Why Repetition Improves Vocabulary

Forgetting is steep at first, then flat

Memory for something newly learned drops sharply in the first day or two, then the rate of loss slows considerably. Most of what you forget, you forget early.

This shape has a direct practical consequence. A review placed in the steep part of the curve — when you are actively losing the word — does far more than a review placed while the word is still fresh, because there is something to recover.

Reviewing immediately feels productive precisely because it is easy, and it is easy because nothing has been forgotten yet. Ease and effectiveness are inversely related here, which is why intuition points the wrong way.

Why massed repetition fails

Going over a list twenty times in one evening builds a strong impression that fades fast. The word is held in short-term storage throughout the session, so almost none of the repetitions require genuine retrieval — you are re-reading something that never left.

The test of whether a repetition counts is whether you had to work to produce the word. If it was already available, the repetition did very little.

This also explains why cramming the night before an assessment can work for the assessment and leave nothing behind a fortnight later. The information never entered a state where it needed to be retrieved.

Intervals that work

You do not need precise scheduling. An approximate expanding sequence captures nearly all the benefit:

• Same day — brief, just to confirm the word is in
• Next day — the most valuable single review
• Three days later
• One week later
• Two to three weeks later
• One month later — after which most words are stable

Six touches over about six weeks is enough for most vocabulary. Compare that with twenty repetitions in one evening, which is more total effort for a worse result.

If a review goes badly, move the word back down the sequence rather than forward. If it goes well twice in a row, retire it.

Doing it without an app

Spaced repetition software exists and works, but it adds a maintenance burden that puts a lot of people off. A paper version that does the same job:

Use five index cards or five sections of a notebook, labelled daily, 3-day, weekly, fortnightly, monthly. New words go in daily. A word you get right moves one box up. A word you get wrong drops back to daily.

Each session, work the boxes that are due. Most days that is daily plus one other, which takes a few minutes. The system self-balances — hard words cycle frequently, easy words disappear upward.

Vary the question, not just the timing

Repeating the same prompt six times teaches recall of that prompt. Varying it teaches the word.

Rotate between:

• Definition → word (the hardest and most useful direction)
• Word → definition
• Scrambled letters → word, which you can generate with an anagram tool
• Sentence with a gap → word
• Related words → the target, using a contains finder to build the group

The scrambled version is worth including even though it looks like a game, because it tests form while the others test meaning, and both need to be solid.

Repetition through play

Regular play delivers a rough version of all this automatically. Common words recur constantly and consolidate quickly; rare words recur occasionally, which happens to approximate a good interval for words you do not need often.

What play does not do is target your gaps. It repeats whatever the game happens to throw at you, which over-serves words you already know. That is the case for a small deliberate list alongside play rather than instead of it.

A practical combination: play normally, note three uncertain words per session, and run those through the box system. Everything else is handled by the game.

Two things that break it

Reviewing too early. If it feels easy, the interval was too short. A small amount of difficulty is the signal that the review is working.

Letting the list grow. Adding words faster than you retire them produces a backlog, the backlog produces sessions that are too long, and long sessions stop happening. Retire aggressively — two correct reviews and the word leaves.