Spaced repetition, explained properly (and how to schedule it yourself)

Most students meet spaced repetition as a piece of software. That is the wrong way round: it is a finding about memory, and software is just one way to act on it. Understanding the finding matters, because it tells you when to break the rules the software gives you.

The finding

If you study something once and then never see it again, your ability to recall it drops fast at first and then slowly. Hermann Ebbinghaus described this shape in the 1880s by memorising nonsense syllables and testing himself at intervals; it is usually drawn as the forgetting curve.

The useful part is what happens when you review. Each successful review flattens the curve — the same material now decays more slowly than it did the first time. So the same total study time produces very different results depending on how it is arranged:

Approach	Total time	Recall three weeks later
One four-hour block, night before	4 hours	Poor — most of it decayed within days
Four one-hour sessions across three weeks	4 hours	Substantially better

The second row is the whole idea. This is called the spacing effect, and it is one of the most reliably reproduced results in learning research — it holds across ages, subjects, and whether the material is vocabulary, anatomy, or physics problems.

Why it works (two honest explanations)

Nobody has settled this completely, but two mechanisms are usually offered, and both are useful to you as a student.

Retrieval difficulty. A review that is slightly hard strengthens memory more than a review that is easy. Immediately after reading a page, recalling it is trivial, so you learn almost nothing from the attempt. Two days later, the same recall takes real effort — and that effort is the thing that works. This is why re-reading feels productive and mostly isn't: it never lets the memory get weak enough for the retrieval to count.

Encoding variability. Each time you come back, you are in a slightly different state — different mood, different room, different associations available. Every review attaches a new set of retrieval cues to the same fact, so in the exam hall you have more routes to it.

A schedule that actually works

Software schedules do this with an algorithm, but you can get most of the benefit with a fixed ladder. For material you need to hold for a term:

1. Same day as the lecture — a fast pass, 10 minutes, just to catch what you didn't understand while you can still ask.
2. Day 2 — first real recall attempt, closed notes.
3. Day 7.
4. Day 16.
5. Day 35.
6. Then roughly monthly until the exam.

The intervals roughly double. That is not magic; it is a decent approximation of "review just before you would have forgotten it". If a review goes badly — you blank, or you get it wrong — you do not carry on doubling. Drop back to the previous interval and rebuild from there.

Rating your recall honestly

Every spaced-repetition system, paper or software, lives or dies on one thing: whether you tell it the truth. Use three buckets and be strict:

• Forgot — you could not produce the answer, or you produced the wrong one. Also use this when you almost had it. Almost is forgot.
• Hard — you got it, but slowly, or you had to reason it out from something else.
• Easy — it came immediately and completely.

The single most common failure is grading "hard" as "easy" because you recognised the answer once you saw it. Recognition is not recall, and the exam asks for recall. If you find yourself flipping the card fast because the honest rating means more work, the system has stopped being a memory tool and become a mood tool.

What spacing does not fix

This is where students lose time, so it's worth being blunt.

Spacing schedules facts. It does not create understanding. If you memorise "the mitochondrion produces ATP" without knowing what ATP is for, you will pass a definition question and fail anything that asks you to reason. Before a fact goes into a review schedule, it should be a fact you can already explain. Spacing preserves understanding; it does not manufacture it.

Cards that are too big don't work. "Explain the causes of the First World War" is not a flashcard, it is an essay. If a card has five things on it, you will half-recall it forever and never grade it honestly. One card, one fact, one answer. Split the essay question into the eight facts it needs, and keep the essay itself as a practice-writing task.

Volume is not progress. Two hundred cards you review honestly beat two thousand you skim. Every card you make is a small debt you have promised to pay every few weeks for the rest of the term.

Applying it to problem-based subjects

For maths, physics, and anything else where the exam asks you to do rather than state, spacing still applies, but the unit changes. Instead of a fact card, space the problem type:

• A card that says "integration by parts — when do you reach for it, and what do you set as (u)?" is a good card.
• A card containing an entire worked integral is not; you will recognise the answer and learn nothing.

Keep a short error log alongside your cards: every problem you got wrong, one line on why — not "silly mistake" but "dropped the minus when distributing". Review the log on the same ladder as your cards. Most students find three or four causes explain almost everything they get wrong, which is a much shorter list than they expected.

Starting this week

If you take one thing from this: pick the subject you are most behind on, spend 20 minutes turning last week's notes into no more than 30 single-fact questions, and put a first review in your calendar for tomorrow and a second for next Monday. That is the entire method. Everything else — algorithms, apps, statistics — is optimisation on top of a habit you either have or don't.

StudyFlow generates the question set from your own notes and schedules the reviews using confidence ratings you give each card, which removes the tedious half of this. The judgement half — honest grading, small cards, understanding before memorising — is still yours.