Nothing To Watch
Cancelling a movement that is already on its way takes you a fixed amount of time. It produces no keypress, no reaction time and no trace of any kind. This page times it anyway.
The last four pages in this lab measured what a decision costs: your hand as a channel, choosing as a channel, the queue that runs one decision at a time, and the price of changing which decision you are making. Every one of them ended in a finger moving. Every one of them had an event to time.
This one does not. This one asks what it costs to cancel a decision that is already on its way out of you.
and the nothing has a duration
An arrow appears. Left arrow, press left. Right arrow, press right. On one trial in four the arrow turns red shortly after it appears, and on those you are to cancel: press nothing at all. That is the stop-signal task, and the interesting part of this page is not the number it prints. It is that the number can exist.
Logan and Cowan (1984) called it a race. One process is trying to move your finger; it starts at the arrow. The other is trying to cancel it; it starts at the red. Whichever finishes first wins, and they do not talk to each other.
Hundreds of ordinary trials give the whole distribution of how long the first process takes. The moment the second one starts is known exactly, because this page chose it. So if you also know what fraction of red trials produced a press anyway, you can find the finish line by sliding a line through your own distribution until the part behind it is that fraction. Everything behind the line was fast enough to escape. The line is where stopping finished.
recovered from the trials where stopping failed
The delay before the red walks itself around under you: later when you succeed, earlier when you fail. You are not meant to win. You are meant to win about half.
Slowing down on every trial to catch the red ones. It is the sensible thing to do and it makes this measurement impossible, because then nothing is racing anything. The page can tell: if the trials where you failed to stop come out reliably slower than your ordinary trials rather than faster, the model is broken for this run and the headline is refused instead of printed.
So answer the arrows at your ordinary speed and let the red ones catch you out. Failing half of them is the design working.
On a phone the two buttons at the bottom do the same job. Nine blocks, about ten minutes, two unscored warm ups first, escape aborts at any time. A run with more than a quarter of its arrows answered on the wrong key is refused rather than scored.
Everything runs in your browser. Nothing is recorded. Nothing leaves the page.