Why kids 'forget' maths overnight (and what's actually happening)

Every primary teacher has had this experience. You teach long division on Monday. The class gets it. They can do it. You finish the lesson feeling great.

Tuesday, you start with a quick recap. Half the class looks at you blankly. The other half remember the procedure but get muddled on the details. By Friday, when you give a written assessment, the success rate has crashed.

It's tempting to call this 'forgetting.' But the cognitive science term is more accurate, and more useful: the gap between PERFORMANCE and LEARNING.

Performance is not learning

When a child gets a question right at the end of a lesson, what you've measured is their performance in that moment. It does not necessarily mean they have learned the underlying concept.

Performance and learning are different things. Performance is what a child can do RIGHT NOW. Learning is what a child can still do TWO WEEKS LATER, in a different context, possibly under stress.

If a child can do long division at the end of the lesson but not on Friday, they didn't learn long division. They performed it. Those are different events.

This isn't a moral failure on the child's part — it's how memory works. Information that's just been taught sits in working memory and short-term memory. Most of it never makes it into long-term memory unless something specific happens to push it there. That something is called retrieval, and it has to be deliberate.

What actually pushes information to long-term memory

Three things, all of them well-evidenced by research:

**Retrieval (active recall).** Trying to remember something STRENGTHENS the memory of that thing. This sounds obvious but contradicts how most people study (re-reading notes, copying out tables). The act of pulling information OUT — answering a question without help — is what consolidates the memory. Looking AT the answer is far weaker.

**Spacing.** Memory consolidates when information is revisited at increasing intervals. Five minutes after the lesson, then the next day, then four days later, then a week, then a fortnight. This is the 'spaced retrieval' that the cognitive science literature keeps banging on about. It's annoying because it doesn't feel like teaching — but it's what actually moves things to long-term memory.

**Interleaving.** Mixing different topics together is harder in the moment than blocking them — but it produces stronger long-term recall. A child who does 30 long-division questions in a row doesn't really learn long division; they learn how to repeat one procedure. The same child doing 10 long-division mixed with 10 fractions and 10 area questions has a harder lesson but learns more durably.

These three are the boring engine of learning. They don't make for an exciting lesson observation. But they are the difference between performance (which fades) and learning (which sticks).

Why this looks like forgetting

When a child can do something on Monday and not Tuesday, what's happened isn't that the memory has been deleted. It's that the memory was never strong in the first place. Monday's success was a moment of high performance — the information was active in working memory, the teacher's scaffolding was visible, the previous example was minutes ago. By Tuesday, all of that had decayed, and the underlying memory wasn't strong enough to stand on its own.

This explains an annoying classroom phenomenon. The child who 'gets it' in the lesson but 'can't do it' in the homework hasn't suddenly become stupid. They just had supports during the lesson — your worked example on the board, the previous question's answer still in their head, the gentle scaffolding of your teaching voice — that aren't there at home. Strip away the supports, and the underlying memory wasn't enough.

The reframe for planning

If you accept that performance ≠ learning, your planning changes.

A lesson that ends with high success but no follow-up has produced no learning. A lesson that ends with moderate success but with planned spaced retrieval over the next three weeks has produced learning. The first feels better. The second works better.

This means your weekly maths plan should look something like: new content, then a planned schedule of retrieval. Not 'we did fractions on Tuesday so we tick that off and move on.' More like 'we did fractions on Tuesday — they get a fluency starter on fractions on Friday, again the following Wednesday, again two weeks later, then again at the end of the half term.'

Most schools' maths plans don't do this systematically. The curriculum plan usually says 'fractions: Week 3.' It doesn't say 'fractions retrieval: weeks 4, 6, 8, 10.' That's the missing piece.

What 'retrieval practice' looks like in primary

It doesn't have to be fancy.

- A 90-second 'do it now' starter that includes 2 questions from last week, 2 from last term, and 1 from today's topic. - A weekly 'flashback Friday' with 6 mixed questions covering everything from the term so far. - A half-termly low-stakes mixed assessment that covers everything covered, not just the most recent topic. - Short multiple-choice quizzes (3 questions, in your back pocket, fillable from any topic at any time).

The key features: low stakes (no big scores recorded), mixed topics (not just the current unit), self-marked or quick-marked (you don't want to drown in marking), and frequent (multiple times per week).

This is essentially what your daily-5 times-tables practice is doing — applied to all of maths, not just multiplication.

Why we don't do this

If retrieval, spacing and interleaving are so well-evidenced, why don't more teachers do them? Two reasons.

One: they don't feel like teaching. They feel like testing. And testing has a bad name in primary, so we shy away from doing it more. But the evidence is clear: low-stakes retrieval is one of the most powerful TEACHING tools we have. Calling it 'practice' or 'flashback' rather than 'test' helps.

Two: they expose problems. A retrieval starter on last term's fractions might reveal that 60% of the class haven't really retained it. That feels like a failure of the original teaching. It is — but it's a failure that ALREADY HAPPENED. The retrieval starter is just showing it. Better to find out now than to discover it in a SATs paper. The retrieval is the reset chance.

The takeaway

When children 'forget' maths overnight, they didn't forget. They never properly learned it in the first place — the lesson produced performance, not learning. The fix is boring and works: deliberately revisit, deliberately mix up, deliberately make children retrieve from memory rather than copy from a model. Do this every week, in small bursts, on every topic.

It's a bigger shift than it sounds. But it's the difference between teaching that fades and teaching that sticks.