Sleep neuroscience
Stages and their circuits, the two-process model, replay and memory consolidation, and what closed-loop stimulation is trying to do. The biology behind the hypnogram your band drew.
You are skimming: the title, the first figure, and the short version. Switch to Read in the header for the full page, or Deep to open every deep dive.
Your band recorded a night and the stager drew a staircase: wake, light sleep, deep sleep, back up, REM, repeat every ninety minutes, with the deep sleep front-loaded and the REM back-loaded. Each step of that staircase is a different brain state with a different circuit in charge, and the sequence is not arbitrary. Sleep is when the brain does maintenance it cannot do while awake, and the rhythms you detected are the maintenance in progress.
The stages, mechanistically
Wake. Brainstem and hypothalamic arousal systems (noradrenaline, acetylcholine, histamine, orexin) bathe the cortex and thalamus in neuromodulators that keep neurons depolarized and desynchronized. Low-amplitude, fast EEG; alpha with eyes closed.
N1. Arousal drive drops. Alpha fades; slow rolling eye movements; theta. Two minutes, easily interrupted, and the stage every stager gets wrong because it is a transition.
N2. The thalamus, no longer held awake, begins to gate: spindlesSleep spindleA half-second burst of 11 to 16 Hz activity in light sleep, generated in the thalamus. Glossary entry from the reticular-relay loop, and K-complexes, single large slow waves. Half the night is N2.
N3. Cortex falls into up and down states: slow wavesSlow waveA large, slow (under 1 Hz) oscillation of deep sleep during which most cortical neurons fall silent and then fire together. Glossary entry, delta, the largest EEG there is. Hard to wake from. Concentrated in the first two cycles because the pressure to enter it, built up during the day, is highest then.
REM. A cholinergic brainstem population switches on, the cortex becomes as active as in wake, the eyes dart, dreams are vivid, and a separate circuit paralyzes the skeletal muscles so the dreams are not acted out. EEG looks like wake with no alpha; the chin EMG goes flat; that is how the stager separates the two.
Why the staircase has that shape
The two-process model: a homeostatic pressure S that rises during wake and falls during sleep, and a circadian rhythm C from the hypothalamus’s clock that gates when sleep is permitted. At bedtime S is high, so deep sleep comes first and hard. As S discharges, N3 shrinks and REM, which C favours toward morning, grows. Caffeine blocks the adenosine that carries S; jet lag is C out of phase with the schedule; both show up in the hypnogram.
During deep sleep the hippocampus replays, at high speed, the sequences of activity from the day. Where in the slow wave does this replay tend to occur?
In the up state, nested with the spindle. Hippocampal sharp-wave ripples carrying replay occur preferentially during the cortical up state and in the troughs of spindles, so that the reactivated memory arrives at cortex when its synapses are ready to change. That three-way nesting (slow wave, spindle, ripple) is the leading account of how memories move from hippocampus to cortex overnight, and it is the reason enlarging slow waves with sound is expected to help memory.
Consolidation
Learn a list of words before sleep and you recall more of them in the morning than after an equal time awake; the effect tracks the amount of slow-wave sleep. Rodent recordings show hippocampal place-cell sequences from the day replayed during subsequent sleep, compressed twenty-fold. The active systems consolidation account says the hippocampus, a fast-learning temporary store, teaches the cortex, a slow permanent one, by replaying during slow waves. A complementary account, synaptic homeostasis, says slow-wave sleep renormalizes synapses that were potentiated during the day, pruning noise and preserving signal. Both are probably partly right, and the closed-loop protocol you built is a test of the first.
What closed-loop stimulation does
Two soft clicks timed to the up-state of a detected slow wave, and the following slow waves are larger and the spindles more numerous, and next-morning recall improves in most (not all) studies. The stimulation is doing what the brain does anyway, a little more strongly, at the moment the brain is already doing it. Clicks out of phase have no effect or a negative one. This is the model for a whole class of neurotechnology: sense a brain state, act only in phase with it, and let the brain’s own mechanisms do the work.
Clinical sleep
Obstructive sleep apnea, where the airway collapses and the sleeper wakes repeatedly, is the most common diagnosis in a sleep lab and the reason polysomnography records breathing and oxygen. Insomnia, narcolepsy (orexin loss), REM behaviour disorder (the paralysis fails; it predicts Parkinson’s disease years ahead), and restless legs round out the clinic. A sleep technologist sees all of them, which is why the clinic thread recommends the job.
Deep dive Why we sleep, honestly 2 min
Nobody knows, fully. Consolidation, synaptic renormalization, metabolic clearance (the glymphatic system flushes the brain’s interstitial space during slow-wave sleep), energy conservation, and immune function all have evidence. Every animal with a nervous system does something like it, and total deprivation kills rats in weeks. That a function so universal is so unexplained is a good reminder of how early the field is.
Deep dive Consumer sleep tech and what it can measure 2 min
Wrist devices infer stages from movement and heart rate, which correlate with sleep stage but do not define it; their agreement with polysomnography is modest, especially for N1 and for distinguishing REM from wake. Forehead EEG bands measure the actual defining signal and do much better. The regulatory line between wellness product and medical device runs right through this space, which Phase 5 covers.
Explain what this page was about to your roommate in three sentences. No jargon they would not know.