Spaced Repetition for GCE: Set Your Review Dates From Your Exam Date

This post is the hub for the revision posts on this site. It covers what the studies found, how to turn them into dates, the card box, the 12-week plan and what to do if you start late. Timetabling, formula recall, mock results and study groups each have their own post, linked where they apply.

What two experiments show about testing yourself

Rereading feels productive because the page looks familiar. Two experiments put numbers on the gap between that feeling and what you can recall later.

In Roediger and Karpicke (2006, Psychological Science), Washington University students studied prose passages. One group then took a written recall test with no feedback. The other group studied the passage again. Five minutes later, the restudy group did better (81% of idea units against 75%). After two days the tested group led, 68% to 54%. After one week it was 56% against 42%. In a second experiment, students who studied once and then took three recall tests remembered 61% after a week. Students who read the passage about 14 times remembered 40%. Source: Roediger and Karpicke 2006 (PDF), opened 15 October 2026.

The first result matters for cramming. Restudying wins if the test is in five minutes. Testing wins if the test is days away.

Karpicke and Roediger (2008, Science) tested a habit most students have: learn an item, then stop practising it. College students learned 40 Swahili-English word pairs. Once a pair had been recalled correctly, it was either kept in the test rounds or dropped from them. One week later, students who kept all pairs in the test rounds recalled about 80%. Students who dropped pairs from testing after one correct answer recalled 36% (still restudied, no longer tested) and 33% (dropped from both studying and testing). Before the final test, all groups predicted they would remember about half. Source: Karpicke and Roediger 2008 (PDF), opened 15 October 2026.

For your cards, this means one correct answer does not retire a card. It moves to a longer interval and stays in the box until the exam.

Dunlosky and colleagues (2013, Psychological Science in the Public Interest) reviewed ten common study techniques. Practice testing and distributed practice received the top rating, high utility. Rereading and highlighting were among five rated low. Source: Dunlosky et al. 2013 (PDF), opened 15 October 2026. The same review rated interleaving moderate, which is why that technique has its own post: How to Memorise Formulas You Keep Forgetting.

All of these were studies of volunteers, mostly university students, using prose, word lists or trivia. No study of this kind used GCE papers. Treat the results as strong evidence about how memory behaves, and test them against your own scores.

How big should the gap be?

Cepeda and colleagues (2006, Psychological Bulletin) combined 839 assessments from 317 experiments. The best gap between study sessions was longer when the final test was further away, and a gap that was too long lowered recall again. Source: Cepeda et al. 2006 (PDF), opened 15 October 2026. The same paper found that expanding gaps (1 day, then 3, then 7) were sometimes better than equal gaps and sometimes no better. The fixed ladder in most blog posts, including the earlier version of this one, is a convention and not a research result.

Cepeda and colleagues (2008, Psychological Science) measured the effect directly. Over 1,350 people learned facts, had one review after a gap that ranged from zero to 105 days (3.5 months), and were tested up to a year later. For recall, the best gaps they tested were 1 day for a test 7 days later, 11 days for a test at 35 days, and 21 days for tests at 70 and 350 days. Smoothed estimates were roughly 3, 8, 12 and 27 days. Compared with reviewing immediately, the best gap raised final recall by 10% (7-day test), 59% (35 days), 111% (70 days) and 77% (350 days). Source: Cepeda et al. 2008 (PDF), opened 15 October 2026.

Turn that into a rule you can use. Take the days left until the paper and read the first review gap from this table.

The ranges run from the smoothed estimate to the best gap actually tested. The study used one review session, so for later reviews the table is an extrapolation, and the rule below is my working method, not a tested one. The studies measured the delay from the review, not from first learning; if you count from first learning, shorten each gap by about a fifth. The worked schedule and late-start rows carry the same caveat.

A worked schedule

This example is illustrative. Say you learn the topic “enzymes” on day 0 and your paper is 70 days away.

1. Day 0: learn it and write the cards (see the box below).
2. Day 0, evening: one quick test of the cards while you still remember them. This stops the first forgetting, and it is cheap.
3. Day 12 to 14: first proper review, with no notes open. Pass the cards you answer correctly to the next box. Return the ones you miss to the first box.
4. About two weeks later, day 26 to 28: second review of the cards that passed.
5. After that, one review every three to four weeks, so that every card is seen at least twice more before the paper.

Now compare a candidate with 21 days left who learns the same topic. The table puts the first review at about 5 to 7 days, not 12. If you keep the 70-day rhythm with 21 days left, the first review lands on day 12 to 14 and later reviews fall after the paper.

Rule: the further away the paper, the longer you can wait before the first review. As the paper approaches, shorten every gap. Over the last week, review daily.

The card box (no power, no data)

A card box needs paper and five envelopes or folded sheets. Label the envelopes 1 to 5. Each envelope has a due date that you set from the rule above, and that moves closer to the exam as it approaches.

1. New and missed cards go in envelope 1. You open envelope 1 at the start of every session.
2. Open envelope 2 when its date arrives, envelope 3 at the next date, and so on.
3. Answer from memory first, then check. Right: the card moves up one envelope. Wrong: it goes back to envelope 1.
4. A card in envelope 5 is not retired. It comes out for one final review in the last fortnight.

Write one question per card. “Explain photosynthesis” is an essay. “Name the two products of the light-dependent reaction” is a card. A subject usually needs a few hundred cards, though I have found no study that sets a number, so use your syllabus topic list to decide what to cover. If you cannot say what a card tests, make it smaller.

Phone apps do the same bookkeeping. They save time once you pass a few hundred cards. Paper works when power or data fail, which is covered in How to Revise Without Reliable Electricity or Data.

How to write cards for calculation subjects is in the formulas post. For essay subjects, build one card per piece of evidence, one per argument and one per counterargument. For languages, test into the target language, because producing the word is harder than recognising it.

The 12-week plan

This assumes about 90 minutes on weekdays and a longer weekend session. Scale it. The shape matters more than the minutes. How to fit the sessions into your real week is covered in How to Build a Revision Timetable That Survives Real Life.

Attempt one past paper per subject in week 3. The score will be low. The paper shows you how examiners phrase questions and where the marks sit, while you still have eight weeks to act on it. What to do with the marks you lose is in Mock Exams: What To Do With the Result.

Sleep

Diekelmann and Born (2010, Nature Reviews Neuroscience) review evidence that sleep after learning strengthens memories, with slow-wave sleep and REM sleep playing different roles. Source: abstract via Europe PMC, opened 15 October 2026. That review is about memory in general. I did not find a study of exam-week sleep schedules for GCE candidates. The practical conclusion is modest: in the last fortnight, do not trade night-time sleep for extra reading.

If you start late

Use the gap table. Fewer weeks does not mean abandoning the method. It means shorter gaps.

If you miss a day, carry on. If you miss a week, the catch-up rule is in the timetable post, not here.

Three ways the system fails

1. Checking the answer too early. Wait ten seconds before you look.
2. Marking yourself generously. If you half remember it, it is wrong.
3. Making cards and not reviewing them. Writing cards feels like work. Reviewing them is the part the studies tested.