Interleaving: mixing problem types while you practise
How learning works · Published · Evidence: Moderate · About this guide
Interleaving means mixing different kinds of problems in one practice session instead of practising one kind at a time. In a 2019 meta-analysis it beat blocked practice by a moderate amount on average, but the benefit depended on the material: largest for paintings, small on average for maths, no reliable benefit for expository texts, and blocks did better for words [1]. It also tends to feel worse while you do it [2].
Most practice comes in blocks. A maths chapter teaches one method and follows it with a page of problems that all need that method; a language course drills one rule before moving to the next. Interleaving means mixing the kinds of items instead, so that the next one is not obviously the same kind as the last. A growing body of research supports it for some material. It is also often oversold: it does not help with everything, and it tends to feel worse while you do it.
What does interleaving mean?
Picture three kinds of problem, A, B and C. Blocked practice runs AAAA BBBB CCCC. Interleaved practice runs something like ABCB CACB ACBA: the same problems, arranged so that two neighbours rarely need the same method. Nothing is added. Only the order changes.
In maths, a blocked assignment after a lesson on, say, the volume of a cone is a page of cone problems. An interleaved one mixes them with problems from earlier lessons, so each problem first asks which formula it needs, a step the blocked page takes for you [3][4].
Interleaving is related to spacing but is not the same thing. Spacing is about when you come back to the same material. Interleaving is about what sits next to what within a session. The two overlap, because once the kinds are mixed, the problems of each kind end up spread apart. That overlap is one reason researchers still debate why interleaving works [4][1].
What did the key experiments find?
In a small experiment, 18 college students learned to work out the volume of four unfamiliar solid shapes [3]. One group practised in blocks: a tutorial on one shape, then four problems on that shape, then the next shape. The other group read all four tutorials first and then worked the same problems in a mixed order. During practice the blocked group did much better, 89% correct against 60%. On a test a week later the result reversed: 63% for the mixed group, 20% for the blocked group. Almost every error on the test was a wrong choice of formula. The blocked students could still do the calculation. What they had not practised was recognising which formula a problem needed.
A study with paintings made a similar point [2]. Students looked at landscape paintings by twelve artists chosen to be unfamiliar to them, six paintings by each. Some artists' paintings were shown one after another; others were shuffled in among other artists' work. After a short filler task, the students saw paintings they had not seen before and had to name the artist. Shuffled study did better: in the first round of the test, about 61% correct against 35%.
Interleaving has also been tested in real classrooms. A preregistered randomised trial ran in 54 seventh-grade maths classes, with 787 students [4]. For about four months, classes periodically did practice worksheets in class that were either mostly mixed or mostly blocked, and then both groups did the same mixed review. On an unannounced test about a month later, the mixed group scored 61% and the blocked group 38%.
Why would mixing help?
The leading explanation is discrimination: learning to tell kinds apart. When every problem on a page is the same kind, the page has already told you which method to use, so you do not practise choosing one. A cumulative exam usually does not label its questions. Mixed practice makes you ask "what kind of problem is this?" at each problem, and it puts different kinds side by side, so the differences between them stand out.
One study, with two experiments, tested this with paintings [5]. Putting more time between one artist's paintings, while still showing them artist by artist, did not help. Showing paintings by different artists side by side on the screen helped as much as shuffling them. That suggests the benefit came from comparing the kinds, not from the delay itself. A later meta-analysis found results consistent with this explanation, although only a few studies had tested it [1].
Is interleaving better than blocked practice?
On average, by a moderate amount, but not for every kind of material. This is where popular advice runs furthest ahead of the evidence. A 2019 meta-analysis pooled 59 studies with 238 results [1]. On average, interleaving beat blocking by a moderate amount (g = 0.42, possibly somewhat smaller once studies that may have gone unpublished are allowed for), but the size of the benefit depended heavily on the material:
- Paintings showed the biggest average benefit. Other pictures, such as photographs of birds or drawings of invented creatures, showed a smaller but still positive one.
- Maths problems showed a small average benefit, with large differences between studies.
- Expository texts, meaning reading passages that explain something, showed no reliable benefit.
- Words, such as vocabulary or word categories, did better in blocks than mixed. Only a handful of studies used words, so the authors asked readers to treat that result with caution.
Most of these studies mixed kinds of problems from one topic. Switching between unrelated subjects is a different practice, and these results do not show that it helps.
Interleaving by kind of material: average effects (Brunmair & Richter 2019)
The average effect of interleaved over blocked study, as Hedges' g, by kind of material, as Brunmair and Richter's meta-analysis (2019) prints it in its published summary: largest for paintings, small for mathematical tasks, and in favour of blocking for words, from few studies. Expository texts are not drawn: the summary gives no value for them, only that the effect was not significant. Point estimates without their confidence intervals; averages across the studies pooled, not a prediction for any one learner.
Data: Brunmair & Richter (2019), Psychological Bulletin 145(11), abstract. [1]
Data table
| Material | Hedges' g |
|---|---|
| All kinds of material | 0.42 |
| Paintings | 0.67 |
| Mathematical tasks | 0.34 |
| Words | -0.39 |
- Point estimates as the published article's abstract prints them; for expository texts and tastes it reports only that the overall effect was not significant.
- The authors advise caution with interleaving, especially for expository texts and words.
Mixing helped more when the categories were similar to one another, that is, easy to confuse. A refinement of the discrimination theory predicts the reverse for kinds that are easy to tell apart and whose own examples differ a lot from one another: there, blocks may work better, because seeing one kind's examples together brings out what they have in common [1]. An earlier review of ten study techniques rated interleaving as moderately useful. It called the technique promising but said the evidence was still too limited for a higher rating, and in the two studies it reviewed, it found no sign that mixing vocabulary practice improves retention [6].
Even for maths the picture is not settled. The meta-analysis found a small average benefit, and its authors were reluctant to recommend interleaving for maths. The classroom trial found a large one [1][4]. Both can be true at once: the effect varies a great deal with how the mixing is done and what the test asks.
Why does mixing feel worse if it helps?
Because during practice it really is harder. Mixed practice brings more mistakes and slower progress. In the volume experiment, the group that later did far better on the test had far lower practice scores [3]. In the painting study, 78% of participants did better with the shuffled artists, yet 78% said blocked study had been as good or better [2]. Blocks feel smooth, and smoothness is easy to mistake for learning. The article on illusions of knowing looks at this trap in more detail.
Mixing also takes longer. In the classroom trial, every teacher reported that the mixed assignments took students more time. Time on task was not measured, but the researchers note that, to the extent the mixed work did take longer, the benefit per minute of practice would have been smaller [4].
How could you try it?
These are reasonable ways to apply the findings, not guarantees:
- Mix kinds you could confuse. The experiments described here mixed related categories within one subject: formulas, artists, problem types. Switching between unrelated subjects is a different thing, and it is not what they tested.
- Meet each kind first. In most of the maths studies, students had already practised each method a little before the mixed practice began. The researchers do not suggest giving up blocked practice entirely [4].
- Check your answers. In the maths studies, students saw the correct solutions after working the problems, and in the classroom trial they were also asked to correct their errors. The trial's authors think feedback may be a necessary part of the effect [4].
- Before a cumulative exam in a subject where you must choose a method, such as maths, practise with a mixed set. Questions drawn from several chapters ask the same question a cumulative exam does, "which method is this?", in a way that another block from the last chapter cannot [4].
- Judge by a later test, not by how the practice felt.
- For vocabulary and reading, lean on other methods. Self-testing and spreading study over time have much broader support across materials [6].
What the research does not say
- It does not say mixing beats blocking for all material. For words, blocking did better; for expository texts, there was no reliable difference [1].
- It does not say to switch between unrelated subjects. The studies in this article mixed related kinds of items within one subject. One study that alternated two texts on different topics, the immune system and female reproduction, found no benefit from mixing them [1].
- It does not give an ideal mix. Studies compared mostly mixed with mostly blocked practice. No particular ratio or order has been shown to be the right one.
- It does not settle why it works. Telling kinds apart, spacing, and the effort of recalling a method may all contribute, and in maths practice they come bundled together [4].
- It does not cover every learner or setting. Most studies used university students. School-age learners and everyday study material, such as textbook chapters, have been studied far less [1].
- It does not predict your result. The percentages in this article are group averages from particular studies. They are not a forecast of how much mixing would change your own scores.
Related articles
-
Spaced repetition and spaced practice
The spacing effect: why spreading study over days helps memory last, how far apart reviews should be, and whether ever-longer intervals beat equal gaps.
-
Retrieval practice (active recall)
Retrieval practice, or active recall: why recalling and then checking the answer tends to beat rereading, how large the testing effect is, and its limits.
-
Feedback and learning from mistakes
Why checking answers after a quiz matters, when a wrong guess helps you learn, and why confident mistakes are often easiest to fix, though they can return.
-
The illusion of competence: when studying feels better than it works
Familiar is not the same as remembered. Why rereading and highlighting feel effective but rate low in research, and how to check what you really know.
-
Exam study plan: planning your reviews before an exam
Start from the exam date: how far apart to space reviews, what to do in each one, and an example four-week revision plan, with the limits of the evidence.
-
Why do we forget what we study?
Why you forget what you study: new learning can disturb fresh memories, similar ones compete, cues change. Forgotten often means out of reach, not gone.
Sources
- [1] Brunmair, M., & Richter, T. (2019). Similarity matters: A meta-analysis of interleaved learning and its moderators. Psychological Bulletin, 145(11), 1029–1052. meta-analysis · doi.org/10.1037/bul0000209 ↩ Back to the text
- [2] Kornell, N., & Bjork, R. A. (2008). Learning concepts and categories: Is spacing the "enemy of induction"? Psychological Science, 19(6), 585–592. experiment · doi.org/10.1111/j.1467-9280.2008.02127.x ↩ Back to the text
- [3] Rohrer, D., & Taylor, K. (2007). The shuffling of mathematics problems improves learning. Instructional Science, 35(6), 481–498. experiment · doi.org/10.1007/s11251-007-9015-8 ↩ Back to the text
- [4] Rohrer, D., Dedrick, R. F., Hartwig, M. K., & Cheung, C.-N. (2020). A randomized controlled trial of interleaved mathematics practice. Journal of Educational Psychology, 112(1), 40–52. experiment · doi.org/10.1037/edu0000367 ↩ Back to the text
- [5] Kang, S. H. K., & Pashler, H. (2012). Learning painting styles: Spacing is advantageous when it promotes discriminative contrast. Applied Cognitive Psychology, 26(1), 97–103. experiment · doi.org/10.1002/acp.1801 ↩ Back to the text
- [6] Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students' learning with effective learning techniques: Promising directions from cognitive and educational psychology. Psychological Science in the Public Interest, 14(1), 4–58. review · doi.org/10.1177/1529100612453266 ↩ Back to the text