How does the Delayed Match-to-Sample Test work?
A pattern shows for 3 seconds and disappears. After a wait of 0, 4 or 10 seconds, 4 patterns appear, and you pick the one you saw. There are 15 patterns, 5 at each wait, in a random order, so you never know how long you will have to hold one. Only the first is sure to have no wait, to show how the test goes.
Each pattern is a 3 × 3 grid of tiles, and each tile is a quarter disc in one of its four corners. Neighbouring discs join into blobs, waves and bites that are easy to see and hard to put into words, which is the point: a pattern you could describe in a sentence would test your words, not your picture of it. A bar fills during every pause, so you can see the wait running. Nothing is timed; only your pick counts.
How is the Delayed Match-to-Sample Test scored?
Your score is how many of the 15 patterns you matched, with your right answers at each wait beside it. Guessing alone gets 1 in 4, about 4 of 15, so a score well above that is memory. Comparing the waits shows whether a longer wait costs you: a score that falls from no wait to 10 seconds is a picture fading.
With 5 patterns at each wait, one lucky guess or one slip moves that wait's score by a fifth, so read small differences lightly. How often guessing alone reaches a score at one wait:
| Right out of 5 at one wait | Chance of at least that by guessing |
|---|---|
| 2 or more | 37% |
| 3 or more | 10% |
| 4 or more | 1.6% |
| 5 or more | 0.10% |
Why are the Delayed Match-to-Sample patterns so alike?
Because a memory test has to be impossible without memory. Each of the 4 patterns is one tile away from a base pattern you never see, each at a different place, so any two differ in exactly two tiles and none stands out from the rest. Comparing them with each other points nowhere; only remembering the one you saw does.
The obvious way to build the choices, three copies of the answer with one tile changed each, gives the answer away: every tile the answer has then appears in at least three of the 4, and a tile-by-tile vote finds it without any memory at all. The tests that check this site's code confirm the 4 options here are interchangeable on every deal.
What is delayed matching-to-sample?
Delayed matching-to-sample is a memory task in which a sample disappears before the choices appear, so the match has to be made from memory, and the wait can be made longer or shorter. Donald Blough (1959) used it to study memory in pigeons, and it has been given to people too.
Barbara Sahakian and colleagues (1988) gave a computerized version, with waits of 0 to 16 seconds, to patients with Alzheimer-type dementia and with Parkinson's disease. The dementia group did worse as the wait grew, but not when the sample stayed in view; the medicated Parkinson's group was impaired at every wait, and with the sample in view too. This test is a recreational version with its own patterns, and it gives no diagnosis.
What can affect your Delayed Match-to-Sample score?
What you do during the wait, most of all: looking away, thinking of something else or talking all crowd the picture out. Naming parts of the pattern (“bite top left, wave across the middle”) can help, though the tiles are chosen to make that hard. Tiredness, a small screen and glare count too. To remember which tiles lit up rather than a whole picture, try the Visual Memory Test; to hold a color instead of a shape, the Color Memory Test.
What doesn't the Delayed Match-to-Sample Test measure?
It measures one kind of memory: recognising a picture you saw seconds ago among close look-alikes. It doesn't measure remembering things for hours or days, recalling something without a choice in front of you, or memory for words and numbers. It isn't a screening test for dementia or any other condition.