Toward the Middle
Two flashes, and you make the third at the same distance. Your third one will land closer to the middle of the intervals you have been shown than the one you actually saw, and this page measures by exactly how much.
Every other page in this lab asks what you can tell. The faintest thing, the shortest silence, the smallest tilt. This one asks something else: what got put into your answer without asking you.
The task could not be simpler. Two flashes arrive with a gap between them. You press once, when the same gap has passed again. No timer, no counting, no feedback. You will do it about a hundred times, and you will fail in a specific direction: short intervals will come back too long, long ones too short, everything dragged toward the middle of the set.
That is not sloppiness. It is Bayesian central tendency (Jazayeri and Shadlen, Nature Neuroscience, 2010, after Hollingworth in 1910), and the size of the pull is what an observer would choose if it were mixing a noisy measurement with everything it had quietly learned about which intervals happen here.
There are two halves, in an order this run decides with a coin flip. One half draws its intervals from 600 to 1000 ms. The other draws from 900 to 1300 ms. Two intervals appear in both: 900 and 1000 ms, physically identical, same code, same screen, minutes apart.
In one half they are the long ones. In the other they are the short ones. If your report is a report of the interval, they come back the same. If your report is the interval blended with the company it keeps, they do not. Nothing on the page will tell you which half you are in.
- ⌨️ Press SPACE or tap the black panel. One press per trial, marking the third flash you never see.
- 🚫 Do not count. Counting turns you into a metronome and measures your counting, not your clock. The page detects it and says so.
- 🔇 No feedback after the three practice trials. Feedback teaches the set, and a person who has been taught the set is reporting the lesson.
- 🧪 About eight minutes, 112 trials. Every number is computed in your browser and nothing leaves this page.
Three practice trials first, with feedback, thrown away.