Brain States and Chemistry

The brain is never off

One of the most persistent images in popular writing about the brain is the computer that gets shut down at night. It is wrong in a way that matters. The brain in deep sleep is not idling, and it is not consuming a trivial fraction of the energy it uses awake. Whole-brain metabolic rate falls in slow-wave sleep, but only modestly relative to quiet wakefulness, and in REM sleep it climbs back to roughly waking levels. The organ is running the whole time. What changes is how it runs.

Consider what the EEG shows. In quiet, alert wakefulness the trace is fast and low in amplitude: cortical neurons are firing, but each on its own schedule, so their contributions largely cancel at the scalp. In deep slow-wave sleep the trace becomes slow and enormous. That is not less activity, it is more synchronised activity: vast populations of cortical neurons alternating together between an active up-state and a near-silent down-state, roughly once a second. The big wave is the signature of a crowd moving in step, not of a crowd going home.

The point that follows from this is the organising idea of the whole section. The brain does not have one mode of operation that occasionally gets interrupted. It has several, it switches between them on a schedule that is only partly under voluntary control, and the circuit that recognises a face, or holds a number in mind, or forms a memory, does a different job in each one.

What sets a brain state

If you wanted to design a system that could flip a hundred billion cells into a new mode of operation, you would not wire each cell individually. You would build a small number of specialised nuclei with enormously branched axons, let each one release a chemical over huge swathes of cortex at once, and change the state by changing the chemical mix. That is, in outline, what the brain does.

The systems that do this are the neuromodulatory systems. Each consists of a small nucleus, often only tens or hundreds of thousands of cells, sitting in the brainstem, hypothalamus or basal forebrain, whose axons project diffusely across the cortex and much of the rest of the brain. They do not carry point-to-point messages in the way that a sensory pathway does. They set a background condition.

Two features of this arrangement deserve emphasis. First, the numbers are small: the locus coeruleus contains on the order of tens of thousands of neurons in humans, yet its influence is brain-wide. A very small structure sets the condition of a very large one. Second, the effect is modulatory rather than instructive. Noradrenaline does not tell a cortical neuron what to represent. It changes the gain, the excitability, the signal-to-noise ratio, the willingness of a synapse to change. The message stays the same; the way the message is handled does not.

Thalamocortical rhythms

The chemical background is half the story. The other half is timing. The cortex and the thalamus are wired into a massive reciprocal loop: cortex projects down to thalamus, thalamus projects back up to cortex, and an inhibitory shell around the thalamus, the reticular nucleus, sits in the middle of the conversation. Loops with delay and inhibition oscillate. This one does, and the oscillations it produces are among the most prominent features of the EEG.

What the loop does depends on the chemical state it is in. When neuromodulator levels are high, in waking, thalamic relay neurons sit depolarised and pass sensory information through to cortex in a faithful, one-spike-per-input manner. As those levels fall at sleep onset, the same neurons hyperpolarise and switch into a bursting mode, in which they fire rhythmic volleys instead of relaying. The thalamus stops being a gateway and starts being a metronome. This is the cellular reason why the sensory world fades at sleep onset, and it is a beautiful example of the general principle: same cells, same wiring, entirely different function.

The wake-sleep continuum

The best-charted axis of brain state is the one running from alert waking down through drowsiness into sleep, and then, in a genuine surprise of twentieth-century neuroscience, back up into the strange wake-like activity of REM. It is worth laying out the whole sequence, because each step is a different chemical and electrical regime.

1. Alert wakefulness

   Noradrenaline, acetylcholine, serotonin, histamine and orexin all active. Cortical EEG fast and desynchronised. Thalamus in relay mode: the outside world gets in. Attention, working memory and decision circuits fully online.

2. Relaxed wakefulness and drowsiness

   Arousal falls. Posterior alpha rhythm becomes prominent, particularly with the eyes closed. Thoughts drift; attention lapses become more frequent. This is a genuinely intermediate state, not merely a weaker version of alertness.

3. Non-REM sleep

   Neuromodulator tone drops sharply. The thalamocortical loop takes over, producing sleep spindles, K-complexes and then the great slow waves of deep sleep. Sensory transmission is gated off at the thalamus.

4. REM sleep

   The paradox. Acetylcholine returns to waking levels while noradrenaline and serotonin fall to near zero. The EEG looks awake, the eyes move, vivid dreaming occurs, and the skeletal muscles are actively paralysed so that the dream is not acted out.

These four regimes are not points on a single dial. REM is not "deeper" than slow-wave sleep, and it is not "lighter" either. It is a third thing, chemically unlike either waking or non-REM. Any model that treats sleep as a simple dimming of the waking brain cannot accommodate it, which is precisely why its discovery in the 1950s forced a rethink of what sleep is. The full architecture is covered in sleep and the brain.

How states are measured

Because states are global, they can be read from a global signal, and the classical global signal is the electroencephalogram. EEG electrodes on the scalp record the summed postsynaptic potentials of large, aligned populations of cortical pyramidal neurons. It is a blunt instrument spatially, but its temporal resolution is superb, and it is exquisitely sensitive to exactly the thing that distinguishes states: how synchronised the population is.

The standard sleep study, polysomnography, combines EEG with eye movement recording and a muscle tone channel, and that combination is not optional: the EEG of REM sleep is so wake-like that without the eye and muscle channels you could not reliably tell them apart. The details of what EEG can and cannot tell you, including the widespread misreadings of "brainwave" frequency bands, are covered in brain waves and EEG.

The stress state and the reward state

Sleep and waking are the states everyone thinks of, but they are not the only ones. Two others are important enough to have their own pages, and both illustrate the same principle from a different angle: a chemical signal, released widely, that changes how the whole system behaves.

The stress state

A threat, real or anticipated, triggers two responses on two timescales. Within seconds, the sympathetic nervous system floods the body with adrenaline and the brain with noradrenaline: heart rate up, pupils wide, attention narrowed onto the threat. Over minutes to hours, a hormonal cascade from the hypothalamus through the pituitary to the adrenal cortex raises cortisol, which crosses back into the brain and acts on receptors that are especially dense in the hippocampus and prefrontal cortex. Acutely this is adaptive. Sustained, the picture becomes more complicated, and the honest evidence is less tidy than most popular accounts suggest. See the stress response.

The reward and motivation state

Dopamine neurons in the midbrain fire when things turn out better than expected, and their signal is broadcast to the striatum and prefrontal cortex. This is not a pleasure signal, whatever the popular press says. It is a teaching signal and a motivational one: it tells the brain that a prediction was wrong in a good direction, and it energises the pursuit of whatever produced the surprise. The distinction between wanting and liking, which turns out to rest on different chemistry, is one of the most important corrections in modern neuroscience. See dopamine and reward.

Why state determines function

Here is the reason this section exists at all. Neuroscience textbooks are largely organised around structures and circuits: this region does that job, this pathway carries that signal. That framing is useful but incomplete, because a circuit is not a machine with one function. It is a machine whose function depends on the state it is embedded in.

The examples are everywhere once you look for them. A thalamic relay neuron transmits sensory information in waking and generates rhythmic bursts in sleep, using the same cell and the same synapses. Cortical synapses undergo long-term potentiation readily when acetylcholine and noradrenaline are present and much less readily when they are not, which means the brain's capacity to learn is gated by state. The prefrontal cortex supports flexible, deliberate control at moderate arousal and is functionally impaired at high arousal, when control shifts toward faster, more habitual systems. And a memory that was encoded during the day is reactivated and consolidated during subsequent slow-wave sleep, a process that simply cannot happen in the waking state.

Common misconceptions

Explore this section