Interleaving¶

Principle · All principles
Evidence · 42 claims (18 for, 22 mixed, 2 against) · 27 studies (11 causal, 6 review, 4 quant-synthesis, 4 theoretical, 1 associational, 1 qualitative), q1–q4 · 12 of 27 report an effect size · 39 claims rest on one study

Conditional relationship¶

This page owns the decision of whether practice mixes kinds or keeps them apart. When a learner must later decide which category an example belongs to, or which procedure a problem calls for, and the candidate categories or problem types resemble one another, practising with examples or problems of different kinds mixed together (interleaved) is expected to produce better later classification and problem choice (near-transfer, delayed-retention) than practising one kind at a time (blocked), even though practice itself usually goes worse when mixed. Interleaving gives the learner repeated chances to compare kinds and to choose, which blocking removes: in a block, the kind of the next item is already known.

The relationship is conditional on four things, each from the evidence below:

How alike the categories are. The benefit grows with similarity between categories and shrinks with similarity within a category. Where the kinds are easy to tell apart, there is little for mixing to teach.
The material. Across 59 studies the benefit was largest for paintings and naturalistic photographs, smaller for mathematical tasks, absent for expository texts and reversed for word lists, where blocking did better.
The goal. The evidence is about discriminating kinds and choosing among them. It is not evidence that mixing helps a learner master a single new procedure, learn one text, or memorise words.
What the learner sees during practice. Mixed practice feels harder and looks worse while it happens; learners in one study judged blocking more effective after interleaving had served them better. A design judged on practice performance or on learners' preference will choose blocking.

The meta-analysis found no reliable moderation by retention interval, by retention against transfer test, or by whether examples were shown together or one after another; finding no moderation is not evidence that these never matter. It found larger effects in samples of university students than in other samples, and a second systematic review found no high-school study among its 26. The one school study in the wiki is a classroom experiment with 126 seventh-graders in mathematics, which found the benefit at 1 day (d = 0.42) and larger at 30 days (d = 0.79). Carrying the size of the effect to other ages, other subjects or adults outside education is untested in the wiki. In the few studies a third meta-analysis could split by ability, interleaved practice had negative effects for high-ability students, so who gains is also not settled.

How this sits beside its neighbours: Spaced Learning holds when practice of the same thing recurs; interleaving changes what sits next to what, and in most mixed schedules it also spaces each kind. A systematic review reports that the benefit survives when spacing is held constant, and one set of experiments found that what matters is that different kinds sit next to each other: putting unrelated material between items removed the advantage over blocking. Sequencing holds the wider order of a unit and already carries the blocked-then-mixed choice as one row; Retrieval Practice holds whether practice asks for recall.

Default design, while the dose is untested¶

The meta-analysis establishes the direction and its moderators, not a dose. This default is a labelled proposal: the order of steps follows the evidence, and the numbers (how many blocked items, how much mixing, how many sessions) are untested choices.

Check the condition first. List the categories or problem types the learner will have to tell apart. Ask: would a competent person confuse them? If the kinds are not confusable, keep practice blocked and stop here: there is little evidence that mixing dissimilar kinds helps.
Introduce each kind on its own with two or three worked examples or items, so each is recognisable before it has to be distinguished (the blocked start is a design choice, see Evidence and qualifications).
Mix from the first practice set after each kind has been met: in a set of 12 items covering three kinds, no two consecutive items of the same kind more than occasionally, in an order the learner cannot predict. Keep the mixed items back to back: in one set of experiments, putting unrelated material between items of different kinds removed the advantage, while spreading each kind further apart with other kinds in between helped.
Ask for the choice explicitly: before solving or labelling, the learner names the kind or the procedure, and gets feedback on the choice as well as on the answer. Do not leave a definition of each kind on screen during the mixed set: in one experiment summarised by a review, a definition on screen removed the benefit of interleaving.
Across sessions, keep mixing earlier kinds into later practice (week 1 kinds A–C mixed; week 2 adds D and keeps A–C in the mix), so the mixed sets also space each kind. The one classroom test in the wiki ran mixed assignments over about three months. Plan at least one delayed, mixed check (a week or more later; that classroom study found the advantage larger at 30 days than at 1) before judging the design.
Judge it on the delayed, mixed check, not on practice accuracy and not on learners' ratings of which practice felt better.

Fitting the design to a situation¶

Three facts change the decision most. Ask for them when the brief omits them: are the kinds the learner must tell apart actually confusable? (if not, interleaving has little to work with); what is the material? (visual categories, mathematical problem types, texts, word lists); and what will the learner have to do later: choose among kinds, or reproduce one?

If the brief says…	Then change…	Basis
The goal is to recognise or classify visual categories (artists' styles, species, skin lesions, rock types)	Interleave from the first practice set after one or two examples of each; this is where the evidence is strongest	Interleaving shows positive effects for all visual stimulus types, largest for paintings [+M] (g = 0.67 for paintings, 0.35 for naturalistic photographs)
The goal is choosing the right procedure among mathematics problem types (`procedure`, several types taught)	Mix types in practice and homework once each has been taught; ask the learner to name the type before solving; expect smaller gains than for visual material in laboratory studies, though the one classroom study found more	Interleaving benefits are small for mathematical tasks [~M] (g = 0.34 for mathematical tasks); Interleaving improves discrimination [+S] (shuffled mathematics practice improved a delayed test); Interleaved practice produces higher mathematics test scores than blocked practice in a classroom experiment [+M] (126 seventh-graders, graph and slope problems over three months); the benefit held for both problem types at 30 days but was not significant for slope problems at 1 day (by problem type and delay [~M])
The kinds are grammatical forms the learner must choose between (verb conjugation classes, tense or case endings)	Treat as choosing among rules: mix the forms in practice once each has been met, rather than blocking as for word lists	Interleaved practice improves delayed posttest performance compared with blocked practice [+W] (one experiment with an Italian conjugation task: fewer trials and less time on a test two days later; accuracy on new forms was not the measure)
The material is expository text or prose passages	Do not interleave passages to improve learning of the passages; keep each text whole	Interleaving benefits are absent for expository texts [-M] (no significant effect)
The material is word lists or vocabulary categories	Block by category; interleaving reversed in this material	For word materials the effect favoured blocking [-M] (g = −0.39)
The kinds are easy to tell apart (low similarity between categories), or each kind varies a lot internally	Interleaving has little to work on: keep blocks, or first teach what varies within a kind	Between-category similarity raises and within-category similarity lowers the interleaving effect [~M]; a systematic review reaches the same boundary, with blocking possibly better where members of a kind are diverse (interleaving is most useful when differences are subtle [~M])
Each kind is defined by a clear verbal rule the learner is given (a definition, a stated criterion)	Blocking may do as well; if mixing, take the definition off the screen during the mixed set and ask the learner to apply it from memory	Interleaving is especially useful without a clear verbal rule [~W] (a review's narrative summary: blocking better in one rule-based study, a definition on screen removed the benefit in another)
Learners are `novice` on every kind and the kinds have not yet been met	Meet each kind on its own first (two or three items), then mix; do not open with mixed practice of unfamiliar kinds	untested proposal; the blocked-then-mixed order is a design choice the meta-analysis did not test; guidance that helps novices may be redundant later (expertise reversal [~M])
Learners are `advanced` on the kinds, or already strong in the subject	Skip the blocked introduction, but check rather than assume that mixing helps them: give a short mixed check first, and if they already discriminate the kinds, spend the time on harder kinds instead	In the few studies that split by ability, learning techniques helped low-ability and not high-ability students [~W] (interleaved practice had negative effects in the high-ability group; 6 studies of high-ability students across all techniques)
Learners are children or adolescents	Use interleaving where the condition holds, but do not promise the laboratory effect sizes: they come mostly from university samples, and one review found no high-school study at all. The one school study in the wiki, in seventh-grade mathematics, found a benefit	Interleaving effects are larger in university student samples [~M]; the literature is dominated by laboratory studies of undergraduates [~W]; at a 1-day delay, interleaved practice gave d = 0.42 in a seventh-grade classroom [+M]
The test is immediate, or the design is judged on practice scores	Expect mixed practice to look worse; judge on a delayed, mixed check instead, a month later where the course allows	Interleaving improves discrimination [+S] (worse during practice, better at the delayed test); the meta-analysis found no moderation by retention interval (null moderators [~M]); in the classroom study the benefit was d = 0.42 at 1 day and d = 0.79 at 30 days, though the difference between delays was not significant (the benefit does not diminish over 30 days [+M]); a review finds the benefit persists at delays of a day to ten days (not caused by spacing, persists over delay [+W])
Learners choose their own practice order (`online-self-paced`, homework, revision)	Build the mixing into the materials or the software; do not rely on learners to choose it, since learners in two studies judged blocking more effective	Interleaving improves discrimination [+S] (Kornell & Bjork: most learners believed blocking had served them better); Learners judge blocking more effective even when their test performance favours interleaving [~M] (about two thirds of 114 undergraduates); the same misjudgement for spacing is on Spaced Learning
The schedule puts other activities between practice items (a quiz, a video, unrelated questions), or a learning app spreads items across the day	Keep items of different kinds next to each other so they can be compared; spread each kind out by placing items of other kinds between its instances, not unrelated material	Spacing and interleaving are subadditive [~M] (with spaced presentation interleaving did not beat blocking); interrupting juxtaposition impairs inductive learning [~M]; within interleaved sequences, a larger lag between same-category items helped [+M] (d = 0.75; all three from one paper with undergraduates and online adults)
Homework or review sets in a school course	Mix earlier problem types into each later set, alongside new ones	Spaced and interleaved homework practice improves learning and retention [+W] (a practice brief summarising research, not a new study)
A single session only	Introduce the kinds, then one mixed set and one mixed check at the end; do not claim retention	untested proposal
One teacher for many learners	Give everyone the same mixed set and the same "name the kind first" step; look at which pairs of kinds are confused across the class and re-teach those pairs	untested proposal
Stakes are high and the kinds are dangerous to confuse (look-alike drugs, similar alarms, similar fault patterns)	Interleave the confusable pairs specifically, and require correct discrimination on a delayed mixed check before sign-off	untested proposal; the condition (high between-category similarity) is where the evidence predicts the largest benefit
The brief names its own unit order or textbook sequence (chapter by chapter, one type per chapter)	Keep the brief's order for introducing kinds; change only the practice and review sets, which mix earlier types back in	untested proposal

Observation, state and explanation¶

These rows are proposals, untested with learners as diagnostics. A check is itself practice and changes what it measures, so record each one.

Response	Plausible explanations	Observation that could change the interpretation
Accurate on blocked practice, many errors on a mixed set	Has learned each procedure but not when it applies; the kinds were mixed before each was recognisable; the mixed items also differ in surface features	Ask the learner to name the kind of each mixed item before solving it; compare naming accuracy with solving accuracy.
Names the kind correctly but still solves wrongly	The choice is learned; the procedure itself is not	Give a blocked item of that kind; if it also fails, the problem is the procedure, not discrimination.
Confuses two particular kinds consistently	Those two share the features the learner attends to; the distinguishing feature was never pointed out	Show a pair side by side and ask what differs; see whether the confusion persists on new examples.
Prefers blocked practice and asks for it	Mixed practice feels slower and less successful; the learner reads difficulty as failure	Show the learner their own delayed mixed-check results under each kind of practice, if both were used.
Mixed practice helps on the practised examples but not on new ones	The learner memorised examples rather than the features of the kind	Test with new examples of each kind, not the practised ones.

Evidence and qualifications¶

Interleaving improves discrimination [+S]: three studies. Kornell & Bjork (2008) found learners classified new paintings better after interleaved study of artists, while most believed blocking had helped them more; Rohrer & Taylor (2007) found shuffled mathematics practice gave far better delayed performance despite worse practice; Brunmair & Richter (2019) found the overall benefit depends on category similarity.
Interleaved practice produces a moderate overall benefit for inductive category learning [+M]: g = 0.42 across 238 effect sizes in 158 samples (Brunmair & Richter 2019). The same finding is recorded on Interleaving category examples improves inductive category learning [~M]; the two claim pages are candidates for a claim merge.
Between-category similarity raises and within-category similarity lowers the interleaving effect [~M]: the condition this page is built on.
Interleaving benefits are small for mathematical tasks and absent or reversed for expository texts and words [~M]: g = 0.34 for mathematical tasks, no significant effect for expository texts, g = −0.39 for words. The limiting claim: interleaving is not a general improvement to any practice.
Interleaving shows positive effects for all visual stimulus types [+M]: g = 0.67 for paintings, 0.35 for naturalistic photographs.
Several candidate moderators show no reliable effects [~M]: no moderation by retention interval, retention against transfer test, simultaneous against successive presentation, or publication status. Not detecting a moderator does not show it has no effect.
Interleaving effects are larger in pure university student samples [~M]: the effect sizes above may overstate the benefit for other learners.
Spaced and interleaved homework practice improves learning and retention [+W]: a practice brief, not a new study; it carries the idea into homework design.
Rohrer, Dedrick & Stershic (2015), the one classroom experiment in the wiki: 126 seventh-graders practised graph and slope problems for about three months, each student getting some types interleaved and some blocked, then took an unannounced test of new problems. Interleaved practice produced higher test scores [+M] (partial η² = .165); d = 0.42 at 1 day [+M] and d = 0.79 at 30 days [+M], with the schedule-by-delay interaction not significant (p = .09), so a growing benefit is suggested, not shown; by problem type [~M], the slope-problem effect at 1 day (d = 0.43) was not significant. Problems were mixed with reviews of earlier types, so spacing is not separated here.
Firth, Rivers & Boyle (2021), a systematic review of 26 studies with a meta-analysis of 17: interleaving improved classification of new examples [+M] (effects up to 0.66) and of studied examples [+M] (up to g = 0.65), reported as maxima, not pooled means; the benefit persists when spacing is held constant and at delays up to ten days [+W]; it is largest where differences are subtle [~M] and weaker or reversed where a clear verbal rule defines each kind [~W]; and no included study used high-school learners [~W].
Birnbaum, Kornell, Bjork & Bjork (2013), experiments with undergraduates and online adults learning bird and butterfly species: interleaving beat blocking only when items were contiguous [~M]; unrelated questions placed between interleaved items impaired learning [~M]; with order held interleaved, a longer lag between same-kind items helped [+M] (d = 0.75); and about two thirds judged blocking more effective [~M]. Together these make juxtaposition of kinds, not spacing alone, the working part, and they are why step 3 keeps mixed items back to back.
In the few studies that split by ability, learning techniques helped low-ability students and not high-ability students [~W]: Donoghue & Hattie (2021), with negative effects for interleaved practice in the high-ability group; only 6 studies of high-ability students across all ten techniques, so this is a caution, not a finding about interleaving alone.
Interleaved practice improves delayed posttest performance compared with blocked practice [+W]: one experiment (132 participants, an Italian verb-conjugation task) in which interleaving reduced the time and trials needed on a test two days later (η² = 0.09); the first wiki evidence of interleaving helping in language material, and it is grammar rules, not word lists.

Taken together: mixing kinds in practice reliably helps learners tell confusable kinds apart, most for visual categories and in one classroom study for mathematics; it does nothing or harm for texts and word lists. The benefit seems to come from comparing kinds side by side, and survives spacing being held constant, but inserting unrelated material between items removes it. No claim in the wiki tests a dose (how much mixing, after how many blocked items), and only one study in the wiki was run in a school.

The evidence by question¶

Should examples or problems of confusable categories be met in blocks or interleaved, and for which materials? Interleaving category examples improves inductive category learning for visual and mathematical materials, but not for expository texts or word categories [~M]
How alike must the categories be for mixing to help, and when does blocking do better? Interleaving effects are stronger when material is more similar between categories, less similar within categories, and more complex [~M]
Does it work in a school classroom, over weeks, on an ordinary test? Interleaved practice produces higher mathematics test scores than blocked practice in a classroom experiment [+M]: one study, seventh-grade mathematics.
Is the benefit just spacing in disguise? The interleaving benefit is not caused by spacing [+W] and spacing and interleaving are subadditive [~M]: the review and the experiments agree that comparing kinds side by side does work spacing does not.
Does it help with new examples, not only the practised ones? Interleaving benefits transfer to novel items from a learned category [+M]

Further evidence, not yet read against this model¶

Learners prefer massed study despite its inferior outcomes [~M]: the same misjudgement as Kornell & Bjork's learners, for spacing; it bears on why interleaving needs to be built into materials rather than left to learners.
After training, d.tech teachers adopted low-stakes practice tests and spaced, interleaved practice [+W]: adoption by teachers, not a learning outcome.
Accuracy in the mixed block decreases across trials while PES also decreases, independent of intelligence [~W] — read 2026-10-19 (not about this page): a "mixed block" of an executive-function task in kindergarten children, where accuracy fell with fatigue; it is not mixed practice of kinds to be learned.
High and low CI practice do not differ significantly in overall relatively permanent gains (Δ pre-retention) [-W] — attached 2026-10-21 from Achraf Ammar et al. (2024), which proposed "Do not assume high contextual interference (random) practice guarantees a longer-term performance advantage in PE and sports"; tests this page's relationship.

Objective and learner-valued goal¶

The designer's objective is usually correct choice among kinds at a later time, without being told which kind it is. Ask separately what the learner wants: a learner preparing for an exam of mixed questions has a goal that interleaving serves directly, while one who wants to feel fluent today may resist mixed practice because it feels worse. Record that preference as an observation. Do not overrule it silently, and do not read it as evidence that blocking works better.

What would revise this model?¶

The expectation should weaken if comparisons with comparable learners find no advantage for mixing on a delayed, mixed test where the kinds are confusable, or if the advantage disappears in further studies that hold spacing constant. If new syntheses find benefits for expository texts or words, the material rows above should change. If the blocked-then-mixed start loses to mixing from the first item for novices, step 2 of the default design should go. If school studies in subjects other than mathematics find no benefit, or studies of high-ability or advanced learners confirm a cost, the learner rows should narrow where interleaving is recommended.

Claims¶

Spaced Learning — when practice of the same thing recurs; usually confounded with interleaving
Sequencing — the wider order of a unit; carries blocked-then-mixed as one choice
Retrieval Practice — practice that asks for recall
Worked Examples — how each kind is introduced before mixing

Examples¶

Interleaving and Interleave practice types are the strategies that carry this principle into an activity.
Include a small block of problems of a new skill at the start of an assignment, then interleave the rest

Key Sources¶

Brunmair, M., & Richter, T. (2019). Similarity matters: A meta-analysis of interleaved learning and its moderators. Psychological Bulletin, 145(11), 1029–1052. doi:10.1037/bul0000209
Rohrer, D., Dedrick, R. F., & Stershic, S. (2015). Interleaved practice improves mathematics learning. Journal of Educational Psychology, 107(3), 900–908. doi:10.1037/edu0000001
Rohrer, D., & Taylor, K. (2007). The shuffling of mathematics problems improves learning. Instructional Science, 35(6), 481–498. doi:10.1007/s11251-007-9015-8
Kornell, N., & Bjork, R. A. (2008). Learning concepts and categories: Is spacing the enemy of induction? Psychological Science, 19(6), 585–592. doi:10.1111/j.1467-9280.2008.02127.x
Firth, J., Rivers, I., & Boyle, J. (2021). A systematic review of interleaving as a concept learning strategy. Review of Education. doi:10.1002/rev3.3266
Birnbaum, M. S., Kornell, N., Bjork, E. L., & Bjork, R. A. (2013). Why interleaving enhances inductive learning: The roles of discrimination and retrieval. Memory & Cognition, 41, 392–402. doi:10.3758/s13421-012-0272-7
Donoghue, G. M., & Hattie, J. A. C. (2021). A meta-analysis of ten learning techniques. Frontiers in Education, 6, 581216. doi:10.3389/feduc.2021.581216
Zeitlhofer, I., & Zumbach, J. (2026). Germane processes moderate the effectiveness of complex desirable difficulty tasks. Instructional Science. doi:10.1007/s11251-026-09807-1

Page metadata

Field	Value
`type`	`principle`
`id`	`interleaving`
`aliases`	`interleaved-practice`, `mixed-practice`
`title`	`Interleaving`
`description`	`When the goal is to tell apart categories or problem types that resemble one another, mixing examples or problems of different kinds in practice (interleaving) is expected to improve later classification and problem choice over practising one kind at a time (blocking); a meta-analysis of 59 studies supports this for visual categories and, more weakly, mathematics, finds no benefit for expository texts and a reversal for word lists, and finds the benefit grows with how similar the categories are.`
`status`	`review`
`canonical`	`True`
`generated`	by `claude/unspecified` on `2026-10-19`
`trust`	unverified — no one has confirmed this page's content
`sources`	8 — `brunmair-richter-2019`, `rohrer-dedrick-stershic-2015`, `rohrer-taylor-2007`, `kornell-bjork-2008`, `firth-rivers-boyle-2021`, `birnbaum-et-al-2013`, `donoghue-hattie-2021`, `zeitlhofer-zumbach-2026`

Frontmatter is the page's machine-readable record — see the Schema & Guide.