Two clocks, not one
You feel sleepy after roughly 16 hours awake regardless of what the clock on the wall says — but you also feel far more alert at 9 PM than at 3 PM, despite having been awake for the same total duration in each case. Neither fact alone explains human sleep timing; together, they point to two separable biological processes running in parallel. In 1982, Alexander Borbély proposed the two-process model: a homeostatic sleep drive that only cares how long you've been awake, and a circadian oscillator that only cares what time of day it is. Sleep happens where the two collide.
Process S: adenosine on a stopwatch
Process S is the homeostatic sleep pressure — it rises while you're awake and falls while you're asleep, tracking a single underlying chemical signal: adenosine accumulating in the basal forebrain as neurons burn ATP. Caffeine feels like it "wakes you up," but it doesn't remove any adenosine — it just blocks the A1 and A2A receptors that would otherwise sense it, which is why a delayed crash arrives once the caffeine's roughly five-hour half-life clears and all that backlogged adenosine binds at once.
// Waking: S rises toward S_max with time constant tau_w ~ 18.2 h S(t) = S_max - (S_max - S0) * exp(-t / tau_w) // Sleeping: S falls toward S_min with time constant tau_s ~ 4.2 h S(t) = S_min + (S0 - S_min) * exp(-t / tau_s) // A normal 16h-awake / 8h-asleep day: // S climbs ~0.20 -> ~0.68 while awake // S falls ~0.68 -> ~0.18 while asleep
Sleep recovers faster than sleepiness accumulates — the sleeping time constant is roughly four times shorter than the waking one — which is why a single good night's sleep can undo a day of fatigue almost entirely. The EEG signature of Process S is slow-wave activity (0.5–4.5 Hz delta power) during NREM sleep: it's highest in the first sleep cycle, when S has just peaked, and decays through the night in near-perfect lockstep with the model's prediction.
Process C: the clock that doesn't care how tired you are
Process C is generated by roughly 20,000 neurons in the suprachiasmatic nucleus (SCN) of the hypothalamus — a self-sustaining oscillator that keeps running on its own near-24-hour rhythm even in total darkness, entrained day to day by light hitting specialised retinal photoreceptors. Unlike Process S, it is completely indifferent to your sleep history; it approximates a simple cosine, peaking for alertness in the afternoon and bottoming out around 4–6 AM.
C(t) = C_mid + A * cos(2*PI * (t - phase) / T) // T ~ 24.2 h (free-running period in constant darkness) // phase peak alertness ~midday, nadir ~4-6 AM
Critically, Process C doesn't just sit passively in the background — in the early evening it actively counteracts the day's accumulated Process S, holding off sleep onset until after dark in what chronobiologists call the "wake maintenance zone." That's why you can feel wide awake at 9 PM despite having been up since 7 AM, and then feel a wall of sleepiness hit at 11 PM once the circadian push fades. Individual variation in the phase of this oscillator produces chronotypes — morning larks are phase-advanced, night owls phase-delayed, and the distribution skews later in teenagers due to a genuine adolescent shift in SCN timing, which is a large part of why early school start times hurt learning.
Where the two lines cross: the flip-flop switch
Sleep onset and offset aren't governed by a single fixed threshold — the thresholds themselves oscillate with the circadian rhythm, opening a nightly "sleep gate" and closing it again near dawn. Neurally, the transition is sharp rather than gradual because it's implemented as a flip-flop switch: the ventrolateral preoptic nucleus (VLPO) inhibits the brain's wake-promoting nuclei, and those wake nuclei mutually inhibit the VLPO right back. Each side suppresses the other, so the system snaps between two stable states — fully awake or fully asleep — rather than lingering indefinitely in between, much like a light switch rather than a dimmer.
Orexin neurons in the lateral hypothalamus act as the switch's stabiliser, reinforcing whichever state is currently active; their loss is the underlying cause of narcolepsy, where the flip-flop switch fires erratically and patients fall between wake and REM sleep with little warning. Lesion studies confirm the two-process architecture directly: destroying the SCN in animals doesn't stop sleep, but it does eliminate the daily rhythm — leaving only short, fragmented sleep bouts paced by Process S alone, with no more coherent day/night pattern.
Deprivation, debt, and jet lag
During sustained wakefulness, Process S climbs monotonically and cognitive performance falls in step — after roughly 24 hours awake, S approaches its ceiling and psychomotor vigilance impairment is comparable to a 0.1% blood alcohol level. Crucially, during chronic partial sleep restriction (six hours a night for two weeks), objective performance keeps declining while subjective sleepiness plateaus — people stop feeling as impaired as they actually are, a genuinely dangerous mismatch that the two-process model predicted before it was confirmed experimentally.
Jet lag is a Process C problem, not a Process S problem — your sleep pressure adapts to a new time zone within a day or two, but the SCN's phase re-entrains far more slowly, at roughly 1 hour per day flying east and 1.5 hours per day flying west. The asymmetry follows directly from the free-running period being slightly longer than 24 hours: westward travel asks the clock to delay, which is the direction it already wants to drift, while eastward travel demands a phase advance against that natural tendency. Shift workers face the cruelest version of this mismatch — because their light exposure never consistently supports one phase, Process C often never fully re-entrains at all, a state of chronic circadian misalignment linked to elevated cardiovascular and metabolic disease risk.
Frequently asked questions
Why do I feel sleepy at 3 AM but wide awake at 11 PM even though I've been awake longer by 3 AM?
Process S (sleep pressure) is indeed higher at 3 AM than at 11 PM — but Process C, the circadian alertness signal, actively fights sleepiness during the evening "wake maintenance zone" and then drops to its lowest point around 4–6 AM. At 3 AM both effects point the same way: high S and low C combine to produce the deepest sleepiness of the whole 24-hour cycle, which is also why that hour is the most dangerous time to be driving.
Why is eastward jet lag worse than westward jet lag?
The human circadian pacemaker free-runs at roughly 24.2 hours, slightly longer than a solar day, so it naturally drifts later. Flying west asks the clock to delay — moving with its natural tendency, which the SCN re-entrains at about 1.5 hours per day. Flying east asks the clock to advance to an earlier phase, fighting that natural drift, and re-entrainment slows to roughly 1 hour per day. A 9-hour eastward flight can take over a week to fully resolve.
Does caffeine actually reduce Process S, or does it just mask it?
It masks it. Process S is driven by adenosine accumulating in the basal forebrain during wakefulness, and caffeine works by competitively blocking adenosine A1 and A2A receptors — it doesn't stop adenosine from building up, it just stops the brain from sensing it. Once caffeine's roughly 5-hour half-life clears, the accumulated adenosine binds all at once, which is the mechanistic basis of a caffeine crash.
Try it live
Every curve above runs live in Two-Process Sleep Model — Circadian & Homeostatic Regulation. Adjust wake/sleep duration and light exposure and watch Process S and Process C interact to open and close the nightly sleep gate.
▶ Open Sleep Model simulation