Interleaving vs. blocked practice
Thirty quadratic equations in a row feels like mastery. Thirty mixed problems feels like drowning. The second one is what the exam is, and it's what you should be practising.
The short version
- Blocked practice — all of one type, then all of the next — makes you good at executing a method you were told to use.
- Exams don't tell you which method to use. Choosing it is a separate skill, and only mixed practice trains it.
- Interleaving produces worse performance during practice and better performance on a later test. Both effects are large.
- It matters most where problems look similar but need different approaches — which is most of maths and physics.
Textbook exercises are almost always blocked. Chapter 7 is on the sine rule; the exercises at the end of chapter 7 are twenty sine rule problems. You do them, you get nearly all of them right, and you conclude that you can do the sine rule. Then the exam presents a triangle with no chapter heading attached and you find you have no idea which rule it wants.
That gap is the thing interleaving addresses, and it's worth being precise about what's missing. You did learn something in chapter 7 — the execution. What you never practised was the identification: given an unlabelled problem, work out what kind of problem it is. In a blocked exercise set, that step is done for you before you start, by the heading.
The evidence
The clearest demonstrations come from maths. In a study by Rohrer and Taylor, students learned to compute the volumes of four solids, half with blocked practice and half with the same problems shuffled. During practice, the blocked group did far better — roughly 89% against 60%. On a test a week later, the ordering flipped hard: about 20% for the blocked group against about 63% for the interleaved one. Same problems, same total time, radically different outcome.
The effect is not confined to maths. Kornell and Bjork had people learn to identify painters' styles, either one artist at a time or shuffled. Interleaved learners were better at classifying new, unseen paintings — and, revealingly, most of them reported afterwards that they thought blocking had helped them more. The subjective and objective results point in opposite directions, which is the signature of a fluency illusion.
Why blocking flatters you
In a block, you set up the method once and then reuse it. Problem two doesn't require you to decide anything; it requires you to repeat. That's why accuracy is high and why almost nothing is being learned by the time you reach problem fifteen.
What interleaving actually trains
Discrimination. Meeting a quadratic right after a simultaneous equation forces you to notice what makes them different. Meeting twenty quadratics in a row gives you no contrast to notice anything against. Learning what a category isn't requires seeing things that aren't it, nearby.
Retrieval of the method itself. In a block, the method is sitting in working memory from the last question. Mixed up, you have to go and get it — which is retrieval practice applied to procedures rather than facts.
Exam conditions. A paper is interleaved by construction. If every hour of practice you've done was blocked, the exam is the first interleaved test you've ever sat, and you'll find out how that goes at the worst possible moment.
How to do it
Block first, then mix
The sequencing matters and is often skipped. When a method is brand new, some blocked practice is the fastest way to get the mechanics working — you can't discriminate between methods you can't yet execute. The rule of thumb: block until you can do the procedure without looking anything up, which is usually about five problems, then stop and start mixing. The mistake isn't doing blocked practice; it's doing twenty of them.
Build a mixed problem set
The practical version: as you finish each topic, pull two or three problems from it into a running "mixed set" for the subject. Once a week, work through a randomised slice of that set. Because it spans everything you've covered, it interleaves and spaces at the same time — the two techniques stack neatly.
Shuffle the deck
For flashcards, the equivalent is simply not reviewing topic by topic. A deck sorted by chapter lets you answer from context — you know the answer is about photosynthesis because everything for the last five minutes was. Shuffled, each card has to stand on its own.
Take the labels off
A strong drill, especially for maths and science: collect twenty problems from across the syllabus and, without solving any of them, write down only which method each one needs. Two minutes of that trains the identification step in isolation, and it will show you very quickly which pairs of topics you're confusing.
Mixing is a filing problem more than a technique. In YourSubjects, a subject's practice questions live in one bank rather than in per-topic piles, and papers can pull questions from across the whole syllabus with topics tagged — so a mixed set is what you get by default, and the results tell you which topics are actually costing you marks.
Open the appThe limits
It needs surface similarity to pay off. Interleaving shines when problems look alike but need different treatments. Mixing French vocabulary with organic chemistry doesn't create any useful discrimination, because nobody was going to confuse them. Mix within a subject, between things that could plausibly be mistaken for each other.
Don't mix during first learning. Interleaving a method you haven't got working yet adds load without adding discrimination. First understand, then execute, then mix.
It will feel bad, and your accuracy will drop. This is expected and it is not a sign to stop. Practice accuracy is the measure that interleaving deliberately sacrifices. If you use it to judge whether the technique is working you will always conclude it isn't.
Some things are genuinely serial. A proof, a long derivation, a piece of music — anything where step four depends on step three — needs to be practised whole. Interleave between such items, not inside them.
If you take one thing
After you finish a topic's exercises, go back and do five problems from three earlier topics, in random order, without looking at which topic they came from. It takes fifteen minutes and it is the closest thing to a free upgrade in this whole set of guides.
Where this comes from
- Rohrer, D., & Taylor, K. (2007). The shuffling of mathematics problems improves learning. Instructional Science, 35.
- Kornell, N., & Bjork, R. A. (2008). Learning concepts and categories: is spacing the "enemy of induction"? Psychological Science, 19(6).
- Rohrer, D., Dedrick, R. F., & Stershic, S. (2015). Interleaved practice improves mathematics learning. Journal of Educational Psychology, 107(3).
- Dunlosky, J., et al. (2013). Improving students' learning with effective learning techniques. Psychological Science in the Public Interest, 14(1). Rates interleaved practice "moderate utility" — promising, with narrower evidence than testing or spacing.