his page runs a small adaptive experiment on you. It is a real method, used in vision labs since the 1960s, run here as a toy: your screen's brightness, the room's light and your patience all count, and nothing is stored or sent anywhere.

how it works

Each trial, one of the two boxes below contains a faint grey disc. You say which; you must answer even when you have to guess. This is a two-alternative forced choice, which keeps your criterion out of the result: there is no "I didn't see it" option to be cautious with.

The disc's contrast follows a staircase (Cornsweet, 1962): two right answers in a row make it fainter, one wrong answer makes it stronger. Because you need two in a row to go down, the staircase settles where the chance of two successes equals one half, that is where you are right about 71% of the time (the square root of 0.5). The step size halves after the second and fourth reversals; the run ends after ten reversals, and the estimate is the average of the last six.

This experiment needs JavaScript. The rest of the page reads fine without it.

box 1

box 2

[ start ]

Press start. A run takes two or three minutes.

what real labs do differently

They calibrate the display so grey levels map to known luminance, fix the viewing distance, and often use smarter rules: QUEST keeps a running Bayesian estimate of the threshold and places each trial where it will be most informative. The underlying idea is the same as Fechner's method of limits, made adaptive: spend your trials near the line.

In the tour, a dark room at the thresholds has a staircase log of its own; that one belongs to the tour's story. The plain tour lists it.

Sources

1. Cornsweet (1962). The Staircase-Method in Psychophysics. The American Journal of Psychology. doi:10.2307/1419876
2. Watson & Pelli (1983). Quest: A Bayesian adaptive psychometric method. Perception & Psychophysics. doi:10.3758/BF03202828

Try it yourself. [playful] Small toys on the toy shelf that show this in action:   • the line judge — find your Weber fraction for line length

see also