Time
In a run of identical things, a rare surprise seems to last longer than it does. Peter Tse, Josée Rivest, Patrick Cavanagh, and I reported this in Perception and Psychophysics in 2004. The two demos below let you measure it on yourself.
Try it: the oddball
Your own expansion factor
Plain black circles come and go, each lasting about a second. Now and then a disk appears that grows. In the paper, the growing disk felt about one and a half times as long as it was.
After each disk that grows, say whether it lasted longer or shorter than the circles. Use the buttons or the arrow keys. It takes about three minutes.
Your curve
In the paper, the standards were still black circles that each lasted 1,050 ms. The oddball was a black disk that grew from 30 to 100 pixels in radius. To feel as long as a standard, it only had to last 675 ms on average.
The Psi method (Kontsevich and Tyler, 1999) picks each disk's length to learn the most from your answers. Your point of equality is the length at which you say "longer" half the time. Your factor is 1,050 ms divided by that length.
About one question in four comes after a plain circle whose length was changed a little. These check trials should come out near a factor of 1, so a habit of saying "longer" shows up there instead of passing for a stretch. You have until the third circle to answer, and a missed question comes back later.
With 7 to 12 circles between questions, as in the paper, 30 oddballs take about fifteen minutes. The paper locked its timing to the screen's refreshes. A browser cannot promise that, so this page times each shape by the frames that drew and cleared it, and reports how well your screen kept up. This is an informal measure on your own screen.
Six kinds of surprise
The paper tested five more oddballs among 1,050 ms standards: a red circle among black ones, a circle among squares, a square among circles, a large circle among small ones, and a still circle among growing ones. Each felt longer than it was, and matched a standard when it lasted about 760 to 860 ms, against 675 ms for the growing disk.
Each run has 12 oddballs and 3 check trials, and takes about three minutes. Each bar is your factor for that kind, with its uncertainty. The dark tick is the paper's average, 1,050 divided by the point of equality it reports: red among black, 797.5 ms; circle among squares, 760 ms; square among circles, 823.5 ms; large among small, 808.4 ms; still among growing, 858.8 ms; and growing among still, 675 ms. The check bar pools the check trials from every kind you have run, and should sit near 1.
The stretch over time
With standards from 75 ms to 2,100 ms, the paper found no stretch until at least 120 ms after the oddball appeared, and at 75 ms the oddball even seemed shorter. The stretch peaked near 225 ms, dipped, and then rose again. The paper's account is a quick burst of attention followed by a slower, steadier one. It found nearly the same time course for sounds.
The sweep runs one short block for each duration you tick, in a random order, with 2 check trials in each. The chart marks only what the paper states in words: no stretch before about 120 ms, and a peak near 225 ms. Nothing here is shown for less than 100 ms, because a browser cannot hold shorter times steady. So at 150 ms the disk can be no shorter than two thirds of a standard, and a factor above 1.5 cannot show. A small arrow marks a point that reaches that floor.
- Perception and Psychophysics 66(7), 1171 to 11892004. "Attention and the subjective expansion of time." Tse, Intriligator, Rivest, and Cavanagh. Read it
The paper's account: an unexpected event draws attention to it, and attention raises how much is taken in while the event lasts. More is registered, so it seems to last longer.
Try it: hold to match
Show it with your hand
Choosing between two words can lean on habit. Holding a button for as long as something lasted is a second way to ask. The paper used this as a control, and it showed the same pattern over time, though without the early peak.
When the circles stop, hold the button down, or the space bar, for as long as the last shape lasted. Twelve holds take about three minutes.
Your holds
Some stops come after a growing disk and some after a plain circle. The disks last 750, 1,050, or 1,350 ms, and the circles 1,050 ms. Holding tends to run long for short times and short for long ones, so the number that counts is a ratio: how long you held a disk for its length, over how long you held a circle for its length. A ratio above 1 means the disks felt longer in your hand too.
In the paper's version, five disks blinked in turn and observers pressed a key for as long as the last one lasted. The last disk either stayed the same size or grew, and the paper compared the two. This is an informal measure on your own screen.