A face flashes for a tenth of a second. Decide whether its mouth was short or long, ninety times over, while some correct answers quietly pay you five cents.
Attention to detailLearningReward response
Unofficial practice. Not the real Pymetrics game and not connected to
Pymetrics, Harver or any employer.
Untimed,
unlimited retries, nothing saved and nothing sent anywhere.
How the Pymetrics lengths game works
A cross appears, then a face with no mouth, then the mouth for a tenth of a second.
Press the left arrow for a short mouth and the right arrow for a long one, or tap the two keys below the face.
The two mouths differ by about a tenth. At that exposure you are not meant to be certain.
Some correct answers pay five cents and say so. Most answers get nothing at all.
A correct answer that does not pay looks exactly like a wrong one. The game never tells you which you gave.
There is no clock. Ninety faces and it ends.
What the lengths game measures
How far your answers drifted toward whichever mouth was paying more often, kept separate from how well you could actually tell the two apart. Those are two different things about you and the game reads both.
Or play all 12 in one sitting
This page is the lengths game on its own, practice only.
The full run is the rehearsal. All 12 back to back in the order the real
assessment most commonly uses them, timed, one attempt each, and no feedback until the end.
About 35 minutes, and nothing is saved, so it needs one sitting.
Answered straight, including the parts people would rather hear differently.
Because that is the whole design, not an oversight. There is something else going on in this game that you are meant to pick up on without being told, and per answer feedback would wipe it out. You would end up learning to see the difference instead, which is not what is being measured. The site tells you what was going on at the end of the run.
One of the two mouths pays out three times more often than the other, and which one is decided at random when your run starts. Nobody tells you. Because the flash is so short that you genuinely cannot be sure, your answers slowly lean toward whichever one has been paying, and how far they lean is the measurement. It has a name in the research, response bias, and it is separate from your eyesight.
No. It is built that way on purpose. The two mouths differ by about a tenth and you see them for a tenth of a second, so most people land somewhere around two thirds right. If you were getting nearly all of them right the game would have nothing to read, because there would be no uncertainty for the payouts to lean on.
That would be answering a different question than the one on screen, and it also is not worth much. The most you can earn across the whole run is $1.80, and nobody is scored on the money. There is no correct answer to aim for here, only a real one.
The three to one ratio and the five cent reward come from the published research task this game is built on, and one reconstruction of the Pymetrics version reports the same ratio. Pymetrics does not publish its own numbers, so treat ours as a faithful reconstruction rather than the official figures.
Browsers slow down or stop pages you are not looking at, and a tenth of a second stimulus is the one thing that cannot survive that. Left alone, the mouth would stay on screen far too long and the whole run would be worthless while still looking normal. So the game stops and waits for you instead of spending faces you were not watching.
No, and this is the one game on the site we cannot make work another way. The whole task is judging a difference too small and too brief to describe in words. Anything that read the face out loud would just be reading out the answer. It is the only test here with that problem and we would rather say so than pretend otherwise.
Pymetrics and Harver are trademarks of their respective owners and are not affiliated with
OpenThinkr. Practice results are for interest and self-reflection, not assessment.
