Textbook exercises usually come in chapters: twenty integration-by-parts problems, then twenty substitutions, then twenty partial fractions. Finish a block and you feel fluent. But exam papers never print “use integration by parts” above a question. The hardest thing you do in the exam hall, recognising which kind of problem this is, is the one thing you never practised.
Interleaving means mixing several problem types or topics within one session, instead of finishing one type before moving on (blocking). Studies in maths, motor skills and visual category learning often find that interleaving produces better results on delayed tests, even though learners make more mistakes during practice and rate themselves as doing worse.
Blocking hides the most important step
An exam problem asks for two things: choosing a method and executing it. Work twenty same-type problems in a row and the chapter heading has already made the choice for you. You only practise execution. When a mixed paper arrives, the choosing step is exposed as a skill you never trained.
This is why so many students say, “I can do the homework but not the exam.” Their execution is fine; their discrimination is not.
Interleaving trains discrimination
When type A and type B sit side by side, you are forced to compare: how is this like A, how is it unlike B? That comparison is what builds the cues you use to recognise problems. In language learning, mixing the simple past and present perfect in one exercise forces attention to time markers that disappear when all twenty sentences use the same tense.
| Blocked practice | Interleaved practice | |
|---|---|---|
| How it feels | Smooth, confident | Clumsy, more errors |
| Skill trained | Running one procedure | Choosing the right procedure, then running it |
| Immediate test | Often good | Can be worse |
| Delayed, mixed test | Often worse | Often better |
| Best suited to | First encounter with a technique | Consolidation and revision |
When blocking is still the right choice
Interleaving does not always win. With a brand-new technique, you need a few consecutive examples to understand the procedure; mixing from the start just creates noise. A practical rule:
- First session with a new type: three to five problems of that type in a row, with worked examples.
- From the second session on: the new type joins older ones in a shuffled set.
- Topics that are too far apart, such as integrals and German grammar, give no discrimination benefit. Interleaving helps when the types are easy to confuse.
Building your own mixed sets from a chapter-based textbook
- List the problem types in the exam scope, say six types in a calculus module.
- Pick three or four problems per type from the end of each chapter, where problems are usually harder than the worked examples.
- Copy each onto a separate card or slip with no chapter name; note the page number on the back.
- Shuffle and draw eight to ten for a session.
- Before solving, write one line: “Type: … Because: …”. This line is the most valuable part of the session.
- Mark both parts: did you pick the right type, and did you solve it correctly? Track wrong-type errors separately, because they are the ones that need the most work.
Past papers are already interleaved. All you need to add is the “Type / Because” line before each answer.
Beyond maths
Physics and engineering
Mix conservation of energy, conservation of momentum and Newton's laws problems. The question is always: what is conserved here, and why?
Accounting
Mix transactions across accounts instead of doing all inventory entries before moving on to receivables. The real accounting skill is recording a transaction correctly when nobody tells you what kind it is.
Medicine and pharmacy
Mix cases with overlapping symptoms. Telling apart two conditions that look alike is exactly what clinicians do every day.
Languages
Mix easily confused structures in one exercise: prepositions of time, articles, neighbouring tenses.
Common mistakes
- Shuffling but still seeing the chapter title. If you know where a problem came from, recognition has been done for you.
- Losing heart as the error rate rises. This is predicted. Judge the method by a test a few days later, not by how the session felt.
- Mixing too many types at once. Three to five related types per session is about right; ten becomes random noise.
- Skipping the “why this type” step. Without it, interleaving is just doing problems in a messy order.
Combining it with spacing and retrieval
The three techniques reinforce each other. Retrieval makes you produce the answer yourself, spacing times those retrievals well, and interleaving makes you distinguish between things that are easy to confuse. A strong revision session usually contains all three: a mixed set (interleaving), no peeking at solutions (retrieval), including topics from last month (spacing).
For your next revision session, take eight problems from your three most recent chapters, copy them onto loose slips without chapter names, shuffle, and write the “Type / Because” line before each one. Count your wrong-type choices separately. That is the number you want to watch fall.
Câu hỏi thường gặp
What is interleaved practice?
It is mixing several related problem types or topics within one study session instead of completing one type before starting the next. It trains you to recognise which method a problem requires.
Why does interleaving feel harder than blocked practice?
Because you must decide the method for every problem, whereas in blocked practice the chapter heading decides for you. More mistakes during practice are normal and often come with better delayed test results.
Should I interleave from the very beginning?
No. For a completely new technique, work a few consecutive examples first to understand the procedure. Start mixing it with older types from the next session.
Does interleaving work for theory subjects?
Yes, when concepts are easy to confuse, such as similar theories or conditions with overlapping symptoms. Mixing them forces you to pin down what makes each one different.
How many problem types should I mix in one session?
Around three to five related types works well. Mixing many unrelated types adds confusion without the benefit of comparison.
