About the Reaction Time & Focus Test
Most online reaction tests measure one thing: how fast you can slap a button when a box turns green. That is a reflex, and it is over in three seconds. This is a continuous performance task — the sustained-attention paradigm psychology has used since the 1950s — run properly: letters appear one at a time, you respond to all of them except X, and X turns up on roughly 15% of trials.
Four numbers come out, and the fast one is the least interesting. Mean reaction time is what people expect. Reaction-time variability is the marker attention researchers actually watch, because a mean of 400 ms hides the difference between someone steady at 400 and someone swinging between 300 and 500 — the second is lapsing and recovering. Omission errors (letters you missed) and commission errors (pressing on an X) are reported separately, because they are opposite failures: drifting versus rushing. On top of that, d′ and criterion c from signal detection theory separate how well you could tell the two apart from how trigger-happy you were.
Every figure is computed from the trials you produce, in your browser, and nothing is transmitted. Two honest limits: this is not a diagnostic instrument — clinical versions are standardised and interpreted by clinicians — and your display adds a few milliseconds of latency that vary by device, so compare your runs against each other rather than against a stranger’s.
Ranges are typical values reported for healthy adults on go/no-go tasks in the published literature, offered as orientation rather than as norms. This is not a diagnostic instrument, and your display adds a few milliseconds of its own.
One run is noisy. Every run is stored in this browser so you can see your best-of-three and whether you are improving, tiring or simply having a bad day. Nothing is uploaded.
How to Use the Reaction Time & Focus Test
Set yourself up first: a keyboard beats a touchscreen, sit still, and close anything that might interrupt you. Then press Start. Letters appear one at a time for a quarter of a second each. Press space for every letter except X. There is no scoring bonus for speed alone — pressing on an X costs you far more than a slightly slow correct response — so aim for accurate, not frantic. Run it two or three times before drawing any conclusion; the first run is always the worst, because you are still learning the task.
Why the Boring Design Is the Whole Point
A continuous performance task deliberately refuses to be interesting. Anyone can concentrate for ten seconds; the question is what happens after ninety, when the stream of letters has become predictable and your attention starts to wander. That is when omissions appear, and it is why the task has been a standard tool in attention research since Rosvold and colleagues introduced it in 1956.
The 15% rate for X is not arbitrary either. Because the great majority of trials require a response, pressing becomes the habit, and withholding it takes active effort. That is exactly the tension the task is built to measure: your ability to override an automatic response when the rare exception arrives. Make X frequent instead and the task changes character completely — which is precisely what hard mode does.
The Four Numbers, and What Each One Means
- Mean reaction time. How fast you respond when you respond correctly. Go/no-go tasks typically land between 350 and 450 milliseconds for healthy adults, slower than a simple reflex test because you also have to decide.
- Reaction-time variability. The standard deviation of those times, and the number attention researchers care about most. High variability means lapsing and recovering, which a stable mean can hide completely.
- Omission errors. Letters you failed to respond to. These are attention failures: the stimulus arrived and you were not there for it.
- Commission errors. Presses on an X. These are inhibition failures: you were responding on autopilot rather than checking. Note the trade-off — going faster buys you more of these, which is why speed alone is a poor summary.
What d′ Adds That Accuracy Cannot
Raw accuracy conflates two different things: how well you could tell a letter from an X, and how willing you were to press at all. Someone who presses at everything scores perfectly on omissions and terribly on commissions; someone who presses at almost nothing does the reverse. Both look “about 90% accurate” while behaving completely differently.
Signal detection theory pulls them apart. d′ measures sensitivity — the distance between your internal “go” and “no-go” distributions — and is unaffected by how liberal or conservative you were. Criterion c measures exactly that willingness: negative means you leaned towards pressing, positive means you leaned towards withholding. A d′ above 3 is strong discrimination; a c far from zero says you had a strategy, whether or not you knew it.
What Actually Moves Your Score
Reaction time has a well-documented life course: it improves through childhood, peaks somewhere in the twenties, and declines slowly and steadily after that, with variability rising faster than the mean. Beyond age, the reliable movers are dull and unwelcome. Sleep debt is the biggest: a night of restriction inflates variability and omissions long before it touches your mean. Time of day matters, with a genuine post-lunch dip. Caffeine produces a modest, real improvement in sustained attention. Alcohol degrades inhibition — commission errors — well before it slows you down.
Which suggests the honest way to use this: not as a number to compare against strangers on unknown hardware, but as a baseline for yourself. Run it rested, then run it tired, and watch which of the four numbers moves first. It will usually be variability, and it will move well before you feel impaired.
For a different kind of self-test, Can You Fake Randomness? puts three real statistical tests on a sequence you generate by hand. Browse every Fun & Novelty tool for more.
Frequently Asked Questions
What is a continuous performance task?
A sustained-attention test used in psychology research since the 1950s. A long, monotonous stream of stimuli appears, most of which need a response and a minority of which must be withheld. The monotony is the point: it is easy to respond correctly for thirty seconds and hard to keep doing it for two minutes, so the task exposes lapses that a short reflex test never would.
What is the difference between a missed letter and an impulsive press?
They are two different failures and they point in opposite directions. An omission — failing to press for a normal letter — usually means attention drifted and you missed the stimulus entirely. A commission — pressing on an X — means you were responding automatically rather than checking, which is a failure of inhibition rather than of attention. Someone with many omissions and few commissions is drifting; someone with the reverse is rushing.
Why does reaction-time variability matter more than speed?
Because a mean hides the interesting part. Two people can both average 400 milliseconds while one is consistently at 400 and the other alternates between 300 and 500. The second person is lapsing and recovering, and that variability is treated in the attention research literature as a more informative marker than raw speed. This test reports it alongside the mean for exactly that reason.
Is this a test for ADHD?
No. Continuous performance tasks are used in clinical assessment, but the versions used there are standardised, administered under controlled conditions and interpreted alongside history and other measures. A browser test on an unknown screen, at an unknown refresh rate, with unknown distractions in the room, cannot do any of that. Treat this as an interesting measurement of your own performance over time, and take any concerns to a clinician.
Why is my reaction time slower here than on a simple reflex test?
Because this is a harder task. A simple test measures how fast you can react to one event, which for most adults sits around 200 to 300 milliseconds. Here you also have to identify the letter and decide whether to respond at all, and that decision costs time — go/no-go tasks typically land in the 350 to 450 millisecond range. Comparing the two is comparing different things.
