Your brain runs on about 20 watts. Roughly a fifth of that — the energy equivalent of a dim nightlight — is spent on circuits that switch on when you stop doing anything in particular. The pattern has a name now, the default mode network, and it was hiding in plain sight inside PET scans for years before anyone believed it was real.
The man who finally named it was Marcus Raichle, a neurologist at the Washington University School of Medicine in St. Louis. In a 2001 PNAS paper titled “A default mode of brain function,” Raichle and his colleagues coined the term “default mode” to describe a baseline state of the resting brain. The phrase was deliberately modest. What it described was not.
The anomaly in the control condition
In the late 1990s, neuroscientists ran experiments the same way pharmacologists run drug trials. There was a task — read this word, watch this dot, remember this face — and there was a control: rest. You compared brain activity during the task against brain activity during the rest, subtracted one from the other, and the difference was supposedly the work the task required.
The subtraction kept producing a strange residue. In trial after trial, the same scattered regions decreased their activity the moment a task began. Medial prefrontal cortex. Posterior cingulate. Precuneus. Angular gyrus. They quieted down when concentration started, and they fired back up the instant the task ended.
Researchers referred to these regions as task-negative areas. The label was almost a complaint. They weren’t supposed to be there.
What the PET scanner kept showing
Raichle had been working with positron emission tomography since the 1980s, watching radioactive water trace blood flow through living brains. PET is slow and blunt compared to modern fMRI, but it measures something fMRI can only infer: actual oxygen consumption, the metabolic cost of thinking.
The task-negative regions weren’t quiet at baseline. They were burning fuel at a rate close to the busiest regions of an actively working cortex. The brain, which is about 2 percent of body mass, draws roughly 20 percent of the body’s total energy at rest — and a disproportionate share of that draw was being spent by these idling, inward-facing areas.
Raichle asked subjects to close their eyes and let their minds wander. He measured what happened. The task-negative regions used more energy than the rest of the brain. That was the paper. That was the discovery.
From scattered regions to a network
Two years later, a team at Stanford led by Michael Greicius generated higher-resolution data and showed that these regions weren’t just coincidentally active at the same time. They were synchronised. The blood-oxygen signal in one region rose and fell in step with the signal in another, even when nothing was happening in the room. They were a network, and the Stanford group gave it the name that stuck: the default mode network.
The finding turned the standard model of brain imaging on its head. As a 2007 review by Fox and Raichle in Nature Reviews Neuroscience later put it, spontaneous blood-oxygen fluctuations are not random noise. They are specifically correlated between functionally related brain regions, and they map onto known anatomical systems. The brain at rest is not a brain at rest. It is a brain doing something else.
And the something else does not stop when a task begins. The same review noted that these spontaneous fluctuations continue throughout task paradigms, contributing to the trial-by-trial variability that experimenters had spent decades trying to scrub out of their data.
Twenty watts, one fifth of everything
The brain is about 2 percent of body mass. It draws roughly ten times its share of metabolic energy. And the lion’s share of that draw is not the visible work of doing things. It is the invisible work of being a self.
The cortex maintains this consumption with remarkable stability. A 2013 PNAS paper by Tomasi, Wang and Volkow showed that brain regions with the highest degree of functional connectivity also carry the highest glucose cost — connector hubs are the brain’s most expensive real estate. The default mode is composed of several of those hubs.
What the network actually does
Two decades of follow-up work have produced a list of functions long enough to feel suspicious. A foundational 2008 review by Buckner, Andrews-Hanna and Schacter implicated the default mode network in mind-wandering, remembering past experiences, thinking about other people’s mental states, imagining the future, processing language, and weighing decisions about reward and punishment. It is the substrate of replay, prediction, and what some researchers call self-referential processing.
Vinod Menon, who directs the Stanford Cognitive and Systems Neuroscience Laboratory, has argued that the apparent grab bag is held together by a single job: building an internal narrative. The default mode helps you think about who you are in relation to other people, retrieve your past, and stitch the result into a coherent story about yourself.
Lucina Uddin, a neuroscientist at UCLA, started studying the network because of its links to self-recognition — identifying your own face, your own voice. She has since shifted her attention to how networks talk to each other. Examining how networks communicate with each other can be more revealing than studying them in isolation. Networks work together dynamically, coming apart and changing their activity patterns over time.
The off switch
The default mode is not always on. It has a specific antagonist: the salience network, anchored in the anterior insula and the anterior cingulate cortex. The salience network’s job is to detect when something in the environment is important enough to pay attention to. When it fires, the default mode quiets down.
Causal evidence for this push-pull comes from animal work using optogenetic stimulation. A 2023 study in Nature Communications by Menon and colleagues showed that stimulating Chronos-expressing neurons in the anterior insular cortex of rats directly suppressed activity in the default mode network and decoupled the two systems. The salience network is, mechanically, the brake.
This is the architecture behind the feeling of snapping out of a daydream. The phone buzzes. The salience network flags it. The default mode releases its grip on the internal narrative. Attention swings outward.
The states where it goes quiet
The states in which the default mode network noticeably loosens are interesting in their own right. Deep concentration on a demanding external task is one. The pattern Mihaly Csikszentmihalyi described in 1975 after interviewing artists and chess players who forgot to eat — flow, the state in which the self drops out of awareness — looks, on a scanner, a lot like a default mode network that has been turned down.
Psychedelics are another. Singleton, Luppi and colleagues, writing in Nature Communications in 2022, used receptor-informed network control theory and existing fMRI datasets to link LSD and psilocybin to what they describe as a flattening of the brain’s control energy landscape. The brain transitions between states more easily, and the default mode’s grip on its usual configuration weakens. The sense of a unified self — the network’s chief output — weakens with it.
Sleep and anaesthesia, by contrast, do not abolish the network. Spontaneous BOLD correlation patterns are largely consistent across resting states, including sleep and anaesthesia. The network is an intrinsic property of the brain, not a side effect of unconstrained mental chatter.
Why dysfunction here matters
If the default mode is the substrate of self-narrative, anything that disrupts it should produce something that feels like a disrupted self. That is roughly what the clinical literature has found, in messy form. Altered default mode activity has been reported in depression, schizophrenia, anxiety, autism, dementia and Alzheimer’s disease.
The findings are not tidy. In depression, some studies find network nodes overly connected, others find the opposite. In some cohorts the network itself looks normal but its interactions with other networks are off. One proposal is that abnormal coordination between the default mode, the salience network and a third system called the frontoparietal network underlies a wide range of psychiatric symptoms.
Deanna Barch, who studies the neurobiology of mental illness at Washington University in St. Louis, has called this framework a useful starting point rather than an explanation. Differences in connectivity, she has said, are a place to look, not a verdict.
What “autopilot” looks like on a scanner
The colloquial sense of going through the motions — driving home without remembering the drive, scrolling without registering the feed, getting to the end of a workday without recalling a single decision — has a plausible neural signature. It looks like a default mode network running loud while the salience network’s attention-flagging is muted. The internal narrative keeps generating itself. The external world keeps getting filed without being noticed.
This is not pathology. It is, by Raichle’s measurement, among the most metabolically expensive things the brain does. The cost of having a continuous self is paid in watts, around the clock, whether or not anything is being attended to.
What Raichle actually changed
Before 2001, the resting brain was a control condition. After 2001, it was a subject. Functional connectivity — the practice of mapping which regions fluctuate together when nothing is being asked of the brain — became one of the most productive techniques in neuroscience. It produced the salience network, the frontoparietal network, the dorsal attention network and a handful of others, each defined by its own coherent rhythm.
None of those discoveries were possible while “rest” meant “nothing.” Raichle’s contribution was less the network than the reframe: the brain at rest is doing the thing the brain mostly does. The tasks are the exception. The default is the rule.
The dim bulb and the inner monologue
The arithmetic is worth holding still. A human brain consumes roughly 20 watts. Around a fifth of that — close to four watts — goes to the default mode and its baseline maintenance of the self. Four watts is less than the standby draw of a television. It is roughly the power of a small LED desk lamp.
Everything you think of as you when you are not actively concentrating — the running commentary, the rehearsal of yesterday’s conversation, the half-formed plan for next week, the sense that you are a continuous person moving through time — costs about as much electricity as a reading light.
It has been running, in some form, since you were old enough to have a self at all. It will keep running tonight while you sleep. Raichle’s instruments saw it first in 2001, but by then it had been there, burning its quiet four watts, for the entire history of the species.
