Each card has a short phrase about something a person did. You drag it by colour only. At the end you see whether the words changed your speed anyway.
Three steps, no questions about yourself, and a result you can read in a minute.
Browsers report time differently on different hardware. Your refresh rate and clock resolution get measured first, so the numbers afterward mean something.
Blue goes one way, yellow goes the other. The words on the cards are irrelevant to the instruction, and you are asked to ignore them entirely.
If the words really were irrelevant, your speed would not depend on them. Where it does, by how much, and whether that is even measurable is the result.
No. It is a demonstration of a research method, not a clinical or psychological assessment. It is not a personality test, and it is not a measure of whether you are a good person. If you are looking for mental health care, this is not that, and it is not a substitute for talking to a professional.
Nowhere. There is no server and no account. Everything is computed in this browser tab and stored only in this browser's local storage, which you can clear from the "How it works" page at any time. No name, no email, no IP address, no analytics on your responses.
Honestly: not very, and the page says so plainly on your result. Reaction-time measures like this one are useful for comparing groups and weak at describing an individual. The full reliability numbers, including the ones that do not flatter this task, are published on the "How it works" page.
You get a practice round first, and errors are expected. If too many trials come back unusable, the task tells you it cannot score this attempt rather than showing you a number that would not mean anything.
Built by Mohith Gajjela as an independent implementation from the publicly registered protocols osf.io/n3kyu and osf.io/v46q9 by Paul C. McKee (Duke University). Not affiliated with, endorsed by, or connected to MADLAB, Duke University, or any research study. mohithgajjela.com
Reaction times collected in a browser overshoot the truth by tens of milliseconds, and by how much depends on your hardware. So it gets measured rather than assumed.
Hold still for a moment.
A card appears in the middle of the screen with a short phrase on it. Ignore the words. Respond only to the colour, as fast as you can.
The card grows as it comes toward you and shrinks as it moves away. Halfway through, the colour rule flips, and you'll get a warning before it does.
Speed matters more than perfection, but errors are recorded. Keep your hand on the card between trials. Press Esc at any time to stop.
A task can produce a large, perfectly replicable group difference and still be useless for telling two people apart. That is the reliability paradox, and it is why most implicit measures disappoint. Rows marked SIM are simulated respondents, including a deliberate 12% who answer faster than a human decision allows, so the quality screens have something to catch.