Spaced Repetition for Language Learning: Intervals That Actually Work

Of all the vocabulary methods I have tested across five languages over 20 years, spaced repetition is the one with the most consistent evidence base — and the biggest personal impact on my own retention. When I started learning Norwegian in 2005, digital flashcard apps did not really exist yet, so I built my own manual three-box system out of shoeboxes. It looked ridiculous. It worked exactly as advertised.

What I want to give you in this deep dive is not a marketing pitch for a specific app. It is the mechanism, the intervals that actually work in practice, and the mistakes I made along the way that cost me months of effort.

What spaced repetition actually does

The technique is built on a well-documented phenomenon: the forgetting curve. When you learn a new word, your ability to recall it drops sharply within hours and continues to decline over days. Each successful review flattens that curve — you forget more slowly the next time. After several well-timed reviews, the word moves from short-term into long-term memory, and the interval between reviews can extend to months or years without loss of recall.

The critical variable is timing. Review too early and you are wasting effort on a word you have not yet forgotten. Review too late and you are effectively relearning it from scratch. Spaced repetition schedules reviews at the point where recall is still just possible with effort — because that moment of effortful retrieval is what actually consolidates memory. Research on retrieval practice, well summarised in the journal npj Science of Learning, has confirmed this pattern across dozens of studies.

What this means practically: spaced repetition is not about intensity. It is about hitting the right moment on the forgetting curve, repeatedly, with less and less effort over time.

The intervals that actually work in practice

There are dozens of proposed schedules — Leitner boxes, Anki’s SM-2 algorithm, more recent adaptive systems. In practice, most language learners can start with a simple manual schedule and only move to software if they exceed roughly 300 active cards.

Here is the schedule I have used across all five of my languages, in its simplest form:

• New card — review the same day (evening after morning learning)
• Passed — review after 1 day
• Passed again — review after 3 days
• Passed again — review after 7 days
• Passed again — review after 14 days
• Passed again — review after 30 days, then 90, then annually

Every time you fail a recall, the card drops back to the “1 day” pile. That is the whole system. It fits on a Post-it. No algorithm required.

Software or paper — which is better?

This question comes up constantly, and my honest answer is: both work. The choice depends on how many cards you are managing and how disciplined you are with the schedule.

Paper works when: you have fewer than about 300 active cards, you enjoy the physical ritual, and you can reliably show up daily to sort cards yourself. My Norwegian manual system worked for about 18 months before I switched to software — and only because the card count made manual scheduling impractical.

Software works when: you cross that threshold, or when you want to review on a phone during commutes. Anki is the reference implementation and is free; commercial courses like 17 Minute Languages* build spaced repetition into the course itself so the algorithm handles scheduling silently in the background. If you find setup and card creation friction blocks you, a pre-built course is worth the money.

There is no correct choice. There is only the choice you will stick with for 12 months without abandoning.

How I built my Norwegian vocabulary manually — and what I would change

My three-box system worked like this: an active box (daily), a warm box (weekly), and a cold box (monthly). Cards moved forward on a successful recall and dropped back to the active box on failure. Simple.

Two things I would do differently today:

First, I mixed word types too randomly. Nouns, verbs, adjectives, phrases — all in the same boxes, shuffled together. Retrieval was inefficient because the brain had to switch types constantly. Today I would group cards thematically within each box (see the vocabulary learning guide for why this matters).

Second, I front-loaded low-frequency words because they seemed “interesting”. This was a real mistake — I spent months learning specialised vocabulary while still lacking basic connectors and everyday verbs. The most common words to learn first guide covers what I should have done: work through a frequency list for the first 1,000-2,000 words before branching into specialised areas.

Common mistakes with spaced repetition

1. Adding too many new cards per day

The temptation is to add 30-50 new cards daily. Within three weeks, the review load becomes unmanageable, and learners quit. My rule across all five of my languages: maximum 10 new cards per day in the first month, ramping to 15-20 once the system is stable. Beyond that, retention drops even though you feel productive.

2. Reviewing without effortful recall

Passive review — flipping the card and reading both sides — does almost nothing. The effort of trying to recall the word before flipping is what produces retention. If you find yourself just reading, you are wasting the session.

3. Ignoring the failure signal

When a card keeps failing, the answer is not to force-review it harder. It is to break the card down: add a context sentence, an image, or an association (see the association method in the vocabulary guide). A card that fails five times in a row is telling you the encoding is wrong, not that you need to try harder.

4. Building cards you never review

Creating cards feels productive. Reviewing them is the actual work. If you have a 500-card backlog, stop adding new cards until you catch up. A smaller reviewed system beats a larger unreviewed one every time.

When spaced repetition does not work

Spaced repetition is a vocabulary and phrase-level tool. It is excellent for words, short phrases, gendered nouns, verb conjugations, and any discrete knowledge with a clear right answer. It is not suited for:

• Grammar concepts that require pattern recognition across many examples — those need reading and speaking exposure, not flashcards
• Listening comprehension — see the listening guide for methods that actually train the ear
• Speaking fluency — spaced repetition builds the vocabulary, but only real speaking practice produces retrieval speed in conversation

Use spaced repetition as one layer of your system, not the whole system. The complete language learning guide shows how to combine it with listening, reading, and speaking practice for full coverage.

A simple starting protocol

If you want to start today without buying software, here is the shortest working version:

• Get a small stack of index cards and three envelopes labelled “1”, “3”, “7”
• Write 5-10 new words per day on cards (word on one side, translation and example sentence on the other)
• Every day: review all cards in envelope 1, then any due in 3 and 7 as they cycle in
• Successful recall: card moves forward. Failure: card drops back to “1”
• After a week of consistent daily reviews, add envelopes “14” and “30”

This is exactly the system I used for the first 18 months of Norwegian. It cost me the price of index cards. It produced the foundation I still speak with today.

Related guides in this series:

• The complete language learning guide — the full system
• How to learn vocabulary effectively — the seven core methods, including thematic grouping and the association method
• How to memorise vocabulary fast — the “stick versus cram” question
• The 1000 most common words to learn first — what to feed your spaced repetition system

Which language would you like to learn?

Find the right course for your language here:

→ View all language courses