Interleaving: why mixing your words beats blocking them

Mixing wins because every item makes you work out its category again, where a blocked batch hands you the answer. The shuffle also spaces each category out for free.

In Rohrer's 7th-grade math classes, interleaved practice scored 72% on a test a month later, against 38% for blocked practice. On plain vocabulary lists, the meta-analysis in the correction above goes the other way.

Say you've got 40 new words from three lessons: food, travel, feelings. The obvious way to study is one theme at a time. Knock out all the food words, then all the travel words, then the feelings.

Clean. Organized. It feels like sorting laundry.

Blocking is also the worse way, by a margin that surprised the researchers who found it. The better way is to shuffle all 40 together so you never know what's coming next. It feels sloppier and slower, and it wins on every delayed test anyone has run.

01 / 06

One shuffle buys you interleaving and spacing at the same time

The tidy method has a name: blocking. All of category A, then all of B, then all of C. The messy one is interleaving: A, C, B, A, B, C, mixed so consecutive items rarely share a category.

Blocked, the batch tells you the answer before the card does Nine cards, three categories, one session, identical in both rows. Only the order differs. Mixed, consecutive cards rarely share a category, so every single card asks the question again. Schematic. Colour stands for category (food, travel, feelings); the two rows hold the same nine cards.

Blocking feels better while you do it, for the same reason cramming does. You get on a roll.

The fifth food word rides in on the momentum of the first four. But that momentum is a crutch, and interleaving kicks it out from under you on purpose.

Interleaving also smuggles in a second effect for free: spacing. Mix 3 categories and each category's items automatically sit further apart in the session than they would blocked.

So every return to a category is a slightly faded, slightly effortful retrieval. Two of the strongest effects in the learning literature, from one shuffle.

Brunmair and Richter's 2019 meta-analysis (a study that pools many studies) across 238 comparisons confirmed the advantage holds broadly, strongest exactly where materials are similar enough to confuse.

02 / 06

Rohrer reordered a worksheet and the delayed scores nearly doubled

Doug Rohrer has run the most convincing interleaving study, in real classrooms with real math. In one study (Rohrer, Dedrick & Stershic, 2015), 7th-graders did their normal practice problems either blocked by topic or interleaved across topics. Same problems, same number, just reordered.

On a surprise test a month later, the interleaved group scored 72% against the blocked group's 38%. That's nearly double, from shuffling a worksheet.

One month on, the shuffled group scored 72% and the blocked group 38% 7th-graders, their normal practice problems. Same problems, same number, same total practice. The interleaved group just met them in a mixed order. Data: Rohrer, Dedrick & Stershic 2015, 7th-grade mathematics, unannounced test one month after practice ended.

The key is that Rohrer's math test was delayed and mixed, which is what a real test (or a real conversation) always is. On an immediate, blocked quiz, blocking often looks fine. It's the delay and the mixing that expose it, which is exactly the trap of judging a method by how it feels in the moment.

03 / 06

Contrast is the teacher: you only see a style beside a different one

My favorite interleaving study has nothing to do with words. Kornell and Bjork (2008) showed people paintings by 12 landscape artists and asked them to learn each artist's style well enough to identify new paintings later. Some saw each artist's works blocked together; others saw them interleaved.

The interleaved group was much better at identifying the painter of brand-new paintings.

And here's the human part: most participants swore blocking had worked better for them. They preferred the method that taught them less. (If you want more on why the harder-feeling option keeps winning, that's desirable difficulty.)

The reason interleaving helped is telling. Seeing a Monet right after a Seurat forces you to notice what makes a Monet a Monet.

Blocking hides those contrasts; interleaving throws them in your face. It teaches discrimination, the skill of telling similar things apart.

04 / 06

The method that looked best during practice scored worst a day later

The Kornell and Bjork participants weren't unusually stubborn. The misjudgment shows up in nearly every interleaving study, at every age, and it survives being shown the results.

Taylor and Rohrer's 2010 study with 4th graders is the cleanest demonstration of why.

During practice, the blocked kids looked terrific: about 99% correct while working, because every problem's method was announced by the batch it arrived in. The interleaved kids slogged along around 68%.

A day later, on a mixed test, the numbers flipped hard: interleaved 77%, blocked 38%.

If you were the teacher watching the practice session, blocking looked like the obviously better method. If you were the kid, it felt like it too. Practice performance is simply a bad witness (the same performance-versus-learning gap that makes cramming so seductive), and blocking bribes the witness.

Textbooks are built blocked for the same innocent reason: chapter 4 is ser, chapter 5 is estar, and every exercise in each chapter answers its own chapter's question. No one's lying to anyone. The format just rewards how learning looks instead of how it lasts.

05 / 06

Where two words keep blurring, mixing them is the training

Language is full of things that need telling apart: ser and estar, por and para, saber and conocer.

A beginner blends them together, because they've only ever met them in separate, blocked lessons.

Interleaved review is discrimination training for exactly these confusable pairs. When por shows up sandwiched between a food word and a travel word, you can't lean on "we're doing prepositions now." You have to actually retrieve which one this sentence needs, which is the skill you're after.

There's direct language evidence too. Studies on interleaving foreign-language grammar and vocabulary (for example, work on Spanish verb structures) find the same delayed-test advantage Rohrer got in math. The effect shows up wherever categories can be confused, numbers or not.

Try the felt version of ser and estar yourself. Ten ser sentences in a row, and by number 4 you've stopped choosing; the hand knows the answer is ser.

Now shuffle 5 ser and 5 estar sentences together, and every single item forces the actual question: permanent trait or current state?

Ten real decisions instead of one decision and 9 echoes. Same 10 sentences, roughly 10 times the learning.

06 / 06

Does a spaced-repetition scheduler already interleave your reviews?

Yes. You basically can't avoid interleaving in a real spaced-repetition system, and that's by design.

When your reviews are scheduled by when each card is due rather than by topic, the deck shuffles itself. A word from Monday's food lesson comes back on the same day as a word from Wednesday's travel lesson, because that's when they're both ripe. The scheduler produces interleaving as a side effect of doing its actual job.

So the practical advice is short:

• Don't sort your review into neat themed sessions. Let the due-date order stand, even when it feels random.
• Resist the urge to "focus on food words today." That's re-blocking, and you'd be trading away the effect on purpose.
• When two words keep blurring together, that confusion is the interleaving working. Grade honestly, let them both come back, and let the contrast teach you.

The messy order is the feature. A study session that feels a little chaotic, where you never quite know what's next, is doing more for you than the tidy one ever did.

And notice what a real conversation is: interleaving at maximum setting.

Your tutor doesn't quiz you on food words for 10 minutes. The dinner story requires a food word, then a past tense, then a feeling, then a number, in whatever order the story demands. Interleaved review is just conversation's demands, rehearsed in advance.

---

Sources