Insight

Practice Feels Good Right Before It Fails You

James·Tuesday, November 3, 2026 Edition

WHEN THE REPS FEEL RIGHT

Open the document. Add twenty problems. Sort by type — all the percentage questions together, then all the ratio questions, then all the rate questions. Work through them in order. Check answers as you go. Notice, somewhere around problem fourteen, that the checking has become a formality. Linh Bùi noticed it too, the evening before her licensing exam, and took it as a good sign.

It wasn't. The smoothness you feel near the end of a blocked session is real — it just belongs to the context, not to you. When every question is the same type, your brain stops selecting a strategy and starts completing a pattern. The retrieval is being done partly by position. Problem eleven looks like problem ten, so the same pathway fires, frictionlessly. That friction. The half-second of not knowing which tool to reach for. Is exactly what learning feels like, and blocked practice removes it. You are measuring fluency. Fluency is not transfer.

The research on interleaving is about as clean as educational science gets. Robert Bjork's work at UCLA has documented the effect across motor skills and cognitive ones: mixing problem types within a session consistently produces worse performance during practice and better performance on delayed tests. The gap between how hard it feels and how well it works is large enough that students who interleave routinely rate themselves as having learned less, even when their test scores are higher. Difficulty during practice is not a sign that something is wrong. It is the sign that retrieval is actually happening.

THE HOUR BEFORE THE EXAM

Say the exam is in fifty minutes and Linh's folder is still sorted by type. The instinct is to keep drilling the category that feels shakiest. To run more percentage problems until they feel locked. That instinct will make the next fifty minutes feel productive and cost her points tomorrow. Instead: shuffle. Literally reorder the remaining problems so a rate question follows a percentage question follows a ratio question. Work three of them. Notice the small drag of reorientation at the start of each one. That drag is the mechanism. The brain is selecting, not just completing, and that selection is what will fire when a random question appears on a page with no category label above it.

The practice that feels most like learning is usually least like it. You are not a bad student, or an undisciplined one. You built a session that hid the work from you, and the session let you.

🎯

Try This

Take whatever you are drilling this week and break the category sequence: after every third problem, switch to a different problem type before returning. Set a timer for twenty minutes — not to finish, but to stay in the discomfort long enough for it to normalize.

How?

Key Facts

*Fluency during practice is not evidence of learning — it is evidence that the current context is doing half the retrieval work for you.

*Interleaving problem types feels harder and slower in the session, but produces significantly better performance when the context changes, which is always when it matters.

*Before your next high-stakes task, deliberately practice a different version of the skill first, so your brain cannot rely on pattern-matching by position alone.