Blocked practice feels productive because success during the session is so visible, but the feeling of mastery is largely an illusion that collapses once the context changes. Interleaving works because it forces the part of the skill that blocked practice quietly removes: deciding what kind of problem you are even looking at.
Why blocked practice lies to you
When you do twenty power-rule problems in a row, the first one or two require genuine problem solving. After that, the page itself is a cue. Seeing the form f(x) = x^n triggers "use the power rule," and the rest is arithmetic. Your accuracy climbs, your speed climbs, and you interpret both as evidence of learning. What you have actually learned is how to apply a known procedure when you already know which procedure it is.
Interleaving breaks that. If a session looks like power rule, chain rule, product rule, power rule, chain rule, product rule, no single problem tells you what to do. You have to inspect the structure of the function and decide. That decision is the part of the skill that an unlabelled exam actually tests.
A concrete example
Imagine a student who has learned three integration techniques: substitution, integration by parts, and partial fractions. After a blocked session of fifteen substitution problems, they finish feeling sharp. After an interleaved session of five of each in a shuffled order, they finish feeling shaky, with perhaps three errors.
A week later, both take a mixed test of fifteen problems. The blocked-practice student, deprived of the topical cue, often stumbles on items that demand a fresh recognition step they never trained. The interleaved student, slower and more error-prone during the session, usually handles the same test more reliably. The trade that looked bad on day one often reverses itself by day seven.
The mechanism: comparing makes things distinct
Putting similar problems next to each other forces you to compare them. Two solved problems of the same type just confirm what you already knew. A substitution problem sitting next to a partial fractions problem forces you to ask what makes this one not that one. That question, repeated across a session, is where the boundaries between categories become sharper. The work is not the arithmetic you can already do; it is the discrimination you have been skipping.
When interleaving does not help
- If you have never seen the basic procedure, mixing in unsolved problems produces confusion, not learning. Build a minimum of fluency in a topic first, then switch to mixing.
- Pure mechanical skills with no decision step, such as single-digit arithmetic, gain nothing from interleaving. The decision is trivial, so removing it changes nothing.
- Strictly sequential material, where topic B cannot be understood without topic A, must be introduced in order before interleaving can do anything useful.
- If the topics are too dissimilar, mixing them adds effort without sharpening any boundary. The contrasts need to be close enough to be confused.
Putting it into practice
- Spend a short blocked session on each new topic, just long enough to know the procedure and to solve one or two problems unaided.
- Build a queue of unsolved problems from two or three related topics and work through them in a shuffled order, not a topical one.
- Judge a session by what you can do on a fresh page a day or two later, not by how many you got right while the topic label was sitting in front of you.
Cram: I finish one topic completely before I start the next. Twenty problems of the same type, then move on. That is how you master something.
Rep: That is called blocked practice, and it is the most convincing bad habit in studying. It makes you fast inside the block and helpless outside it.
Cram: But by problem fifteen I am getting them all right. That has to mean something.
Rep: It means you stopped solving problems. After the first few you already know which method to use, because the page told you. You are practising the arithmetic, not the decision.
Cram: What decision? The problem is right there in front of me.
Rep: The real exam never labels the question. The hardest step is looking at something cold and working out what kind of problem it even is. Blocked practice removes that step entirely.
Cram: So what am I supposed to do instead, jump around at random?
Rep: Mix the types. Do a few of one, then a few of another, then come back. Interleaving. Every problem now starts with a question you had been skipping.
Cram: I tried something like that once. It was slower and I made more mistakes.
Rep: It will be. Studies of mixed practice show worse performance during the session and much better performance on the delayed test. Sometimes several times better.
Cram: That sounds like a bad trade while I am sitting there struggling.
Rep: It only looks bad because you are measuring the wrong day. Speed during study predicts almost nothing. What you can do a week later, unprompted, is the thing you actually want.
Cram: Why does mixing help at all though? Nothing about the topic changed.
Rep: Because sitting them side by side forces you to notice the differences. You learn what makes this one not that one. Similar things only become distinct when you compare them.
Cram: So my neat one topic per session plan was training the easy half of the skill.
Rep: Exactly that. Keep the topics, break the blocks. Shuffle the problems until you no longer know what is coming, because that is the state you will be in when it counts.