What one bad night actually costs you

If you take the tests on this site for a fortnight and look at your history, the largest single source of variation will almost certainly not be practice. It will be sleep.

Why reaction time is the first thing to go

Sustained attention is unusually sensitive to sleep loss, and reaction tasks measure it more directly than almost anything else. This is why the psychomotor vigilance task — essentially a longer, stricter version of our reaction time test — became a standard instrument in sleep research.

The characteristic finding is not what people expect. Sleep loss does not slow everything down evenly. It produces lapses: most responses stay near normal, and then occasionally one is very slow, as though attention simply dropped out for a second.

That pattern has a practical consequence for reading your own results. Your median may barely move while your slowest rounds get dramatically worse — so if you only ever look at the median, you will underestimate what a bad night did to you. The mean and the spread are more revealing here than usual.

Why you cannot judge it yourself

This is the part that matters most, and it is well documented.

People underestimate their own impairment when sleep-deprived, and the gap widens as the deprivation continues. Objective performance keeps declining while subjective sleepiness plateaus — you stop feeling increasingly tired long before you stop getting worse.

The reason it matters is that the judgement you would use to decide whether you are fit to drive, or to make an important decision, is itself produced by the impaired system. This is the single strongest argument for having a number rather than an impression, which is most of what this site is for.

The interaction with caffeine

Caffeine blocks adenosine, the molecule that accumulates while you are awake and produces sleepiness. That is the entire mechanism, and it explains both the benefit and the limit.

It suppresses the signal that you are tired. It does not remove the tiredness. Adenosine continues accumulating while its receptors are blocked, which is why the drop when caffeine clears can feel abrupt.

Two consequences worth planning around:

• The half-life is roughly five to six hours. A coffee at 4pm still has about half its caffeine working at 10pm, and the sleep you lose that night costs more than the alertness you gained.
• Tolerance builds quickly. Regular heavy use tends toward a state where caffeine restores you to baseline rather than lifting you above it — a much less useful transaction than the one you think you are making.

On a reaction test, caffeine reliably improves a sleep-deprived result and does much less for a rested one. If your history shows a big caffeine effect, that is mostly information about your sleep.

What recovers it

Sleeping. Unglamorously, this is the answer, and one night of adequate sleep restores most of a single night’s deficit.

Accumulated debt takes longer. Several consecutive short nights do not resolve in one long one; recovery lags behind, which is why performance can stay depressed into a weekend that feels like it should have fixed things.

Naps help, and they have a catch. A short nap of ten to twenty minutes improves alertness without much grogginess. Longer naps risk waking mid-deep-sleep, which produces sleep inertia — a period of markedly worse performance for perhaps half an hour after waking. If you nap before something that matters, keep it short or allow the inertia time to clear.

Using the tests to see it

The comparison that works is you against yourself, on the same device, at roughly the same time of day. That third condition matters: most people are measurably slower in the early afternoon and shortly after waking, so comparing a 9am result against a 3pm one confounds time of day with everything else.

The daily round combines eight tests into one score, which smooths some of the noise from any single measurement. Over a fortnight, the nights you slept badly are usually visible without needing to remember which they were.

Two caveats on reading that. Expect your first several attempts at any test to improve sharply through familiarity rather than anything else — start the comparison after that plateau. And a browser test measures your hardware as much as you, so a change of laptop resets the baseline entirely.

What “sleep debt” actually means

The phrase is used loosely, and the loose version is misleading in a specific way.

Short nights accumulate: the impairment after several nights of six hours is worse than after one, and it keeps compounding for some days before levelling off. That much is well supported, and it is the useful sense of the term.

What is not supported is the ledger model — the idea that you can precisely repay a quantified debt with an equivalent lie-in. Recovery is real but it is not arithmetic. A long weekend improves things substantially without returning you to baseline, and the recovery lags the sleep rather than tracking it.

The practical consequence is that consistency beats catching up. Seven hours every night produces better numbers than five on weekdays and ten at the weekend, even though the totals are similar.

What it looks like on the other tests

Sleep loss does not hit every task equally, and knowing which ones move is useful for reading your own history.

Sustained attention suffers most. Reaction time and the Stroop test — both of which require holding a state of readiness — are the most sensitive. Expect these to move first and by the most.

Short-term memory suffers moderately. Digit span and sequence tasks decline, though less dramatically than vigilance. The characteristic failure is losing the middle of a sequence, which is where attention has to be maintained rather than merely captured.

Well-practised motor tasks hold up best. Typing speed is comparatively resilient, particularly for a fluent typist, because motor sequences run with relatively little executive involvement. Accuracy usually degrades before speed does.

That pattern is itself informative. If your reaction scores have collapsed while your typing is unchanged, that points at attention rather than at anything more general.

Two practical uses

Calibrating your own sense of it. Take the tests on a day you know you slept well and on a day you know you did not, and look at the difference. Most people find it larger than expected, and having seen it as a number makes the subjective feeling easier to trust later.

Deciding what to do with a bad day. A measurably slow morning is a reasonable argument for moving the demanding work and doing the mechanical work instead — rather than pushing through and producing something you will redo.

When it is not sleep

If you are sleeping seven or eight hours and still exhausted, that is a question for a GP rather than a timer. Anaemia, thyroid problems and sleep apnoea all present first as an inability to concentrate, and all are straightforward to test for. Sleep apnoea in particular is easy to miss from the inside, because the sleep feels long — it is the quality that has gone.

Nothing on this site is diagnostic, and a run of poor scores is a prompt to ask a question rather than an answer to one.