How Does Spaced Repetition Work? Mechanism and Examples

Hermann Ebbinghaus tested his own memory in the 1880s. Alone in a Berlin apartment, he memorised lists of nonsense syllables, then tracked precisely how fast he forgot them. The result was a curve — steep at first, then flattening — mapping forgetting against time. Without rehearsal, roughly half of what you learn disappears within a day.

That curve still drives every serious revision method used today. Spaced repetition is the most direct response to it.

The forgetting curve

Ebbinghaus's core finding was not just that we forget, but that forgetting follows a predictable pattern. Memory strength drops quickly in the hours after learning, then more slowly. Review the material at the right moment — just before the memory drops below a usable threshold — and you reset the curve. More importantly, each successful recall makes the next forgetting event slower.

The practical implication is that timing reviews matters as much as the number of reviews. Two sessions at the right intervals outperform six sessions clustered on the same night.

What happens in the brain

When you retrieve a piece of information successfully, two things occur:

1. The neural pathways associated with that fact become more efficient — the memory trace is strengthened.
2. The forgetting curve for that item flattens — the memory now takes longer to drop below the threshold again.

Space those reviews correctly, and each one pushes the next review further into the future. The first review might come after one day, the second after three, the third after a week, the fourth after a fortnight. By the fifth review, the material can often sit for a month without loss.

This expanding interval is the whole mechanism. Spaced repetition feels harder than re-reading because you are attempting recall when the memory is genuinely weak — which is exactly when strengthening it produces the most benefit.

Robert Bjork at UCLA calls this desirable difficulty: study conditions that feel less productive in the moment often produce the strongest long-term retention. A session where you struggle to remember something is a session where memory consolidation is actively happening.

A concrete example: French vocabulary

Say you learn 20 French words on a Monday.

• Monday evening — self-test; 14/20 correct
• Wednesday — review the 6 missed words, spot-check 4 from the correct group; now 18/20
• Sunday — full test of all 20; 20/20, with 2 still marginal and flagged
• Following Sunday — flagged words only; both correct
• Five weeks later — spot-check 5 random words from the batch; still solid

Total time: roughly 35 minutes spread across four weeks. A student who instead re-read the same list for 35 minutes in one sitting would score well on a test two days later and fail most of the same words a month on. Same effort. Different distribution.

Three subject examples

Language vocabulary is the clearest application. Each word either comes to mind correctly or does not — the binary response fits perfectly into a card-based system. A Leitner box (three or five physical compartments, cards moving forward when correct and back to box 1 when wrong) implements spaced repetition with no software required. A student working through French GCSE vocabulary of 600 words can sort those words into review groups over a term and revise each word only when the interval schedule demands it.

History dates and facts respond similarly, with one important caveat: students frequently confuse recognition with recall. Seeing a date on a timeline and thinking "I remember that" is recognition. Writing the date from a cue — "What year did the Treaty of Versailles end?" — is recall. Spaced repetition cards force the second. Recognition produces false confidence; recall produces durable memory.

Maths problem types work well for procedural knowledge — completing the square, implicit differentiation, the sine rule applied to non-right-angled triangles. The card structure changes slightly: one side shows the problem type and initial setup, the other shows the method steps. A student building an A-level maths revision timetable can categorise problem types from past papers and run them through spaced intervals alongside formula recall, so both procedural steps and the conditions for applying them get distributed practice.

Why students avoid it

Two patterns emerge consistently in the research. First, studying material already partially known feels productive. Second, studying material recently forgotten feels uncomfortable. Both reactions push students toward blocked practice — working through one topic thoroughly before moving to the next — which produces higher performance during revision and lower retention one month later.

A student who re-reads the same chapter three nights before an exam and scores 85 percent on a practice test is experiencing strong performance and weak learning. A student who attempts the same practice test after 10 days without reviewing and scores 62 percent, then reviews targeted gaps, is experiencing the reverse. The exam at the end of the term will reveal which approach held.

Implementing without an app

A physical Leitner box uses five compartments. New cards start in box 1. Review box 1 daily, box 2 every other day, box 3 weekly, box 4 fortnightly, box 5 monthly. A correct answer moves a card up one box; a wrong answer returns it to box 1. The cards hardest to remember accumulate in box 1 and receive daily attention; cards you know well reach box 5 and appear only once a month. Study time distributes exactly where it is needed.

For students managing multiple subjects in a 12-week GCSE revision plan, a spreadsheet tracking last-correct-retrieval date and next-review date per topic achieves the same result across subjects.

Where software adds value

Anki, Quizlet, and AI-powered platforms automate the interval calculation across hundreds of cards simultaneously. An algorithm tracks every response and surfaces each card at its individual optimal interval. A student with 300 cards across five GCSE subjects does not manage which card is due; the software does.

AI tutoring platforms take this further by building spaced retrieval into session design — the system tracks which concepts were tested in previous sessions and resurfaces them at the right intervals, without students needing to create or manage their own card decks. For younger learners in particular, this removes the organisational overhead that makes manual spaced repetition difficult to sustain beyond the first week.

What spaced repetition does not replace

Two limits matter here.

Spaced repetition handles retrievable facts — discrete answers testable with a clear correct/incorrect outcome. It does not produce the kind of deep processing needed for genuine conceptual understanding. A student who can recall that mitosis produces two genetically identical daughter cells has a fact. A student who can explain why that matters for tissue repair has understanding. Spaced repetition supports the first; techniques like elaborative interrogation and practice testing under exam conditions build the second. The two work well together, and effective study methods should combine both.

Starting spaced repetition the night before an exam is also not spaced repetition — the intervals need time to expand. Benefits accumulate over weeks. Starting four to six weeks before an exam, running 15-minute sessions daily, and pairing retrieval practice with active problem-solving produces more durable results than any revision strategy that concentrates effort in the final 48 hours.

That is the mechanism: intervals timed to the forgetting curve, retrieval strengthened under pressure, expanding gaps that concentrate effort where it is genuinely needed. Most students who commit to the approach for two weeks do not go back to re-reading.