What the cognitive-science literature measured, with the papers named and the numbers shown.
Book a Free ConsultationEffective revision is retrieval practice and spaced practice, the only two of 10 techniques Dunlosky and colleagues rated high utility in their 2013 review for Psychological Science in the Public Interest. Students who tested themselves recalled 61% of a passage a week later; students who re-read it recalled 40%, having read it 14.2 times against the tested group's 3.4. Re-reading and highlighting were both rated low utility.
Five psychologists — John Dunlosky, Katherine Rawson, Elizabeth Marsh, Mitchell Nathan and Daniel Willingham — read the experimental literature on 10 study techniques and graded each one for utility: how reliably it works across different ages, abilities, materials and types of exam. Their verdict is the closest thing the field has to a league table, and it is the table almost every revision article alludes to without printing.
| Technique | Utility assessment | What it means in a revision session |
|---|---|---|
| Practice testing | High | Answering questions or recalling from a blank page, without the notes in front of you |
| Distributed practice | High | The same total study time broken into sessions separated by days |
| Elaborative interrogation | Moderate | Asking why a stated fact is true and answering it in your own words |
| Self-explanation | Moderate | Explaining each step of a worked example to yourself before moving on |
| Interleaved practice | Moderate | Mixing problem types within one session instead of blocking by topic |
| Summarisation | Low | Writing condensed summaries of what you have just read |
| Highlighting and underlining | Low | Marking text while reading it |
| The keyword mnemonic | Low | Linking new terms to keyword images, mostly used for vocabulary |
| Imagery use for text learning | Low | Forming mental pictures of what a passage describes |
| Rereading | Low | Going back over notes or a textbook chapter |
Source: Dunlosky et al., Improving Students' Learning With Effective Learning Techniques, Psychological Science in the Public Interest, retrieved 16 August 2026 — read the paper.
The wording of the top tier matters, because it is a claim about generality rather than raw size. Practice testing and distributed practice, the authors write, “received high utility assessments because they benefit learners of different ages and abilities and have been shown to boost students' performance across many criterion tasks and even in educational contexts”. The middle tier is not a criticism of the techniques so much as of the evidence behind them: elaborative interrogation, self-explanation and interleaved practice “fell short of a high utility assessment because the evidence for their efficacy is limited”.
The bottom tier is where most students live. The authors are blunt about it: “Five techniques received a low utility assessment: summarization, highlighting, the keyword mnemonic, imagery use for text learning, and rereading.” Two of those five — highlighting and rereading — are the default revision method in most sixth forms in the country.
Because over a short enough horizon it does work, and that is a measured result rather than a rhetorical concession. Henry Roediger and Jeffrey Karpicke gave students prose passages and then either let them study again or made them take a free-recall test. When the final test came 5 minutes later, the re-studying group won: 81% against 75%. The second experiment sharpened the point. Students who studied four times scored 83%, students who studied three times and tested once scored 78%, and students who studied once and tested three times scored 71%. On that timescale, every extra reading bought another point.
Students also know it. In the same experiment the group that only re-read was the most confident of the three: they “were more confident that they would remember the passage in 1 week than were subjects in the SSST or STTT condition”. They were also the group that remembered the least. Confidence and retention pointed in opposite directions, which is why the method survives every attempt to argue students out of it.
The survey evidence says the argument has not been won. Karpicke, Butler and Roediger asked 177 undergraduates to list what they actually do when they study. Re-reading was “by far the most frequently listed strategy with 84% of students reporting it”, and 55% named it their single most-used method. Retrieval practice was named by 11% of them, and by 1% as their first choice. Flashcards were reported by 40% and practice problems by 43%, and the authors note that even counting both as self-testing, “these percentages are dwarfed by the 84% of students who repeatedly read while studying”.
Roediger and Karpicke ran the same comparison at three delays, and the ordering flips somewhere between five minutes and two days. This is the single most useful table on the page, because the interval that matters to a candidate is the one between revising a topic and sitting the paper — never five minutes.
| Gap before the final test | Re-studied the passage | Took one recall test | Difference |
|---|---|---|---|
| 5 minutes | 81% | 75% | Re-studying ahead |
| 2 days | 54% | 68% | Testing ahead by 14 points |
| 1 week | 42% | 56% | Testing ahead by 14 points |
Source: Roediger and Karpicke, Test-Enhanced Learning, Experiment 1, Cognition and Learning Lab, Purdue University, retrieved 16 August 2026 — read the paper.
One consequence is worth stating in its own sentence. The tested group's score after a week matched the re-studying group's score after two days: taking a single recall test “prevented forgetting of information for an additional 5 days relative to repeated study”. One test, five days of forgetting bought back.
The second experiment is the one to quote at a student who insists more reading is more learning, because it counts the readings. Everyone had the same total study time; what differed was how that time was spent.
| What the group did | Times they read the passage | Recall after 1 week |
|---|---|---|
| Studied four times, no test | 14.2 | 40% |
| Studied three times, tested once | 10.3 | 56% |
| Studied once, tested three times | 3.4 | 61% |
Source: Roediger and Karpicke, Test-Enhanced Learning, Experiment 2 and Table 1, Cognition and Learning Lab, Purdue University, retrieved 16 August 2026.
The group that read the passage a quarter as often remembered half again as much. As the authors put it, the repeated-testing group “recalled much more after a week than did students in the repeated-study condition (61% vs. 40%), even though students in the former condition read the passage only 3.4 times and those in the latter condition read it 14.2 times”. That is the whole case for closing the book, and it is worth more than any amount of advice about study atmosphere. Our GCSE tuition sessions are built around it: the tutor asks, the student produces, and the notes stay shut until the answer is on the page.
Revising hard and scoring the same? The method is usually the problem, not the hours.
A free consultation covers what your child is currently doing in a revision session, where the retrieval is missing, and how to rebuild the week around spaced practice before the next set of mocks.
Book a Free Consultation Message us on WhatsAppEvery revision article says to space it out. Almost none says how far, which makes the advice unusable. Nicholas Cepeda, Edward Vul, Doug Rohrer, John Wixted and Harold Pashler answered it directly: they taught 1354 people a set of facts, brought them back for a review after a gap of anywhere up to three and a half months, then tested them at delays of up to a year. The gap that produced the best final score was not the shortest and not the longest, and it moved with the delay.
| Days until the test | Best gap between the two sessions | Recall gained against no gap |
|---|---|---|
| 7 days | 1 day | 10% |
| 35 days | 11 days | 59% |
| 70 days | 21 days | 111% |
| 350 days | 21 days | 77% |
Source: Cepeda, Vul, Rohrer, Wixted and Pashler, Spacing Effects in Learning: A Temporal Ridgeline of Optimal Retention, Psychological Science, via ERIC, retrieved 16 August 2026 — read the paper.
Read the third row again. For a test 70 days away, revising a topic twice with three weeks between the sittings produced more than twice the recall of revising it twice on the same day — same material, same total minutes, different calendar. The authors summarise the pattern as an optimum gap of “about 20% of the test delay for delays of a few weeks, falling to about 5% when delay was one year”, and they are candid about the implication: “The interaction of gap and test delay implies that many educational practices are likely to be highly inefficient.” Across the study the best gap produced a 64% increase in final recall against no gap at all.
There is a caution buried in the top row. When the test is only a week away, the best available gap was a single day and it bought 10%. Spacing is a technique for the student who starts early; it has very little to give the student who starts on the Friday. That is the honest reason to build a schedule in the autumn rather than the spring, and it is the arithmetic behind our advice on how long GCSE revision actually takes.
Interleaving is the technique students abandon fastest, because a mixed problem set feels worse to do than a blocked one. Doug Rohrer, Robert Dedrick and Sandra Stershic tested it where it counts: 126 seventh-grade students in a Florida middle school, over a 3-month period, taught by their own teachers, with graph and slope problems either blocked by type or shuffled together. Every student got the same problems. Only the order changed. The test was unannounced.
| Delay before the test | Interleaved practice | Blocked practice | Effect size (Cohen's d) |
|---|---|---|---|
| 1 day after review | 80% | 64% | 0.42 |
| 30 days after review | 74% | 42% | 0.79 |
Source: Rohrer, Dedrick and Stershic, Interleaved Practice Improves Mathematics Learning, via ERIC, retrieved 16 August 2026 — read the paper.
The lead holds over the delay rather than closing, which is the reverse of what most study tricks do. Split by problem type at the 30-day test, interleaving produced 84% against 54% on graph problems and 65% against 29% on slope problems. The authors also note that the interleaved group barely forgot anything: “the 30-fold increase in test delay reduced test scores by less than a tenth (from 80% to 74%)”.
The practical translation for a maths or science student is small and specific. A revision set of 20 questions on one topic teaches the method; it does not teach recognising which method a question wants, because that has already been given away by the heading on the worksheet. Mixing three topics into one set puts that decision back. Our maths tutors build mixed sets from past papers for exactly this reason, and the same logic drives the topic ordering in our A-level tuition.
It is worth being straight about the limits, because the confident tone of most revision writing is not supported by the literature it borrows from. The Education Endowment Foundation, the body that funds and reviews classroom trials in England, published a review of cognitive science approaches in the classroom in July 2021, restricted to studies run in real schools. Its conclusion is more cautious than any of the individual experiments above: “Even approaches with indicative evidence of promise like retrieval practice, spaced practice, and the use of worked examples are, as yet, only supported by a few studies that examine their impact in everyday classroom conditions—delivered by teachers over long periods of time.”
On interleaving it goes further: “there are studies with promising results but they have almost exclusively been tested in one subject area (mathematics)”, and “there are serious gaps and limitations in the age ranges, subjects, and learning outcomes studied for most of the strategies we reviewed”. That is worth holding next to the Florida study, which was itself a maths study. Rohrer and colleagues say the same thing from the other direction: their classroom effects were “much smaller than interleaving effects observed in the laboratory”.
None of that overturns the ranking. It changes how firmly it should be stated. Retrieval and spacing are the safest bets a student can make with their time, tested repeatedly and in more than one setting. Interleaving is well evidenced in maths and thinly evidenced elsewhere. Anyone who tells a family that a technique will deliver a specific grade uplift is going beyond what has been measured.
The findings above convert into four decisions, and none of them is about stationery or study playlists.
Set the gap from the exam date, not from habit. With a paper 35 days away, the measured best interval between two sittings of the same topic was 11 days; at 70 days it was 21. A topic covered in the first week of a half term and revisited three weeks later has done more work than the same topic covered twice in one afternoon, and it has cost the same number of minutes.
Make the first pass a retrieval pass. Close the book and write what you can before opening it. The 61% against 40% comparison held even though the tested group had read the material a quarter as often, so a retrieval-first session is not a slower route to the same place.
Mix the set. For quantitative subjects, build practice sets that draw from three or four topics at once. Expect it to feel harder and score lower during practice; the 74% against 42% result appeared 30 days later, not on the day.
Stop marking your own confidence. The students most sure they would remember were the ones who remembered least. The only reliable check is a closed-book attempt marked against a mark scheme, which is also why a mock is worth more than an evening of notes. Families preparing for the final year often find the sequencing in our Year 11 revision guide and the subject-specific plan in our GCSE Biology revision guide easier to act on than a general timetable.
The techniques rated low utility are not forbidden. Highlighting a specification to plan the week is fine; it is only a problem when it is mistaken for the revision itself. The distinction the evidence draws is between arranging material and retrieving it, and only the second one is learning.
Practice testing, with distributed practice alongside it. Those are the only two of the 10 techniques Dunlosky and colleagues assessed that earned a high utility rating, awarded because they “benefit learners of different ages and abilities and have been shown to boost students' performance across many criterion tasks and even in educational contexts”. In Roediger and Karpicke's second experiment, students who tested themselves recalled 61% of a passage a week later against 40% for students who only re-read. Elaborative interrogation, self-explanation and interleaved practice sit one tier below, rated moderate.
For an exam in the next few hours, re-reading is defensible: on a test taken 5 minutes after studying, the re-reading group scored 81% against 75% for the tested group, and in the second experiment scores rose with every extra reading, from 71% up to 83%. Past about two days that advantage has gone and reversed. Dunlosky and colleagues place rereading in the low utility tier alongside highlighting, summarisation, the keyword mnemonic and imagery use for text learning. Use it to refresh, never as the main method.
Set it from the exam date. Cepeda and colleagues taught 1354 people a fact set and varied the gap between the first sitting and the review. For a test 7 days away the best gap was 1 day; at 35 days it was 11 days; at 70 days and at 350 days it was 21 days. The payoff rose sharply with the delay: 10% better recall at the shortest interval, 59% at 35 days and 111% at 70 days, compared with doing both sessions on the same day.
Because it puts back a decision that a blocked worksheet removes. When every question on a page uses the same method, the method has been supplied by the heading; when the questions are shuffled, the student has to identify the type first. That extra step is the point. Rohrer, Dedrick and Stershic found interleaved practice produced 80% against 64% on a test 1 day after review, and 74% against 42% after 30 days. The authors report that the advantage does not diminish over the delay, and describe it as perhaps growing larger — the interaction itself is not statistically significant, so the safe reading is that interleaving holds its lead rather than losing it.
Not by how confident it makes you feel. In the Roediger and Karpicke experiment the group that only re-read rated themselves most likely to remember the passage in a week, and then recalled 40% of it while the tested group recalled 61%. The usable check is a closed-book attempt marked against a mark scheme, at least a day after the material was covered. If the score is high with the notes shut and a day's delay behind it, the method is working.
Retrieval and spacing, yes; the rest with caution. The Education Endowment Foundation's July 2021 review of classroom studies concluded that retrieval practice and spaced practice are “only supported by a few studies that examine their impact in everyday classroom conditions”, and that interleaving has “almost exclusively been tested in one subject area (mathematics)”. Rohrer and colleagues agree that their classroom effects were much smaller than laboratory ones. Treat the ranking as a guide to where to spend time, not as a guarantee of a grade.
Only if the answer is produced before it is seen. Karpicke, Butler and Roediger found 40% of students reported using flashcards and 43% reported answering practice problems, but noted that a student can read a question and copy the answer without ever attempting recall. The same survey found 84% of students re-read while studying and only 11% of 177 described practising retrieval. A flashcard used properly is retrieval; a flashcard flipped quickly is rereading with extra steps.
A tutor supplies the two things a student cannot easily do alone: mark a closed-book attempt honestly, and hold a spacing schedule across a term when nothing is due in. Our specialist subject tutors run sessions retrieval-first, build mixed problem sets from past papers, and revisit each topic on a planned gap rather than when it next appears on a timetable. Leading Tuition is rated 4.8 out of 5 on Trustpilot.
Reviewed August 2026: every utility rating, percentage and effect size on this page was re-read and checked against the published papers and reports named beside each table, and no figure is carried over from a secondary summary
Retrieval-first sessions, mixed practice sets and a spacing schedule set by the exam date, run by specialist subject tutors.
Rated 4.8 out of 5 on Trustpilot. See our GCSE tuition and A-level tuition.
Book a Free Consultation