Here's a piece of arithmetic worth doing.
Twenty questions on a test. A student gets fifteen right, and of the five they lost, three were errors they could have caught — a sign flipped, a unit dropped, a question where they answered the wrong part.
Catch those three and the mark goes from 75% to 90%.
Nothing new was learned. No extra studying happened. The material was already known well enough. Fifteen points were sitting on the desk in front of them for the last ten minutes of the test.
Almost nobody picks them up.
Because finishing feels like being done, and going back feels like doubting yourself. The paper's complete, the pen's down, and every instinct says hand it in and stop thinking about it.
There's also a real cognitive reason: re-reading your own work is nearly useless. Your brain reconstructs what you meant to write rather than what's on the page. So students who do check often do it in the least effective way — a quick skim that confirms what they already believe — and conclude that checking doesn't help.
It does help. It just has to be done differently.
Check against the question, not against your answer. Read the question again — properly, not a glance — and ask whether what you wrote answers that. Not whether the maths is right. Whether it answers the ask. This catches the most expensive class of error, where perfect work went to the wrong question.
Check the answer for reasonableness. Is a negative length plausible? Is a probability above 1? Did the population of a town come out at 4.6 people? Alberta's markers note that students often struggle to connect their answer back to the context of the question and should be checking it's reasonable within that context. Ten seconds, and it catches errors nothing else catches.
Check the instructions you were given. Rounding is the classic. If a question specifies to the nearest tenth, that's a mark. If no rounding instruction appears in a math question, an exact value is expected. Units, labels, "show all work" — those are all marks that don't require knowing anything.
Check the ones you rushed. Not all twenty. You know which three you hurried. Go there first. If time runs out, you've spent it in the right place.
"Check your work" is advice every student has heard ten thousand times and it changes nothing, because it's framed as a character trait — be careful, be thorough — rather than as a task.
Make it a task with a start time.
That last one matters most. A student who checks homework every night for a semester doesn't have to remember to check on exam day. They'd have to remember not to.
This is the part worth taking seriously as a parent.
Reading a question carefully and checking work before handing it in look like the same thing — "being careful" — and they're not. They're two different habits, and a student can be excellent at one and never do the other.
Which is why "just be more careful" fails as advice. It names a mood, not an action. The student agrees, means it, and then does exactly what they did last time, because nothing specific changed.
Specific changes: last ten minutes, four checks, every night on homework. That's actionable. "Be careful" isn't.
Look at the last returned test with your kid and ask one thing: "How many of these could you have caught?"
They'll know. They almost always know. And the number is usually big enough to make the point without anyone having to make a speech about it.
Verification is one of the four steps Joseki works on separately — because knowing the material and checking the work are different jobs.