Mental rotation: the one cognitive skill that reliably improves with practice
The task has one rule, and it is easy to state. You look at a figure, then at four more. One of those four is the same figure, turned. The other three are flipped copies. Which one is the turn?
It looks like spot-the-difference. Timing data from 1971 says otherwise, and that small study is still one of the tidiest results in cognitive psychology.
The line that made this famous
Roger Shepard and Jacqueline Metzler sat people down with pairs of block figures and asked them one question: same object, or mirror image? People got it right. The part Shepard and Metzler cared about was how long each answer took.
Response time climbed in a straight line with the angle. A figure turned 60 degrees took longer than one turned 30, consistently so. At 120 people took longer again, and the line held right out to 180.
The straight line is the strange part. If you recognised the shape, or ticked features off a checklist, a bigger angle would cost you no extra time. A list is a list. A line like that points to something continuous instead: a representation turned at a roughly steady rate, nearer to handling an object than to consulting a memory. In some functional sense, people rotated a picture.
Accuracy tells you whether someone can do it. The slope tells you how they do it.
Why the flipped ones are the whole task
The distractors are mirror images rather than different shapes, and that choice is the whole design. You cannot rotate a chiral figure into its mirror image, however far you spin it. Turn the thing in your head and the reflections fall away on their own. Match features instead (blob on the left here, blob on the left there) and a reflection sails through your check, and you pick it.
Our own rotation test puts two reflections in every set of four, and we drew every figure asymmetric in both axes. Make one symmetrical and you have handed the player several correct answers, and nobody will notice.
The part that makes it worth practising
Most of what these tests measure barely budges. Sleep better, eat better, cut the distractions, and you have about used up your options. Spatial skill is the exception, and a conspicuous one. Mental rotation answers to practice more than almost anything else on this site, and it answers fast: most people get quicker within a few sessions, and the gains hold up better than the "brain training" literature would lead you to expect.
Researchers still argue about how far any of it transfers, and that argument is live. Getting quicker at rotating block figures makes you quicker at rotating block figures. Map reading, packing a boot, reading a technical drawing: results land on both sides, and anyone who tells you the transfer is settled is running ahead of the evidence.
Two strategies, and one is faster
You can solve these two ways, and you can usually feel which one you are using.
Holistic rotation turns the whole figure at once. That is the behaviour the Shepard result describes, and it is the faster route once you learn to trust it.
Feature matching picks a landmark, hunts for it in the candidate, then works outward from there. It feels safer and it runs slower, and on reflections it will mislead you, because a landmark survives a flip perfectly well.
Catch yourself hunting for the dot on the end and you are using the second one. Turning the whole shape in one go feels worse at first. Stick with it anyway.
What it is not
Mental rotation is one narrow spatial ability. General intelligence it is not. It says nothing about how visual a person you are, and it does not settle what work suits you. Researchers report group differences here more than in most areas. Those differences shrink a good deal with practice and swing about with the way the task is set up, which is worth knowing before you treat any single score as a fact about a person.
Ours runs to fourteen figures and takes about five minutes. For the number, take the test. To get better at it, take it twice a week for a fortnight, and watch the second number rather than the first.