How does the multitasking test work?
A shape with dots inside appears in the top or the bottom half of a frame. In the top half, answer its shape: left for a diamond, right for a rectangle. In the bottom half, count its dots: left for two, right for three. Both rules use the same two buttons, so the same figure can need opposite answers depending on where it is.
There are three parts: 16 figures that only appear on top, 16 that only appear below, then 32 that can appear in either half, so you never know which rule comes next.
How is your switching cost calculated?
Your result is your multitasking cost: how much slower you answered in the mixed part, on figures that kept the same rule as the one before, than in the single parts. It measures the price of having to be ready for either task. The switch cost, shown beside it, is how much slower you were when the rule did change.
Only right answers that followed a right answer count, between 150 ms and 3 seconds, and the first figure of each part is left out. That is how the study this test is built on cleaned its times.
Where does this test come from?
It follows the design of a 2013 study by Gijsbert Stoet, Daryl O'Connor, Mark Conner and Keith Laws, in BMC Psychology: the same frame, the same two rules and the same left and right answers. They tested 120 women and 120 men. After practice, people answered 319 ms slower on average when the tasks were mixed than when doing one at a time (763 ms against 444 ms).
The slowdown was a little larger in men than in women (336 ms against 302 ms), a modest difference, and the authors warned against strong generalisations from it. In everyday tasks on paper, men and women did about as well, except that women were better at planning how to find a lost key.
Is multitasking a myth?
Doing two things at once, in the sense of answering two rules in the same instant, isn't what happens here: each figure needs one rule. What costs time is being ready for either rule. Nearly everyone is slower in the mixed part, which is why the result is a cost, not a score to max out. A smaller cost means you switch more smoothly.
For ignoring what doesn't matter, try the Stroop Test or the Flanker Test.
What can affect your switching cost?
Less than you might think, because the result compares you with yourself. A slow screen or a wireless keyboard adds the same delay to every answer, so it cancels out when the single parts are subtracted from the mixed one. What does matter is the first few figures of each part, while you get used to the rule (the first of each part is left out for that reason), and mistakes: an answer after a mistake doesn't count either.