How long to wait before reviewing: the gap is a ratio, not a fixed number of days

The gap that works is a share of the time you need the material

The spacing effect is old and well established: material returned to across separated sessions is held longer than the same material studied in one block. What took much longer to pin down was the gap itself. Asking how many days to leave between reviews has no single answer, because the question is missing a variable. Cepeda and colleagues, working across a wide range of delays, found that the gap producing the most retention is not fixed at all. It moves with the retention interval, which is the distance between now and the moment you actually need to recall the material. The relationship they measured puts the useful gap at roughly a tenth to a fifth of that distance. That single sentence rearranges most revision timetables, because timetables are usually built from the calendar of the course rather than from the date of the test. The practical consequence is that the same deck of cards deserves different treatment depending on why you are learning it, and that a schedule copied from someone sitting a different exam is likely to be wrong in a predictable direction.

A test in one week puts the first return about a day out

If the exam is seven days away, the proportion lands the first gap somewhere near a day, and the following gaps only slightly wider. This is the case where cramming looks most defensible and still is not, because the comparison that matters is not effort but placement. Four separated half hours across the week hold more at the end of it than one continuous two hour session the night before, even though the total time is identical. The short horizon also forgives imprecision. At this scale the difference between a gap of one day and a gap of two is small, so the decision worth getting right is simply whether the material is met again at all before the exam rather than exactly when. If you are working to a deadline this tight, spend your planning effort on making sure every topic gets a second pass, not on tuning the interval.

A test in one month moves the gap to somewhere near a week

Stretch the horizon to thirty days and the same proportion produces first gaps of roughly three to six days. This is the range where most students go wrong in the opposite direction from cramming. Daily review of everything feels diligent, and it is what a habit tracker rewards, but at this horizon it packs the returns far tighter than the evidence supports and it fills the month with sessions that mostly confirm what you already hold. Spreading the same number of returns across wider gaps covers more topics with the same hours. It also feels worse while you do it, because a gap long enough to be useful is long enough that recall is effortful when it arrives. That discomfort is not a sign the schedule is failing. It is the same signal that separates retrieval from rereading, and it points the same way.

Material you need for a year sits on gaps measured in weeks

For a subject you will carry into next year, a professional registration, or a language you intend to keep, the retention interval is measured in months and the proportion pushes the gaps out to weeks and eventually months. Almost nobody schedules this by hand, and that is the honest reason review software exists. The arithmetic is simple but the bookkeeping is not, because every card is at a different point in its own sequence and the whole set has to be re-sorted every day. This is also where the long horizon changes what counts as a good session. Covering forty cards you half remember does more for a year long horizon than perfecting four, since the cards approaching the edge of recall are the ones the gap is meant to catch. Our spaced repetition works on exactly this bookkeeping problem.

The first day is where most of the loss happens

The reason the first gap matters more than the ones after it is the shape of forgetting. When Murre and Dros repeated Ebbinghaus's original experiment more than a century later, the curve came back with the same shape. The steepest loss arrives in the first day, and roughly two thirds of what was learned is gone by the end of it. After that the decline flattens out considerably, and everything following is a slower slide. This is why a review placed on day one is worth more than the same review placed on day four, and why the common instinct to leave the first pass until the weekend gives away the part of the curve where intervention costs least. It also explains a familiar experience. Material that felt secure when you closed the book and vanished by the next lecture did not fail because you learned it badly. It failed on schedule, and the schedule is the thing to plan around.

The gap only pays if the return is a retrieval

Spacing and retrieval are two separate findings and they are easy to confuse, because a spaced schedule of the wrong activity produces very little. The comparison is direct. In Roediger and Karpicke's experiments and the follow up published in Science, the delayed test told a clear story. Students who read a passage and then practised recalling it closed book reached about eighty percent, against roughly a third for students who reread the passage repeatedly. The rereaders predicted the opposite about themselves, which is the part worth remembering: fluency while reading is a poor guide to whether anything will be there next week. Put the two findings together and the instruction is short. Space the returns by a share of your horizon, and make each return an attempt to produce the answer from memory before checking it. A spaced schedule of rereading spends the calendar correctly and the hour badly.

What this research does not settle

Three limits are worth stating, because they bound how far the numbers above can be pushed. First, the proportion is a broad regularity rather than a formula, and the studies behind it used defined material tested at set delays. Your syllabus is not that tidy, so treat a tenth to a fifth as a range to aim at rather than a figure to calculate to the day. Second, most of this work measured recall of discrete items, and a mathematics course or a design portfolio asks for something the experiments did not test. Spacing appears to help there too, but the size of the benefit is not established by these studies and anyone quoting a precise figure for it is going beyond the evidence. Third, none of it removes the need to understand the material first. Spacing governs how long something stays available once it is in. It does not put it in, and a well spaced schedule of material you never understood produces a well spaced record of failing to recall it.

What to photograph, and what happens next

The pages worth turning into cards are the ones where you will need a specific answer under time pressure: definitions, formulae, dates, labelled diagrams, vocabulary. Photograph the page rather than copying it out, because that hour is worth more spent on the returns. From there the intervals are the software's problem rather than yours, and the cards you miss come back sooner than the cards you get. Related reading: building a review schedule, how many cards a day, and flashcards by subject.

By Mert Erarslan (Exam prep and subject routes)

Last updated: 2026-08-15 · Written with AI assistance and reviewed before publishing.