What a bad night actually costs you tomorrow

The cost compounds, and it compounds quietly

The cleanest demonstration is Van Dongen and colleagues, who kept people on restricted sleep for two weeks and tested them daily. Six hours a night — not four, not an all-nighter, the amount a great many people consider a normal working week — produced cognitive deficits *equivalent to going up to two full nights without sleep at all*.

Two weeks of what looks like a slightly short night adds up to the same measurable damage as not sleeping for two days. Nobody feels like they have been awake for two days at the end of a normal fortnight, which is precisely the problem.

The finding that should bother you

In the same study, participants' own ratings of their sleepiness did not track their actual performance. They adapted to feeling somewhat tired, stopped noticing it getting worse, and carried on underestimating their impairment while the objective measures kept sliding.

This is the uncomfortable centre of the whole subject. The instrument you would naturally use to decide whether you are fit to drive, to make a decision, to have a difficult conversation — your own sense of how you feel — is the one instrument the evidence says stops working exactly when you need it.

And it is not only cognition

Spiegel and colleagues restricted healthy young men to four hours for two nights and measured the hormones that govern appetite. Leptin, which signals fullness, fell 18%. Ghrelin, which signals hunger, rose 28%. Self-reported hunger rose 24%, with the sharpest increase in appetite for calorie-dense, carbohydrate-rich food.

Two nights. Not a month of shift work — two nights. If you have ever wondered why a bad week reliably comes with more snacking and blamed your discipline for it, there is a hormonal component and it has been measured.

What this does and does not mean for your data

It means a short night is worth noticing even when you feel fine, because feeling fine is not evidence. It means the cost of a run of six-hour nights is not visible as a single bad morning; it is a slope. And it means the value of tracking sleep is mostly in the trend across a fortnight rather than the verdict on any one night.

It does not mean a low score should decide your day. These are population findings from controlled laboratory restriction, and none of them tells you what you specifically can do this morning. A study establishes that an effect exists and roughly how big it is; it does not license a device to hand you a verdict.

Which is why we withhold the score

There is a feature in Vitra, off by default, that hides today's scores until you have rated your own morning. The reason is exactly the Van Dongen finding: a number you read first colours the answer you give afterwards, and an anchored self-rating cannot be compared to anything.

Rate blind for a couple of weeks and the comparison becomes available — whether the measurement agrees with how you feel, and in which direction it disagrees. That is a genuinely useful thing to know about yourself, and it is only obtainable if the app shuts up first. It is also the only honest use we could find for the observation that people misjudge their own state: not to correct you, but to let you find out.

Frequently asked questions