The best study techniques by evidence are practice testing (quizzing yourself from memory) and distributed practice (spreading study over days), the only two rated "high utility" in Dunlosky et al.'s 2013 review of ten methods. Interleaving, elaborative interrogation and self-explanation rated moderate; summarising, highlighting, the keyword mnemonic, imagery for text and rereading rated low, even though rereading and highlighting are the methods students use most.
- Practice testing and distributed practice are the only two "high utility" methods of the ten Dunlosky's team rated.
- Rereading and highlighting, the most popular methods, both rated low.
- Space the first review at roughly 10 to 20% of the time left to the exam (from Cepeda et al. 2008).
- Interleaving beat blocked practice by g = 0.42 across 59 studies, but lost on word lists.
- Testing feels worse and works better: students who restudied were more confident and remembered less a week later.
The ranking below uses one stated measure: the utility rating each technique received in Improving Students' Learning With Effective Learning Techniques by John Dunlosky, Katherine Rawson, Elizabeth Marsh, Mitchell Nathan and Daniel Willingham, published in Psychological Science in the Public Interest (volume 14, pages 4 to 58).
The authors judged each method on whether its benefit held across four kinds of variables:
- learning conditions;
- student characteristics;
- materials;
- the type of test.
A high rating means the benefit generalised widely and had been shown in real classrooms. Within a tier, I order techniques by the strength of later meta-analytic evidence where it exists, and by the review's own wording where it does not.
This matters in India because the stakes are concentrated. NEET UG 2026 alone had about 22.8 lakh registered candidates, by the National Testing Agency's count reported by Careers360. Most of them will spend the bulk of their hours on the two lowest-rated methods.
Which study techniques does the evidence rank highest?
Practice testing and distributed practice are the clear winners, because they helped learners of different ages and abilities, worked across many materials and test formats, and held up in classrooms. Everything else either lacked classroom tests or helped only some students on some tasks.
See the data as a table
| Utility (3 = high, 2 = moderate, 1 = low) | |
|---|---|
| Practice testing | 3 |
| Distributed practice | 3 |
| Interleaved practice | 2 |
| Elaborative interrogation | 2 |
| Self-explanation | 2 |
| Summarisation | 1 |
| Imagery for text | 1 |
| Keyword mnemonic | 1 |
| Highlighting | 1 |
| Rereading | 1 |
Source: Dunlosky, Rawson, Marsh, Nathan and Willingham, Psychological Science in the Public Interest, 2013 · The review gives categorical ratings (high, moderate, low). The 3-2-1 scale is mine, for plotting only.
| Rank | Technique | What you actually do | Dunlosky 2013 rating | Key extra evidence |
|---|---|---|---|---|
| 1 | Practice testing | Recall answers from memory: flashcards, past papers, blank-page recall | High | Roediger and Karpicke 2006; Karpicke and Roediger 2008 |
| 2 | Distributed practice | Spread reviews of the same topic over days or weeks | High | Cepeda et al. 2006 (839 assessments); Cepeda et al. 2008 (1,350+ learners) |
| 3 | Interleaved practice | Mix problem types in one session instead of blocking them | Moderate | Brunmair and Richter 2019: g = 0.42 overall, 0.34 for maths |
| 4 | Elaborative interrogation | Ask "why is this true?" for each fact | Moderate | Few classroom tests at the time of the review |
| 5 | Self-explanation | Explain each step of a solution to yourself | Moderate | Few classroom tests at the time of the review |
| 6 | Summarisation | Write summaries of what you read | Low | Helps some students on some tests only |
| 7 | Imagery for text | Form mental pictures while reading | Low | Benefits limited to imagery-friendly text |
| 8 | Keyword mnemonic | Link a word to a sound-alike image | Low | Works for some vocabulary, fades over long delays |
| 9 | Highlighting | Mark text with a highlighter | Low | No consistent gain on performance |
| 10 | Rereading | Read notes or chapters again | Low | Raises confidence more than recall |
In 1879–80, and again in 1883–84, the German psychologist Hermann Ebbinghaus ran a memory laboratory with one subject: himself. To stop meaning from helping him, he built about 2,300 nonsense syllables, each a vowel sound between two consonants. He read each series aloud at 150 beats a minute, 0.4 seconds a syllable, paced by a ticking watch and counting repetitions on a cord of wooden buttons, until he could recite it fluently.
Then he waited, from minutes to a month, and measured how much work relearning took. After an hour he needed about half the original effort again; after eight or nine hours, about two-thirds. After 31 days, about a fifth of the first learning still persisted. He published it in 1885 as Über das Gedächtnis, and every spacing study in this post is, in a sense, an answer to his curve.
Hermann Ebbinghaus: Memory, A Contribution to Experimental Psychology (1885; tr. Ruger and Bussenius, 1913), chapters 3 and 7, Classics in the History of Psychology1. Practice testing
Practice testing means pulling information out of memory before you look at it, and it is the single most reliable study technique in the literature. The point is not to measure what you know but to change it: each successful retrieval makes the memory easier to retrieve the next time.
The experiments behind it
The cleanest demonstration is Roediger and Karpicke's 2006 experiment in Psychological Science. Students read prose passages, then either restudied them or took free-recall tests without feedback.
- Five minutes later: the restudy group did better.
- Two days and one week later: the tested group retained substantially more, even though the restudy group felt more confident.
Two years later, in Science, Karpicke and Roediger showed the same with foreign vocabulary: once a word had been recalled once, more studying did nothing for delayed recall, while more testing produced a large gain.
Students' own predictions of their performance were uncorrelated with what they actually remembered.
That last finding explains why Indian coaching culture often gets this backwards. Rereading a module feels like progress; attempting a past paper and getting 40% feels like failure. The data says the second session taught you more.
- Dunlosky rating: high utility
- Best format: free recall or short answer, then check against the source
- Use for: NCERT facts, formulae, reactions, dates, vocabulary, case law

2. Distributed practice
Distributed practice, or spacing, means reviewing the same material across separate sessions instead of cramming it into one, and its benefit is among the most replicated effects in psychology.
Cepeda, Pashler, Vul, Wixted and Rohrer pooled 839 assessments from 317 experiments in 184 articles in 2006 and found that spaced study beat massed study, with one important twist: the best gap between sessions grows as the time until the test grows.
How long the gap should be
Their 2008 follow-up put numbers on that twist. More than 1,350 people learned a set of facts, got a review after a gap of up to 3.5 months, and were tested up to a year later. The best gap was about 20 to 40% of the test delay when the test was a week away, falling to about 5 to 10% when it was a year away. Too short a gap wastes the review; too long a gap means you have forgotten too much for the review to rebuild.
Use the calculator to turn that into a plan. It applies a 10 to 20% band, which sits between Cepeda's one-week and one-year figures and suits exams one to six months out. Treat it as a starting rule, not a law.
When should your first review be?
- Dunlosky rating: high utility
- Evidence base: 839 assessments, 317 experiments (Cepeda 2006)
- Rule of thumb: first gap roughly a tenth to a fifth of the time to the exam
3. Interleaved practice
Interleaving means mixing different kinds of problems in one session, for example alternating integration by parts, substitution and partial fractions instead of doing twenty of each in a row. It rated moderate in 2013 because the research was young; the evidence has since grown.
Brunmair and Richter's 2019 meta-analysis in Psychological Bulletin covered 59 studies with 238 effect sizes and found a moderate overall benefit (Hedges' g = 0.42).
- Visual materials such as paintings: largest effect, g = 0.67.
- Maths: smaller, g = 0.34.
- Word lists: reversed, blocking won, g = -0.39.
Interleaving helped most when categories were similar to each other and hard to tell apart. That is exactly the JEE problem: knowing which method a question needs is half the work, and blocked practice never trains that choice.
- Dunlosky rating: moderate
- Meta-analytic effect: g = 0.42 overall, 0.34 maths, -0.39 words
- Use for: maths, physics problem types, organic reaction types, look-alike diagnoses
4. Elaborative interrogation
Elaborative interrogation means asking "why would this be true?" about each fact and answering it, which links new facts to what you already know. Dunlosky's team rated it moderate because its benefits generalised across some variables but it had not been adequately tested in real classrooms.
It works best when you already have enough background to produce a sensible answer. A Class 12 student asking why alkali metals get more reactive down the group can reason from atomic size and ionisation energy; a beginner cannot, and the technique stalls. Use it on the second pass through a topic, not the first.
- Dunlosky rating: moderate
- Limitation: needs prior knowledge
- Best for: factual lists that have a reason behind them
5. Self-explanation
Self-explanation means explaining to yourself how each step of a worked example or text connects to the last, and it rated moderate for the same reason as elaborative interrogation: promising, but thinly tested in classrooms at the time. It costs time, which matters when a syllabus is as large as UPSC's.
Pair it with worked examples in quantitative subjects. Cover the next line of a solution, predict it, explain why, then uncover it. That turns a passive read of a solved paper into a sequence of small retrievals, which borrows some of the power of practice testing.
- Dunlosky rating: moderate
- Cost: slower per page
- Best for: derivations, proofs, solved examples

6. Summarisation
Summarisation rated low because it has helped some students on some kinds of test, but under limited conditions. A summary that copies sentences in shorter form gives little benefit; one written from memory with the book closed is really a form of practice testing.
- Dunlosky rating: low
- Fix: write the summary from memory, then check
7. Imagery for text
Forming mental images while reading rated low because the conditions under which it helps are narrow: text that describes something concrete and picturable, tested soon after. It does little for abstract material such as constitutional principles or thermodynamics.
- Dunlosky rating: low
- Works for: anatomy, geography, processes you can picture
8. Keyword mnemonic
The keyword mnemonic, linking a new word to a sound-alike image, rated low because it is hard to apply to many materials and its benefit fades at longer retention intervals. It can still be a useful patch for a stubborn list of terms, but it is not a study system.
- Dunlosky rating: low
- Problem: short-lived benefit, narrow use
9. Highlighting
Highlighting rated low because it does not consistently improve performance. Most students report relying on it, which is exactly why the authors included it: a method that common deserved a test, and it did not pass.
- Dunlosky rating: low
- Better use: mark what to turn into a question later
10. Rereading
Rereading rated low and the review's authors explicitly suggest replacing it with practice testing. It is the default method for a reason: it is fast and it produces fluency, the feeling that a page is familiar. Familiarity is not recall, and Karpicke and Roediger's students could not tell the two apart.
Familiarity is not recall.The lesson of Karpicke and Roediger, 2008
- Dunlosky rating: low
- Replace with: close the book, write what you remember, then check
Why do students keep using the low-rated methods?
Because the low-rated methods feel effective in the moment and the high-rated ones feel hard. Rereading raises fluency and confidence immediately, while testing exposes gaps, and spacing means returning to a topic after you have partly forgotten it, which feels like failure.
The effort is part of the mechanism, not a side effect. Three links explain why the easy path wins:
- Reward timing. If you cannot study but can scroll for hours, part of the reason is that the phone gives instant reward and testing gives delayed reward; I go into that in why you can scroll for hours but can't study.
- Working memory. It holds only a few items at once, so cramming overloads it, as explained in the working memory limit.
- Sleep. Spacing works partly because sleep consolidates what you learned between sessions; see how sleep consolidates memory.
A 55-page monograph reviewing ten study techniques chosen because they are easy to use or because students rely on them. Each was judged on whether benefits generalised across learning conditions, student characteristics, materials and test types. Practice testing and distributed practice rated high; interleaving, elaborative interrogation and self-explanation moderate; summarisation, highlighting, keyword mnemonic, imagery and rereading low. Limit: the ratings are expert judgements, not a pooled effect size, and the moderate techniques were downgraded for missing classroom evidence rather than for failing.
Read the paperMore than 1,350 people learned facts, received one review after a gap of up to 3.5 months, and took a final test up to a year later. For any test delay, lengthening the gap first improved and then slowly reduced final scores. The best gap was about 20 to 40% of a one-week delay and about 5 to 10% of a one-year delay. Limit: the material was trivia-style facts, not problem-solving, so the exact percentages may shift for maths or physics.
Read the paperIn two experiments, students studied prose passages and then either took one or three free-recall tests without feedback or restudied for the same number of times. Restudying won on a test five minutes later; testing won clearly on tests two days and one week later. Restudying increased students' confidence despite worse retention. Limit: university students and prose passages; it shows the effect, not the best schedule.
Read the paperA multilevel meta-analysis of 59 studies, 238 effect sizes and 158 samples comparing interleaved and blocked practice. Overall effect g = 0.42; paintings and visual material 0.67; maths 0.34; words -0.39, where blocking was better. Interleaving helped more when categories were similar to each other and the material more complex. Limit: few studies used expository text, so the advice for theory-heavy subjects is uncertain.
Read the paperHow should an Indian exam aspirant use this ranking?
Build the week around retrieval and spacing, and use the moderate techniques to deepen understanding. A workable plan for NEET, JEE, CAT or UPSC looks like this.
- Learn once, actively. Read or watch the lesson once, using self-explanation on solved examples.
- Test the same day. Close the book and write everything you remember, or do ten questions without notes. Check and mark what you missed.
- Schedule spaced reviews. Use the calculator above for the first gap, then roughly double the gap each time.
- Interleave in practice sets. After a chapter is learned, mix its question types with earlier chapters instead of doing chapter-wise sets.
- Replace rereading. Every time you reach for the module to reread it, attempt a past-paper question on that topic instead.
- Protect sleep. Spacing works across nights, not within them.
What's new
- A 2026 study finds worked examples beat retrieval practice only when the retrieval items repeat The Learning Scientists, Jun 2026
- Meta-analysis of maths learning: spacing helps (g = 0.28), the retrieval-practice effect is not yet conclusive White Rose Research Online, 2025
- NEET UG 2026 held in 566 cities for about 22.79 lakh candidates Careers360, May 2026
Questions people ask
Is the Feynman technique backed by research?
Not under that name, but its parts are. Explaining a concept in simple words without notes combines self-explanation (moderate utility) with retrieval (high utility). The retrieval part is doing most of the work, so do it with the book closed.
Are flashcards and spaced repetition apps worth it?
Yes, if you use them to recall the answer before flipping. Flashcards are a form of practice testing, and apps such as Anki automate distributed practice. They work best for discrete facts; for problem-solving subjects, past-paper questions are the better "card".
Is rereading ever useful?
A first careful read is necessary, and a quick reread can help after a test shows you got something wrong. What the evidence rules out is rereading as the main revision method, because it raises confidence faster than it raises recall.
How many hours a day should I study?
Dunlosky's review does not set a number, and neither does Cepeda's. The research says how to spend the hours: more of them on recall and spaced review, fewer on rereading. Six well-structured hours with testing will beat ten hours of passive reading.
Does highlighting hurt learning?
Mostly it just fails to help. The review found no consistent benefit. Use it sparingly, to mark what you will later turn into a question.
Sources
- Dunlosky J. et al.: Improving Students' Learning With Effective Learning Techniques, Psychological Science in the Public Interest 14(1):4-58 (2013)
- Cepeda N.J. et al.: Distributed practice in verbal recall tasks: A review and quantitative synthesis, Psychological Bulletin 132(3):354-380 (2006)
- Cepeda N.J. et al.: Spacing effects in learning: a temporal ridgeline of optimal retention, Psychological Science 19(11):1095-1102 (2008)
- Roediger H.L. and Karpicke J.D.: Test-enhanced learning, Psychological Science 17(3):249-255 (2006)
- Karpicke J.D. and Roediger H.L.: The critical importance of retrieval for learning, Science 319:966-968 (2008)
- Brunmair M. and Richter T.: Similarity matters: A meta-analysis of interleaved learning and its moderators, Psychological Bulletin (2019)
- Ebbinghaus H.: Memory, A Contribution to Experimental Psychology, tr. Ruger and Bussenius, Classics in the History of Psychology (1885; English 1913)
- Ministry of Health and Family Welfare: Tele-MANAS
- Rowland C.A.: The effect of testing versus restudy on retention: a meta-analytic review of the testing effect, Psychological Bulletin 140(6):1432-1463 (2014)
- Murray, Horner and Göbel: spacing and retrieval practice in mathematics, meta-analysis (2025)
- The Learning Scientists: Retrieval Practice vs. Worked Examples: When to Use Which (2026)
- Careers360: NEET UG 2026 concludes, 22.79 lakh candidates (2026)
