Psychomotor Vigilance Task

Dinges and Powell built this task in 1985 for sleep research. It has two properties almost no other cognitive test has. Practice barely helps, so repeating it does not inflate your score. And being clever barely helps either.

What is left is a test that responds almost entirely to how you are right now. That is why sleep researchers made it their standard tool.

Van Dongen and colleagues ran the best-known study. They cut people's sleep for two weeks. Lapses climbed the whole way through. The volunteers' own ratings of how sleepy they felt levelled off early. People stop noticing the damage long before it stops piling up.

The classic ten-minute version counts any response slower than 500 ms as a lapse. This is the three-minute version, tested by Basner and colleagues in 2011.

It changes two things to make up for the shorter run. Signals arrive 1 to 4 seconds apart rather than 2 to 10. And a lapse now means anything slower than 355 ms.

Those two sets of numbers cannot be swapped. A lapse count from a three-minute run at 355 ms is not the same thing as one from a ten-minute run at 500 ms. Mixing them up is a common way to be wrong about your own data. This tool labels which version you ran, and the reference distribution covers the three-minute one only.

In the original machine the thing you react to is a counter. It starts running and stops when you press, so you watch your own time pile up.

Most web versions drop that detail. Dropping it removes the feedback that keeps people going through a deliberately dull three minutes.

It also makes false starts obvious. Pressing before the counter appears is scored on its own rather than folded into your times. More than a handful of false starts ruins a run. It usually means you were guessing the gap rather than reacting. Beating that guess is exactly what the random gap is for.

This is the one test on the site genuinely worth repeating. Run it at the same time of day, on the same hardware, for a week.

Lapse counts that climb over a fortnight are the clearest signal about your own sleep debt a browser can give you. Unlike almost everything else here, practice does not muddy the number.

It is not a medical test and it cannot tell you whether you are safe to drive. A research version is run under controlled conditions and read alongside other measures. If you are falling asleep during the day despite sleeping enough at night, that is a matter for a doctor.

For how much any reaction-based measure varies between people, see average reaction time by age and what counts as a good reaction time.