Mechanical Reasoning Test

Two gears with their teeth in each other cannot turn the same way. Push the top of one forward and the tooth it presses on has to go backward.

So a row of meshed gears alternates. The second turns against the first, the third turns with it, the fourth turns against it again.

That makes the whole question odd or even. Count the meshes between the handle and the wheel you were asked about. An odd count reverses the direction and an even count keeps it.

A chain does nothing of the sort. It drags the far wheel round the same way, which is why the pedals and the back wheel of a bicycle go round together.

One chain in the middle of a train breaks the alternation. Counting wheels stops working at that point. Counting meshes still works.

• Meshed wheels touch, teeth in teeth. The direction flips.
• Chained wheels sit apart, with a loop drawn round both. The direction holds.
• The arrow marks the handle. The question mark marks the wheel to answer for.

Mary Hegarty timed people reading drawings of pulley systems in 1992. They did not take a machine in at a glance.

They worked it outwards from the handle, one join at a time. The further the question sat from the handle, the longer the answer took.

Hegarty and Sims followed that in 1994 and found the reach differs from person to person. Some run out after a few components and some keep going.

So the honest number is where you run out, rather than how many you happened to get. This page looks for it by adding a wheel each time you clear a rung.

Two machines to a rung, both right, or the run ends there. A coin gets one machine right half the time, so a rung that took only one would hand a guesser a five.

Nothing here asks about force, speed, load or how many turns. Every machine asks the same question, and it is always which way.

That is narrower than a real mechanical paper, and the narrowness is deliberate. A question about gear ratios is arithmetic wearing a picture, and the numerical reasoning test already asks the arithmetic without the picture.

Wheel size is the giveaway that had to go. Every wheel on this canvas is the same size, so nothing in the drawing hints at a ratio you were not asked about.

The trains curl round rather than running in a line. That is a drawing decision and not a puzzle: eight wheels in a row on a phone are eight wheels too small to see.

Mechanical comprehension papers turn up in apprenticeships, in the armed forces, and in the sift before a technician interview.

Those papers are much wider than this. They ask about levers, springs, hydraulics, pulleys under load and which of two spanners takes less effort.

So use this page for the part it covers. Reading a linkage is what the direction items rest on, and it is what separates a slow paper from a quick one.

The rest of a paper is worth practising elsewhere. One browser canvas is not an aptitude battery, and saying otherwise would be selling you something.

There is no bar under your number, because there is no distribution to draw one from.

Hegarty published times and error rates from small samples on drawings of her own. Those are findings rather than a curve, and her pulley systems are not these gear trains.

The publishers who do hold norm tables for mechanical papers keep them with the item sets they belong to. None of those items is on this page.

So the page reports the depth and says plainly that the comparison does not exist. Where one does exist we publish the figures and the arithmetic behind them, as with how processing speed changes with age.

Our own norm goes up when a cohort here reaches a thousand runs, and not one run before that.

It means you read a six-wheel train twice in a row. It does not mean you would be good at fixing a gearbox.

Nothing on this page measures mechanical skill, aptitude for a trade, or how an apprenticeship would go for you.

Getting better at these trains is getting better at these trains. That is what researchers call near transfer, and whether brain training games are worth it covers what happened when the wider claims were tested.

A short run is a short measurement. One slip ends a rung, so a single number here is a snapshot rather than a level you have reached.