What time of day are you actually alert?

Predicted alertness across 24 hours

Two-Process Model (Borbély 1982). Curve assumes the schedule below run to steady state.

Your current pattern

What is the Two-Process Model?

Process S is the homeostatic sleep-pressure variable. It rises whenever you are awake — slowly, exponentially toward an upper asymptote — and falls when you sleep, faster, toward a lower asymptote. The mechanism is the build-up and clearance of adenosine and related sleep-regulating substances in the basal forebrain. The longer you stay awake, the higher S climbs; the more you sleep, the lower it falls.

Process C is the circadian alerting signal. Independent of how long you have been awake, it traces a near-sinusoidal pattern over 24 hours with its peak near 17:00 and its trough around 04:00. The pacemaker is the suprachiasmatic nucleus; the entraining input is light. C explains the counter-intuitive finding that you are most alert in the late afternoon — even on days when you slept badly — and most vulnerable around 03:00–05:00, when the homeostatic and circadian pressures align.

Net alertness is what you feel: roughly (S_upper − S) + C. The two processes can compound or cancel. In the morning, the circadian rise pulls you out of the sleep dip even though S is still moderate. In the afternoon, a brief dip in C produces the post-prandial slump despite plenty of remaining wake capacity. Right before bed, a final circadian wake-maintenance peak fights the climbing S — which is why people often feel a second wind around 21:00 and then crash. Insomnia is often a phase-misalignment problem in disguise; the curve makes the structure visible.

How to read the curve

• The circadian nadir sits around 04:00–06:00 — the lowest predicted alertness of the day. Cognition, reaction time, and emotion regulation all bottom out here. If you have to drive, this is when you crash.
• The post-prandial dip is the early-afternoon slump around 14:00–16:00, driven by a temporary fall in Process C. It is real biology, not a lunch problem — fasting subjects show the same dip.
• The wake-maintenance zone is the 1–3 hours before bedtime when Process C peaks one last time, producing a final burst of alertness even as Process S climbs. This is why an earlier bedtime often fails — you are biologically primed to stay up.

Caffeine pharmacokinetics

Caffeine follows first-order elimination kinetics. After oral intake, plasma concentration peaks at Tmax — roughly 30 to 60 minutes — then falls with a median half-life of about 5 hours in healthy adults (Institute of Medicine 2001). The model here approximates intake as a 45-minute linear ramp to the peak dose, then a pure exponential decay anchored at that peak. Real absorption profiles vary with stomach contents and formulation, but the bolus approximation introduces less than 5 percent error past the 90-minute mark — well before any sleep-relevant window.

Caffeine is an adenosine receptor antagonist. Adenosine is one of the molecular substrates of Process S — it accumulates while you are awake and signals sleep pressure to the basal forebrain. Caffeine occupies the same receptors without activating them, which is why it masks rather than reduces homeostatic pressure (Fredholm et al. 1999). Process S is still climbing in the background; the brain just stops reading it. When caffeine clears, the unread pressure arrives all at once — the post-caffeine crash.

The 5-hour half-life is a median, not a constant. CYP1A2 polymorphism produces fast and slow metabolizer phenotypes that differ by a factor of three. Cigarette smoking induces CYP1A2 and roughly halves the half-life. Oral contraceptives roughly double it. Late-trimester pregnancy can triple it. None of this is in the visualization — the curve assumes a 5-hour median (Institute of Medicine 2001). If you metabolize caffeine slowly or are on a hormonal contraceptive, treat the curve as optimistic and your real bedtime residual as higher.