has been the Human Connectome Project, launched in 2009 and funded with $40 million from the National Institutes of Health. Over five years, researchers comprehensively mapped long-distance brain connections in 1,200 healthy young adults using multiple imaging technologies combined for the first time. Beyond brain scans, they collected genetic, demographic, and behavioral data to examine how genes and environment influence connectivity patterns.
The project drove real technical advances: researchers developed a faster, more powerful MRI scanner system that shortened scan times while maintaining high resolution. They also created new data-sharing tools and analytic methods that the broader research community continues to use. The project has since expanded to map brain connectivity across the entire human lifespan, providing a reference atlas against which disease-related changes can be compared.