Why timing
beats repeating
Most students review material once, maybe twice, then move on — and a week later most of it is gone. Spaced repetition doesn't add more review time. It times each review to land right before you'd otherwise forget, so each one costs less and sticks longer.
The forgetting curve
In the 1880s, Hermann Ebbinghaus measured how quickly he forgot lists of nonsense syllables and found that forgetting follows a predictable curve: steep at first, then flattening out. Most of what you forget, you forget within the first day or two — after that, the rate of loss slows down considerably.
The implication is easy to miss: a review that happens too early is mostly wasted, because you still remember almost everything and the review feels pointless. A review that happens too late requires relearning from scratch, which is expensive and demoralising. There's a narrow window in between where the review actually does work.
Retention after a single exposure, with no review: ~100% immediately, ~58% after 20 minutes, ~44% after 1 hour, ~26% after 31 days.
Each successful retrieval at the right moment resets the curve and makes it flatten further next time — which is why the gaps between reviews can grow over time without losing the memory.
The useful window
The useful window sits right at the edge of forgetting — where recall is effortful but still possible. That effort is the point: a review that requires real work to retrieve strengthens the memory far more than a review where the answer is still sitting in working memory.
This is also why a spacing system and active recall are usually discussed together: spacing tells you when to review, and active recall tells you how — by retrieving the information yourself rather than re-reading it. Spacing without retrieval is just a re-reading schedule; retrieval without spacing is cramming with extra structure.
Try it yourself
Drag the slider below to pick a day since you first learned something, and watch where it lands on the forgetting curve.
Building
a deck that works
A spacing system is only as good as the items inside it. Badly written items make every review slower and less useful — well-written items make the whole system run on autopilot. This module covers what makes an item good, and the schedule that decides when each one comes back.
What makes a good item
A good item is atomic — it tests one fact or one idea, not three. "Explain the causes of World War I" is an essay prompt, not a review item; it can't be answered cold in a few seconds and can't be marked clearly right or wrong. "Name one long-term cause of World War I" is atomic.
A good item is also phrased as retrieval, not recognition. If the cue contains most of the answer, you're not retrieving anything — you're filling in a blank you can already see. And it should be unambiguous: if you could reasonably give three different correct answers to the same cue, the item is testing your guess at what the card-maker meant, not your knowledge.
Bad: "The forgetting curve, discovered by Ebb—" → tests nothing, the answer is in the cue. Also: "Describe spaced repetition" → too broad, no single correct answer, can't be marked cold.
Good: "Who ran the original forgetting-curve experiments?" → Ebbinghaus. One fact, one answer, no leakage. "What happens to a Leitner card when you get it wrong?" → it goes back to Box 1. Atomic, unambiguous, retrieval-based.
The Leitner system
The classic low-tech implementation of spaced repetition is the Leitner system: physical flashcards sorted into boxes numbered 1 through 5. Every card starts in Box 1, which is reviewed daily. A correct answer moves the card to the next box up, which is reviewed less often. A wrong answer sends it straight back to Box 1, no matter which box it was in. Digital flashcard apps automate this with the same underlying logic — a shoebox and index cards work identically.
If you'd rather think in calendar days than boxes, the same idea maps to an expanding schedule. Each successful recall roughly doubles the gap before the next review; a failed recall resets the item back to Day 1.
Try it yourself
Six cards, all starting in Box 1. Reveal each one, decide honestly whether you had it, and watch where it lands.
Running
the review session
A well-built deck can still be wasted by a badly run session. The mechanics of how you review — the order, the honesty of each attempt, what you do with a missed day — determine whether the schedule actually does its job.
The session itself
A review session should take 15–30 minutes, not hours. If it's taking longer, you're either adding new items faster than you can absorb them, or your items aren't atomic enough. Go through the items that are due — in any order — and for each one, try to produce the answer before you look. A guess counts as an attempt; flipping the card immediately does not.
This single rule — answer first, check second — is the difference between a spacing system and an expensive way to re-read your notes. If you read the question and immediately flip to the answer "to refresh your memory," you're not retrieving anything.
18 items are due today. For each one: read the cue, say the answer out loud or write it down, then flip. 12 come back correct on the first try — those get promoted to their next interval. 4 are partially right or slow — those are marked for a same-session re-run at the end. 2 are blank — those reset to Day 1.
After the first pass, the 4 "shaky" items get re-run once more. Total time: about 20 minutes for 18 items, most of it on the 6 that needed extra attention.
Try it yourself
Eight items from a mixed deck. For each one, commit to an answer in your head, reveal it, then mark it honestly — that's pass one. Anything you weren't sure about comes back for a second pass.
What goes wrong
Overload
& exam compression
Every spacing system eventually meets two pressures: a review queue that's grown faster than you can clear it, and an exam date that arrives before the long intervals have finished playing out. Both have a fix — and neither fix is "do more."
When the backlog grows
A growing backlog almost always means new items were added faster than the daily review queue could absorb them. The fix isn't to power through a 90-minute session — it's to stop adding new items until the backlog shrinks back to a normal size, then resume at a slower pace.
When you do sit down to clear a backlog, work through overdue items first, oldest first. These are the items most likely to have decayed the most — and clearing them restores the schedule's integrity faster than working through today's items and leaving the backlog for "later."
Try it yourself
Set how many new items you add per day against how many you can actually review, and watch what happens to your queue over two weeks.
Compressing under exam pressure
In the final week before an exam, the long-interval logic that makes spacing efficient stops being the priority — you need maximum retention right now, not the most efficient use of review time over the next two months. This is the one situation where the normal rules bend. The schedule below highlights which phase matches the slider you just set.