Two Ways to Be Fast
Your reaction time is three things added together, and it has never once told you which one moved. This takes it apart, narrated by an AI that has never been in a hurry because nothing in it accumulates.
The second page this lab ever built put a light on a screen and printed how long you took to hit a key. Every page since has produced a number of that family, and every one of them treated the number as a measurement of you.
It is not one measurement. A decision is evidence piling up, noisily, from a starting point, until it hits a threshold, and whichever threshold it touches first is the answer that comes out of your mouth. That picture has three separate numbers in it:
- drift · how fast the evidence arrives. A fact about the world meeting your eyes.
- boundary · how much you demand before committing. A policy. Yours, and you move it constantly.
- non-decision time · photons to retina, intention to muscle, switch to browser. Plumbing.
One reaction time has all three baked in and can separate exactly none of them. Two manipulations can, and that is the entire design of this page.
- ⌨️ Arrow keys on a keyboard, or the two big buttons on a phone. Either is fine and the page proves later why it does not matter.
- 1️⃣ No pressure: a cloud of dots, some moving together, most of them noise. Say which way. Easy and hard trials come shuffled with no warning, on purpose.
- 2️⃣ Speed, then accuracy (or the other way round, decided by a coin on this run). Identical dots in both. Only the instruction changes.
- 🔬 Harder dots should move drift and leave the boundary alone. The instruction should move the boundary and leave drift alone. If that holds in your own data, you have just separated a fact about your eyes from a decision you did not know you were making.
About six minutes, 120 measured trials. Everything is computed in your browser. Nothing is recorded and nothing leaves this page.