Spaced repetition for language learning: how it works

What spaced repetition is

Spaced repetition is a review method: instead of studying a word many times in one sitting, you review it at increasing intervals, each one timed for the point where you’re about to forget it. Each successful review makes the memory more durable, so the next gap can be longer.

In practice it runs on flashcards. A program tracks every card, shows you the ones that are due today, and schedules each card’s next review from your answer. A card you remember comes back later; a card you forget comes back soon. Over time, the words you know well take up almost none of your study time, and your effort goes to the words you’re still learning.

The forgetting curve and the spacing effect

The idea goes back to Hermann Ebbinghaus, who in 1885 published experiments in which he memorized lists of nonsense syllables and measured how quickly he forgot them. His results gave us the forgetting curve: memory fades fast at first, then more slowly. He also found that reviewing material again before it was fully forgotten made it fade more slowly the next time.

More than a century of research has confirmed the core finding, now called the spacing effect: the same amount of study produces better long-term memory when it’s spread out than when it’s massed together. A large meta-analysis by Nicholas Cepeda and colleagues (2006) pulled together hundreds of comparisons and found the effect again and again.

A second finding makes flashcards especially effective. Henry Roediger and Jeffrey Karpicke (2006) showed that actively retrieving information, testing yourself, produces better long-term retention than restudying the same material. Seeing the front of a card and trying to recall the back is retrieval practice. Spaced repetition combines the two effects: retrieval, at spaced intervals.

How the algorithms work

Every spaced-repetition system answers one question: when should this card come back? Three approaches are worth knowing.

Leitner boxes. In the 1970s, Sebastian Leitner described a system of physical boxes. New cards start in box 1, which you review often. Answer a card correctly and it moves to the next box, which you review less often; miss it and it goes back to box 1. It’s simple and transparent, and it works.

SM-2. In the late 1980s Piotr Wozniak developed SuperMemo, and its SM-2 algorithm became the basis of many flashcard apps. Each card gets an “ease” value. You grade your answer, the grade adjusts the ease, and the next interval is the last interval multiplied by the ease, so easy cards spread out quickly and hard ones slowly.

FSRS. The Free Spaced Repetition Scheduler is a newer, open-source algorithm, offered in Anki since 2023. It models each card’s memory stability and difficulty from your full review history, then schedules the next review to hit a retention rate you choose, such as 90%.

The difference between them matters less than it seems. All three exploit the same two effects, and any of them beats rereading a word list.

An example schedule

Here’s what a simple ladder looks like for one word you keep remembering. The intervals are those of LingunAI’s default pattern; the later steps show what an extended pattern adds.

If you forget the word at any point, it goes back to the start and climbs again. Words that are genuinely hard for you get more reviews automatically, without you having to track them.

What makes a good language flashcard

The best cards are words you met in something you read or heard. They come with a context your memory can hold on to, and they’ve already proved worth learning by turning up in real use.

• One word or phrase, one meaning. If a word has several meanings, make a card for the one you met.
• Include an example sentence. A sentence shows how the word behaves: its grammar, its usual partners, its register.
• Keep useful phrases together. “Make a decision”, “fall asleep”, “on purpose”: learn collocations as one unit.
• Recall in the useful direction. Recognizing a foreign word is easier than producing it. For words you want to use, test yourself on producing them.
• Skip rare words. If you’re unlikely to meet a word again, it doesn’t deserve a card yet.

How many new cards a day, and avoiding pile-ups

Every new card creates a stream of future reviews, so your daily intake decides your future workload. On the default 1, 3, 7, 15 ladder with the last step repeating, a word you keep remembering comes back about 26 times in its first year. On the extended pattern in the table above, it comes back seven times.

That difference compounds. Add 10 new cards a day for a year and you’ll have about 3,650 cards. If every card keeps coming back every 15 days, the mature ones alone generate well over 200 reviews a day. With long late intervals, the load stays a fraction of that.

Three rules keep reviews manageable:

1. Start with 10 to 20 new cards a day, and adjust to the time you have.
2. Lengthen the late intervals once a deck holds a few hundred cards you know well.
3. When reviews pile up, stop adding. Clear the backlog before taking on new words.

A ten-minute daily routine

Spaced repetition only works if you show up, so make the habit small enough to keep:

1. Review what’s due first. Due cards are the ones about to slip, so they come before anything new.
2. Answer honestly. If you hesitated or half-remembered, count it as forgotten. A generous answer pushes the card further out than your memory can carry it.
3. Read for a few minutes. Meeting reviewed words in a new text is the best reinforcement there is, a kind of spaced repetition that happens on its own.
4. Add a handful of new words from what you just read, then stop.

Ten minutes a day, every day, beats an hour on Sunday, for exactly the reasons Ebbinghaus found.

Common mistakes

• Adding too many cards. The most common reason people quit. A sustainable pace beats an ambitious one.
• Cards without context. Isolated words from frequency lists are harder to remember and less useful than words from texts you’ve read.
• Recognition only. If you only ever recall the meaning of a foreign word, producing it in speech will still be hard.
• Cramming before a test. It works for a day and fades within a week. Spacing is the point.
• Abandoning after a missed week. A backlog is normal. Clear it gradually and keep going.

How spaced repetition works in LingunAI

LingunAI’s flashcards use a Leitner-style ladder with two buttons: “Remembered” moves a card up one step, “Forgot” sends it back to the first. The default pattern brings a card back after 1, 3, 7 and then 15 days, and the last step repeats. You can design your own patterns of up to 20 steps, each from 1 to 3,650 days, set one as your default, or give each deck its own. For a deck you’ll keep for years, a pattern such as 1, 3, 7, 15, 30, 60, 120, 240 keeps the review load light.

Cards come due at your local midnight. Most start inside something you read: select a word in a story or article and it becomes a card in one step. Free quizzes and AI Review add retrieval practice in different forms. If you’re coming from a dedicated flashcard app, LingunAI vs Anki explains how the two approaches differ. To build the reading habit that feeds your cards, see how to learn a language by reading.