Methodology

A test that runs in a browser can only be as exact as the browser. Each press is stamped with the time the browser recorded when you pressed, not the time the page got round to reading it. The signal is stamped when it was drawn on the screen, not when the page asked for it. Between the two stamps lies your reaction.

A screen shows a new picture only so many times a second, so a signal can be shown a little later than it was planned. If the page stalls while a signal is showing, that try does not count and you do it again.

A browser does not know how long your screen takes to light up, how long a touch screen takes to notice your finger or how long a mouse or keyboard takes to send a press. Together these add a delay that is hidden from us, and it is different for every device.

Robot tests of everyday computers and phones measured it^[1][2]. For the reaction tests we allow for what they found beyond the study's own equipment: 44.1 ms to 115.1 ms with a keyboard or mouse, and 39.8 ms to 52 ms with a touch screen or pen. No study timed a mouse, so a mouse gets the keyboard's figures. That is why we show a range, on the quicker side of the time you see, and not a false exact figure, and why a result from a phone is not the same thing as a result from a computer. For the fairest look at how you are changing, use the same device each time.

• The first try of a timed test is for practice and does not count.
• A press that comes before the signal, or so soon after it that it must have been a guess, is not counted, and you do that try again.
• A try that takes far too long is shown as too slow and is left out of your result.
• If you switch to another tab or app during a try, that try does not count.
• A run that looks as if a machine made it, for example clicks spaced exactly evenly or faster than a hand can go, is kept but not compared with anyone.

Your result is usually the middle one of your counted tries, so one lucky or unlucky try does not decide it.

We compare you only with a named sample: a published study, with its authors, its year, how many people took part and who they were. What we may say depends on how good the sample is. Where a study is good enough, we say roughly how many of the people in it you were quicker than, as a range, because of the hidden delay. Where it only gives rough levels, we show a level. Where it gives only a published range, we mark the range.

Some figures are not studies: a number that a website publishes about its own visitors, or a range that players report for clicking. We call these reported figures and community ranges, we draw them differently, and we never call them a norm. Where there is no sample at all, we show only your own earlier results.

These are the verified samples in our registry. A test is compared only with a sample that fits what it measures, in the language and on the kind of device it was taken with.

Each questionnaire is scored the way its authors describe. Your answers are added up or averaged for each trait, some statements count backwards because they say the opposite of the trait, and the rules of that questionnaire decide what happens when you skip a question. A result is only given when enough questions are answered.

The page of each questionnaire lists its statements, its scoring key, its licence and its source. Where a questionnaire has no published comparison group for your language, the result shows where you sit between the two ends of each trait and says that it does not compare you with other people.

Selfspecs is written in English first. The pages in other languages are translations made by Selfspecs and have not yet been checked by native speakers. For every questionnaire, its page says whether the statements are the original wording, a translation by Selfspecs or a translation that a native speaker has checked.

A figure that we have not been able to check against its source is not shown. Some of the explanations on this site were drafted with the help of a machine and are still waiting for a careful read by a person; the plain facts they contain come from the sources above.