Study organised neatly by chapter raises performance inside the session and lowers it on the paper. The evidence for spacing and interleaving is old and consistent; the reason learners avoid both is the more interesting half of the story.
Blocked practice is the default nearly everywhere. A chapter is read, problem after problem of one kind is worked in sequence, and the session closes with a sense of competence that is not wholly false. Accuracy really does rise across such a block. The rise is the problem. After the first two or three items the method is already in mind, and every item after that asks only that it be executed, not that it be found. What improves is the ability to run a procedure already open; what the examination asks for is the ability to recover a closed one.
Soderstrom and Bjork's 2015 review set out the distinction most revision timetables ignore: performance during training and learning are different quantities, and conditions that raise the first often lower the second. A learner watching accuracy climb inside a block is measuring the wrong thing honestly. Nothing in that session tests whether the method can be produced a fortnight later, in a hall, with no chapter heading to name it.
The spacing effect is among the oldest reproducible findings in psychology, first recorded by Ebbinghaus in the 1880s. The 2006 synthesis of distributed practice by Cepeda and colleagues found that spaced repetitions beat massed ones, and that the useful gap scales with how long the material has to last: days for material needed next week, and much longer gaps for material that must still be there at an examination the following year.
Robert and Elizabeth Bjork's account of storage strength and retrieval strength explains the mechanism. A memory recovered after it has begun to slip gains more than one refreshed while still fresh: the difficulty of the recovery is what strengthens the route, and partial forgetting is the condition that makes a repetition worth anything. Roediger and Karpicke's 2006 experiments on retrieval practice sharpened the point. Learners tested on a passage did worse than re-readers minutes later, and clearly better a week on. They predicted the reverse.
Interleaving applies the same bargain to order rather than timing. Shea and Morgan's 1979 experiments on motor skills found that randomly ordered practice produced worse performance during acquisition and better retention afterwards than practice blocked by movement type. Rohrer and Taylor found the same in mathematics, where learners given mixed problem sets scored lower during practice and far higher on a delayed test than those given identical problems in blocks. The mechanism is plain: a block supplies the identification of the problem for free, leaving only execution.
The structure of the examination makes the same case from the other side. In an archive of 8,267 authentic JEE Main questions from papers between 2002 and 2025, 117 topics carry enough questions to describe a stable pattern: 48 in Physics, 41 in Chemistry, 28 in Mathematics. A candidate meets those topics in an order designed to prevent runs, with nothing announcing which of several dozen methods applies. Blocked preparation rehearses every part of that task except the first.
The cost of failing that first step is not only a mark. Marking schemes set break-even guessing accuracy at penalty divided by reward plus penalty: 20 per cent for JEE Main and NEET, 25 per cent for UPSC Prelims. A blind pick from four options is therefore neutral at UPSC and mildly positive at the other two, so the value of a guess is small and roughly fixed. What is not fixed is the time lost forcing a misidentified problem through the wrong method before arriving at that same guess.
Kornell and Bjork's 2008 study, in which participants learned painters' styles from blocked or interleaved examples, shows why the better method loses. Interleaved learners identified unseen works more accurately. Most participants still judged blocking to have served them better, and many held that view after seeing their own results. The signal learners trust is fluency, and fluency is exactly what interleaving removes.
The second cost is administrative. Blocked study schedules itself: the chapter is the unit, and finishing it marks progress. Spaced, interleaved study has no natural unit. It needs a record of what was practised, when, and how well it went, plus a daily decision about which old material is due. That bookkeeping is real work, and where no system carries it, learners drift back to the chapter because the chapter is legible.
Institutions reinforce the default. Textbooks are blocks, timetables are blocks, and the test at the end of a chapter is blocked by construction, returning a flattering score just as a learner decides whether the method worked. Two limits deserve honest statement. Interleaving helps when items are confusable and each has been encoded at least once, since mixing wholly unfamiliar material produces confusion rather than discrimination. And no arrangement of practice repairs an explanation that was never understood.
Frequency evidence changes the allocation question rather than the method. Matrices and determinants account for 8.8 per cent of JEE Main Mathematics since 2016, 275 questions across 23 years. Kinematics in one and two dimensions accounts for 6.6 per cent of recent Physics, 180 questions. Coordination compounds account for 8.7 per cent of recent Chemistry, 171 questions. Weights like those describe near-certain topics rather than occasional ones, and a near-certain topic must be available on demand at a date months away.
Availability at a distant date is the specific product of spacing, and the thing a block never delivers. A topic of that weight is better served by short returns scattered across the preparation than by one heavy week ending in a chapter test. The trade deserves stating plainly, because learners not warned of it abandon the method within a fortnight. Spaced, interleaved practice lowers the score of the practice session and raises the score of the paper. Only one of those is marked.
The full set, with progress tracking and five agent perspectives per question, is in the JupiteX app — browse the exam catalogue or browse the Learn library.