Never the Waiting
Five words vanish under fifteen seconds of arithmetic, and everybody blames the fifteen seconds. On the first trial of a new category they survive it almost untouched. The page runs the manipulation that tells those two apart, and commits to a number before it turns the knob.
Five words arrive one at a time. Then a three digit number appears and you count backwards from it by threes, typing each answer, for either four seconds or fifteen. Then you type the five words back. That is the oldest experiment in the short term memory literature and it is famous for showing that a handful of items is nearly gone after fifteen seconds.
It is famous for the wrong reason. On the first trial of a fresh category almost nothing is lost at any wait. The steep curve everyone quotes gets built up over trials, by the trials themselves, and nothing about the clock changed while it happened.
The trace fades. Fifteen seconds costs you what fifteen seconds costs you, on the first trial and on the fortieth, because a clock does not care what came before it.
The wait only takes what the earlier trials gave it to take. Nothing competes on trial one, so nothing is lost. By trial three the same fifteen seconds is expensive.
- Forty eight scored trials plus two for practice, about seventeen minutes.
- Five words, then the counting, then type back what you have. Any order.
- At most five words per trial. There are only ever five, so a sixth is a guess.
- The counting is not filler. It is the only thing stopping you rehearsing, and every answer is checked.
- You are never told how long the wait will be, because being told would let you plan.
After eight blocks it stops, reads your own forgetting two different ways, and prints what each reading says about a trial nobody has run yet. Then it runs it. One of those two numbers is going to be wrong on your data, and which one is the entire point.
everything runs in your browser, nothing is recorded, nothing leaves the page