What Are Brain Networks and How Do They Function?

Brain networks are groups of anatomically or functionally linked brain regions that coordinate their activity to produce everything from perception and movement to abstract thought and daydreaming. Rather than any single region handling a task alone, the brain parcels work across distributed circuits whose members fire in synchrony, even when they sit on opposite sides of the skull. Decades of neuroimaging have identified several reliable networks, and the interplay among them turns out to explain a surprising amount about cognition, mental health, and aging.

What Counts as a Brain Network

When neuroscientists say “brain network,” they mean a set of regions whose activity rises and falls together in a statistically reliable way. Two kinds of evidence define these networks. Structural connectivity refers to the physical white-matter fiber tracts that wire one region to another, a bit like the cables between servers. Functional connectivity refers to the correlation in activity between regions over time, measured while a person rests in a scanner or performs a task. The two maps overlap but are far from identical: regions with tight physical wiring can show only weak functional coupling, and some functionally linked regions have no direct anatomical cable between them at all.¹

Both structural and functional brain networks share a layout that mathematicians call “small-world” topology: most regions connect to their immediate neighbors, but a handful of long-range shortcuts link distant clusters, keeping the average number of relay steps between any two regions surprisingly low.² Within that architecture sit highly connected “hub” regions, nodes that serve as major traffic intersections. These hubs are metabolically expensive to run but are critical for integrating information across the brain.

Connectivity at Different Scales

Brain networks exist at every zoom level. At the microscale, individual neurons form local circuits with branching dendrites that receive thousands of inputs. At the macroscale, entire cortical regions communicate through bundled white-matter tracts visible on MRI. Research bridging these levels has found that the two scales follow strikingly parallel rules of organization. Areas of the cortex that handle complex, multi-sensory processing, known as association areas, have both more elaborate dendritic trees at the cellular level and larger, more complex white-matter branching patterns at the whole-brain level.³ In other words, the architecture that makes a region a busy hub appears to be fractal-like, repeating at the level of individual cells and at the level of whole brain lobes. Datasets supporting these features range from complete cellular-level maps of simple organisms all the way to whole-brain connectivity maps in humans.⁴

The Three Networks That Run the Show

Among the many networks identified in imaging studies, three have attracted the most attention because their interactions seem to organize much of everyday cognition. Researchers refer to them collectively as the “triple network model.”⁵

The default mode network (DMN) is active when you are not focused on the outside world: daydreaming, recalling memories, imagining the future, or thinking about yourself and other people. In normal individuals, DMN activity drops when you switch to a demanding external task.⁶ Within the DMN, specific “core-self” regions, centered on medial prefrontal cortex, posterior cingulate cortex, and left inferior parietal lobule, light up more strongly during explicit self-referential thought than during simple rest.⁷ The DMN’s role extends to social cognition: the neuropeptide oxytocin modulates connectivity within this network, strengthening processes related to empathy, autobiographical memory, and social bonding.⁸

The central executive network (sometimes called the frontoparietal network) does roughly the opposite job: it engages when you need to concentrate, plan, hold information in working memory, or solve a problem. The salience network, anchored in the anterior insula and anterior cingulate cortex, acts as a switchboard. It monitors incoming sensory signals and internal states, decides what is important right now, and toggles the brain between the inward-facing DMN and the outward-facing central executive network.⁹ This switching role has been described as context-dependent modulation, where the salience network adjusts its influence on the other two networks depending on moment-to-moment demands.¹⁰

Networks Are Not Static

A common misconception is that brain networks are fixed circuits that sit quietly until needed and then “switch on.” In reality, they reconfigure on the timescale of seconds. Researchers studying dynamic functional connectivity have shown that the brain cycles through distinct states even when a person is lying still in a scanner, and that these rapid transitions can be tracked with millisecond precision using techniques like EEG and MEG.¹¹ New modeling frameworks can automatically detect these state transitions in both fMRI and MEG data, revealing that the brain’s resting-state connectivity is constantly fluctuating rather than holding steady.¹²

This constant shuffling probably explains why you can shift your attention, switch tasks, or snap out of a daydream in a fraction of a second. Your brain’s networks are not booting up from scratch each time; they are flexibly recombining, like musicians in an orchestra who form different ensembles for different pieces without anyone leaving the stage.

The Energy Bill for Staying Connected

Running all of this connectivity is expensive. The brain accounts for roughly two percent of your body weight but consumes about twenty percent of your energy. Research measuring glucose use alongside functional connectivity has found that hub regions, those heavily connected traffic intersections, consume disproportionately more energy as they take on more connections. Spontaneous brain activity, the background chatter of networks at rest, accounts for about 70% of the brain’s total energy consumption; the remaining 30% goes to basal metabolic maintenance unrelated to connectivity.¹³ Some hubs are more energy-efficient than others: deep-brain and cerebellar hubs get more connectivity per unit of glucose than cortical hubs do. This uneven cost structure has consequences for aging and disease, as the most metabolically active hubs tend to be the first to show wear.

How Networks Change Across a Lifetime

Brain networks are not born fully formed. During childhood and adolescence, structural network modules become increasingly segregated, meaning connections within a module strengthen while connections between modules weaken. This increasing modularity supports more efficient information processing and is directly linked to the improvement of executive functions like planning and impulse control during youth.¹⁴

In older adulthood, the trend reverses in specific ways. Brain signaling becomes more segregated overall, with the balance between integrated and segregated processing shifting toward segregation as people age. The size of that shift varies across populations, with one large dataset showing a correlation between age and segregation as high as 0.40.¹⁵ At the same time, overall modularity of functional networks tends to decrease with age, suggesting less distinct functional divisions across the whole brain. Different networks age at different rates: the default mode network typically loses local efficiency, while the sensorimotor network actually shows increased connectivity in older adults.¹⁶

There is also a decline in how tightly structural and functional connectivity track each other. In younger brains, the physical wiring and the functional activity patterns tend to line up more closely. With age, that coupling weakens, particularly in sensory-motor, cognitive, and subcortical networks, even though the fundamental small-world and hub architecture remains intact.¹⁷

When Networks Degrade in Disease

The brain’s hub regions, the same ones that burn the most fuel and route the most traffic, are disproportionately vulnerable to neurodegeneration. In Alzheimer’s disease, highly connected hubs in the medial and lateral prefrontal, parietal, and temporal cortex are selectively and severely affected, showing increased amyloid-beta buildup, tissue loss, reduced metabolism, and disrupted connectivity.¹⁸ Computational models suggest this is not random bad luck: the sheer volume of activity flowing through hub nodes accelerates their degeneration compared to less-connected regions, a pattern called activity-dependent degeneration.¹⁹ Recent work tracking the tau protein, another hallmark of Alzheimer’s, shows that its spread follows the brain’s connectivity architecture: amyloid patterns, structural connections, and functional connections all predict how tau distributes itself across the cortex as the disease progresses.²⁰

Schizophrenia presents a different kind of network breakdown. Rather than losing hubs outright, the triple network model’s coordination falls apart: the salience network, central executive network, and default mode network fail to switch on and off in the normal choreographed way, and this disrupted communication is tied to the cognitive impairments seen in the disorder.²¹ These conditions show that network-level thinking, rather than focusing on a single damaged region, often gives a better account of how symptoms arise.

Your Networks Are Uniquely Yours

One of the more striking recent findings is that the pattern of functional connectivity across your brain is stable enough over time to serve as a kind of fingerprint. Researchers can scan someone twice, weeks apart, and match the two scans to the correct individual purely from their connectivity profile. Beyond simple identification, these “connectome fingerprints” predict individual differences in higher-level cognitive abilities like fluid intelligence, executive function, and language comprehension.²² Regions of the brain that show the most variation between individuals tend to be the most informative both for identification and for predicting cognitive behavior.²³

This matters practically because it means brain-network measurements could eventually complement behavioral tests in clinical settings. If a patient’s connectivity fingerprint deviates from their own baseline, that shift could flag early cognitive decline or track response to treatment in a way that is less dependent on mood, motivation, or test-taking skill on a given day.

Measuring Networks and the Trade-Offs Involved

No single imaging tool captures everything about brain networks. The two workhorses are functional MRI (fMRI) and electroencephalography (EEG), and they have nearly opposite strengths. fMRI provides spatial resolution on the order of millimeters but captures brain activity indirectly, through blood-flow changes that lag several seconds behind the actual neural firing. EEG records electrical activity at the millisecond level but cannot pinpoint where in the brain a signal originates, because the electrical fields blur as they pass through skull and scalp.²⁴

Both techniques are also vulnerable to noise. Head movements in an MRI scanner can produce artifacts that mimic or mask real connectivity. EEG recordings pick up muscle twitches, eye blinks, and ambient electrical interference. These practical problems explain why published connectivity maps sometimes disagree: two research groups using slightly different preprocessing steps can reach different conclusions from the same raw data. The field addresses this through large-scale datasets with standardized methods, but readers should treat any single brain-network study with healthy skepticism.

Whole-brain computational models try to bridge these gaps. They simulate the dynamics of every brain region using mathematical descriptions of local neural activity, then couple those simulations together along the structural wiring measured with diffusion MRI in a real individual.²⁵ These models are already being used in neurosurgical planning, where predicting how a tumor resection will affect network function can guide the surgeon’s approach.

What Happens When a Stroke Hits a Network

A stroke destroys tissue in one location, but its effects radiate outward through the networks that pass through the damaged zone. Lesion-network mapping is a technique that embeds a patient’s lesion into a map of normal connectivity to estimate which distant regions lose their functional link to the damaged area.²⁶ This approach helps explain why two patients with similar-looking lesions on a scan can have very different symptoms: what matters is not just the local damage but which networks the lesion disrupts.

Intriguingly, the brain shows measurable resilience after stroke. When researchers simulate a second virtual stroke in stroke patients’ brains and compare the result to the same simulation in healthy controls, patients’ networks hold up better. This resilience in network efficiency appears as early as two weeks after the initial stroke and lasts at least a year, suggesting the brain actively reorganizes its connectivity in response to damage.²⁷ However, using lesion-network mapping to predict an individual’s behavioral deficits remains imperfect; improved methods are getting closer but still explain only a portion of the variability in outcomes.²⁸

Why Human Networks Look the Way They Do

Comparing brain networks across species reveals something about what makes the human brain distinctive. It is not simply that our brains are bigger. The regions of the cortex devoted to primary senses like vision and touch are almost the same size in humans and chimpanzees. The cortical areas that have expanded disproportionately in humans are the association cortices, the regions that integrate information from multiple senses and support abstract reasoning, language, and social cognition.²⁹ This expansion comes with increased anatomical connectivity between these areas, providing the physical wiring for networks that can bind together far-flung regions of cortex.³⁰ In short, human cognition is less the product of any single brilliant brain area and more the result of having a richer, more densely interconnected set of association networks than other primates.

Targeting Networks With Therapy

The realization that symptoms map onto networks rather than single regions has opened the door to network-targeted therapies. Transcranial magnetic stimulation (TMS) uses magnetic pulses to stimulate a spot on the scalp, and when that spot is chosen based on its network connectivity rather than just its location, the effects can propagate through the relevant circuit. Early work using this approach in people with mild cognitive impairment has shown promise for slowing early-stage decline, with researchers emphasizing that personalizing the stimulation site to each patient’s connectivity map improves outcomes.³¹ Clinical trials are now using network maps derived from lesion studies to select TMS targets for conditions as varied as cervical dystonia, depression with co-occurring anxiety, and anorexia nervosa.³²

Another avenue is neurofeedback, where a person watches a real-time display of their own brain activity and learns to nudge it in a particular direction. fMRI-based neurofeedback can modify connectivity across broad neural networks, including the default mode and executive control networks, even when the feedback targets only a single region.³³ In studies of depression, training people to increase amygdala activity while recalling happy memories reversed abnormal patterns of connectivity and produced lasting brain changes, with improvements in self-reported mood.³⁴ These results are still early-stage, but they illustrate a principle: because brain networks are plastic and responsive to experience, it is at least possible to retrain them.

Psychedelics and Network Reconfiguration

Few recent findings have generated more public curiosity than the effect of psychedelic drugs on brain networks. Under psilocybin, brain connectivity does not simply “dissolve” into chaos, as sometimes portrayed. Instead, the connectome decomposes into dissociable patterns. One recent study identified two distinct connectivity traits during acute psilocybin effects: a primary pattern that scaled with the concentration of psilocin in the blood, and a second, independent pattern linked to impaired performance on a visual creativity task.³⁵ The fact that these two patterns were statistically independent suggests the psychedelic state involves multiple co-occurring network shifts rather than one global scramble. That distinction matters for ongoing clinical trials exploring psilocybin for depression and PTSD, because it implies that specific therapeutic connectivity changes could be separable from the disorienting perceptual effects.

Chemical Signals That Tune the Networks

Networks are not governed solely by their wiring. Chemical messengers, particularly norepinephrine and dopamine, act as volume knobs that tune how prefrontal circuits process information. Both transmitters reach the prefrontal cortex extensively, but they bind to different receptor types and activate different signaling pathways, meaning each one adjusts network behavior in its own way. Norepinephrine tends to sharpen attention and alertness, while dopamine is more tightly linked to working memory and reward-based learning. Critically, both must stay within an optimal range: too little or too much of either can degrade the very networks they are supposed to support. This inverted-U relationship helps explain why stimulant medications can improve focus at one dose but cause agitation at another, and why stress, which floods the prefrontal cortex with catecholamines, can temporarily impair executive function even in healthy people.