Shuffled, Not Sorted: Interleaving
Mixing related problem types in one practice session forces you to choose the right method for each problem — and that practiced choice, not extra time, is what roughly doubles delayed test performance. · 11 min
Open almost any textbook and look at the practice set. Twenty problems, every one solved by the method the chapter just taught. You never have to ask which tool a problem needs — the page has already answered that. Researchers call this arrangement blocked practice, and it has a rival: interleaved practice, in which several related problem types share one shuffled session. The two arrangements contain the same problems and take the same time. They differ only in order. That difference sounds cosmetic. Measured a week later, it is anything but — and the arrangement that wins is the one that feels worse while you are in it.
Guess before you learn
Rohrer and Taylor gave two groups the same geometry problems — four related types about volumes of solids. One group practiced them blocked, one type at a time; during practice they averaged 89% correct. The other group practiced the identical problems shuffled together, averaging just 60%. One week later, both groups took the same mixed test. Who scored higher?
On the week-later test the blocked group managed 20%; the shuffled group scored 63% — more than three times higher. If you backed the blocked group, you agree with most students, and with how the practice session felt from the inside. This folio is about why that feeling misleads.
9–12
3–5
A worksheet with all multiplication is easy to race through — after the first problem, you already know what every next one wants. Mix multiplication with division and something changes: each problem now asks two things. First, which kind is this? Then, do it. Real tests, and real life, always ask both.
Mixed practice feels slower and messier. It is. That extra work of deciding is exactly what makes the memory stick.
6–8
Blocked practice groups problems by type: all of one kind, then all of the next. Interleaved practice shuffles related types into one session. In Rohrer and Taylor's 2007 experiment, blocked practicers hit 89% during practice but only 20% on a mixed test a week later; interleaved practicers managed 60% during practice and 63% at the test. Same problems, opposite outcomes.
The reason: a blocked page never asks you to identify a problem. Interleaving makes every item a two-step act — classify, then solve — and classifying is the skill the final test secretly grades.
9–12
The result held at scale. Rohrer, Dedrick, Hartwig, and Cheung (2020) randomized 54 seventh-grade mathematics classes to blocked or interleaved worksheets for months, then gave a surprise test one month after the last session: interleaved classes scored 61% against the blocked classes' 38% — an effect size of d ≈ 0.83, enormous by the standards of education research.
The proposed mechanism is discrimination. Mixed problems force you to notice what distinguishes each type and to pair each type with its method. Blocked practice lets you execute a method twenty times without once choosing it — fluent performance, thin learning.
K–2
Here are ten cards. Five say plus, five say take away. If all the plus cards come first, you can stop reading the signs. Your hand already knows what to do.
Now shuffle the cards. Every card makes you look: plus, or take away? That looking is the practice that matters.
Undergrad
Two accounts survive scrutiny. The discriminative-contrast account holds that interleaving juxtaposes confusable categories, sharpening the boundaries between them; Kornell and Bjork (2008) found interleaved paintings taught artists' styles better than blocking did — and most participants, shown their own results, still voted for blocking. The second account notes that interleaving inherently spaces each type; studies that equate spacing still find an interleaving advantage, so discrimination carries independent weight.
Metacognition reliably inverts here: practice accuracy is higher under blocking, so judgments of learning follow it — a fluency signal misread as knowledge, the same error folio 4 dissected.
Postgrad
Brunmair and Richter's 2019 meta-analysis (g ≈ 0.42 overall) shows sharp moderation: robust benefits for confusable, similar materials — mathematics problem types, artists' styles, bird species — but null to negative effects for unrelated materials such as foreign-vocabulary lists. Carvalho and Goldstone's sequential-attention account predicts the pattern: interleaving directs attention to between-category differences, blocking to within-category commonalities.
The design rule follows. Interleave what learners must discriminate between; block briefly when the task is abstracting a shared structure, or when the learner cannot yet execute any single type. Interleaving is a discrimination tool — it earns nothing where there is nothing to confuse.
interleaving
Arranging practice so that related, confusable problem types share one shuffled session — forcing each item to be classified before it is solved.
Why is this true?
Why can a blocked practice set never exercise the choice of method?
Because the block answers the question in advance: every problem in the section uses the section's method. Choice only exists where the next problem's type is uncertain — which is exactly what shuffling restores.
The result scaled. Rohrer, Dedrick, Hartwig, and Cheung (2020) ran a randomized trial across 54 seventh-grade classrooms over several months. On a surprise test one month after the final worksheet, interleaved classes scored 61%; blocked classes scored 38%. The standardized difference, d ≈ 0.83, is among the largest ever measured for a classroom adjustment that costs nothing: the problems were identical, and only their order changed.
Interleaving has a right moment and a wrong one. On first contact with a brand-new problem type, a short blocked run — a worked example, then two or three problems — is fair: you cannot choose between methods you cannot yet execute. Once each type runs on its own, shuffle them together and keep them shuffled. Interleave related, confusable types — the ones you might mistake for each other — not arbitrary subjects. Mixing calculus with French vocabulary sharpens nothing, because nothing is being discriminated. And expect the session to feel worse: slower, less certain, more errors. That drop in practice polish is the price of the test-day gain — folio 11 gives the trade its proper name.
One detail deserves notice before you leave: a shuffled session does not just mix types — it also spaces them, since each type returns only after a gap. Interleaving carries folio 7 inside it. The next folio assembles everything this unit has built — retrieval, spacing, a criterion for success — into the strongest study protocol on record.
Practice — new ink and old, interleaved
1.Tonight's set is ten problems, all on the new topic. In one sentence, rework it into interleaved practice.
2.From folio 7: the best review gap is roughly 10–20% of how long you need to remember. For a test 30 days away, the middle of that range is a gap of about how many days?
3.From memory, folio 7's headline finding: what happens when the same total study hours are spread across days instead of massed into one sitting?
Distributed sessions produce far more durable memory than one massed sitting of equal length — cramming holds up tomorrow and has largely evaporated within weeks.
How close were you? Grade yourself honestly — it sets your review date.
4.Rereading a chapter mostly exercises which act of memory?
5.From folio 8: a card passes at an interval of 1 day, then 6 days, with an ease factor of 2.5. About how many days is the next interval?
6.A week after studying, the repeated-recall group beat the rereaders. What is the best explanation?
7.A meta-analysis reports retrieval practice at g ≈ 0.61. What does that number mean?
8.Turn this highlighted sentence into a self-test question: 'The hippocampus replays the day's learning to the cortex during slow-wave sleep.'
9.Which homework plan is interleaved?
10.Match each activity to what it actually practices.
11.Order the life of a reviewed memory, first to last.
- Learn the list to full strength
- The curve falls steeply through the first day
- A review restores full strength
- The new curve falls more slowly than the first
12.From folio 5: solving mixed problems beats rereading solved examples for the same reason that —