The Complete Spaced Repetition Guide for Vocabulary: From First Card to Advanced Settings

Spaced repetition is a review method in which you revisit each word at gradually longer intervals, for example after 1 day, 3 days, a week, two weeks and a month, so that every review happens just as the memory starts to fade. Each successful recall at that point makes the memory more durable, which lets the next gap be longer. For vocabulary, it is the single most reliable way to keep the words you learn.

This guide covers the whole path: why it works, a schedule you can run on paper, how to predict your daily workload, what the main settings in apps like Anki do, and how to combine flashcards with reading so the words become usable rather than just memorized.

Why spacing works

Two well-established findings sit underneath every spaced repetition system:

• The forgetting curve. Ebbinghaus (1885) showed that newly learned material is lost quickly at first, then more slowly. Each successful review flattens the curve.
• The spacing effect. The same amount of study spread over time produces better long-term retention than the same study crammed together. Cepeda and colleagues (2006) reviewed hundreds of experiments and found the effect to be one of the most robust in learning research, and Cepeda et al. (2008) showed that the best gap grows as the time you need to remember something grows.

A third principle, active recall, is what makes reviews count. Pulling an answer out of memory strengthens it far more than re-reading it (Roediger and Karpicke, 2006). Bjork (1994) called this kind of productive struggle a desirable difficulty: reviews should feel slightly hard. If every card is instant, your intervals are too short.

Level 1: A fixed schedule you can run on paper

You do not need software to start. Use this expanding schedule for every new word:

If you fail a review, the word goes back to review 1. The spaced repetition scheduler generates these dates for any start day and exports them to your calendar.

The Leitner box version

The Leitner system (Leitner, 1972) turns the schedule into physical boxes. Box 1 is reviewed daily, box 2 every 2–3 days, box 3 weekly, box 4 every two weeks, box 5 monthly. A correct card moves up one box; a wrong card drops back to box 1. It is crude but self-correcting, and it is a good way to feel how spacing works before trusting an algorithm.

Level 2: Know your daily load before it knows you

The number one reason people quit spaced repetition is not boredom. It is a review pile that grew faster than they expected. Reviews are a delayed consequence of new cards: what you add today, you pay for over the next months.

With the fixed six-step schedule above, every new word generates six reviews over 90 days, plus extra reviews whenever you fail one. If you add the same number of new words every day, after about three months each day contains reviews for words added 1, 3, 7, 14, 30 and 90 days earlier:

Add time for the new cards themselves, which take longer than reviews. Adaptive algorithms change the shape (easy cards drift to long intervals, hard ones come back often), but the rule of thumb holds: daily reviews settle at several times your daily new cards. Choose the new-card rate from the time you have, not from your ambition. The 10 words a day plan works through this math in detail.

Level 3: Making good vocabulary cards

A perfect algorithm cannot save bad cards. For vocabulary, these rules matter more than any setting:

1. One fact per card. "mitigate = make less severe" is one card. Do not cram synonyms, etymology and three example sentences onto the answer you have to recall.
2. Use a real sentence. Cards built from sentences you met while reading are easier to remember and teach usage. This is the core of sentence mining.
3. Decide the direction. Recognition cards (word → meaning) are fast and build receptive vocabulary. Production cards (meaning or cloze sentence → word) are harder and build productive vocabulary. Make production cards only for words you want to use.
4. Be honest when grading. If you got the gist but not the word, that is a fail on a production card. Generous grading inflates intervals and you pay later.

Level 4: How algorithms schedule (SM-2 and FSRS)

Software replaces your fixed table with per-card intervals based on how you answer.

SM-2

The SM-2 algorithm (Wozniak, 1987) gives each card an "ease" factor. After a successful review, the next interval is roughly the previous one multiplied by the ease; pressing Hard lowers the ease, Easy raises it, and a failure sends the card back to short relearning steps. It works, but it has known weaknesses: ease factors that drift down after a few failures can trap cards in short intervals ("ease hell"), and the multipliers are the same for everyone.

FSRS

FSRS (Free Spaced Repetition Scheduler) models each card's memory with estimates of stability and difficulty, and fits its parameters to your own review history. Instead of choosing multipliers, you choose a desired retention: the probability you want to remember a card when it comes due. Anki offers FSRS as an option in its deck settings, with a default desired retention of 90%.

Level 5: Settings that matter (in general terms)

Menus differ between apps and versions, so here are the concepts rather than click paths:

Two rules beat any tweak: grade with only "Again" and "Good" if you are unsure about Hard and Easy, and once you have a few hundred reviews, let FSRS optimize its parameters on your history rather than editing numbers by hand.

Raising retention costs a lot

Workload does not rise linearly with retention. Moving from 90% to 95% means intervals have to shrink noticeably, so daily reviews can grow considerably for a few points of retention. For general vocabulary, 85–90% is the sweet spot; you will re-meet many words in reading anyway.

Combining spaced repetition with reading

Flashcards are a memory tool, not a language. A word that lives only on a card is brittle: you know meticulous means "very careful", but you may not know it describes people and work (meticulous records) more than actions, or that it sounds positive. Reading supplies that depth.

1. Read at 95–98% coverage and capture useful words with their sentences, as described in how to learn vocabulary from reading.
2. Turn the best captures into cards, one sentence each. Skip words you will meet constantly anyway; reading will handle them.
3. Review daily in one short session, before adding anything new.
4. Notice the words again in the wild. Each time a carded word shows up in your reading, that is a free, high-quality review. Some tools make this visible: Wordie, for instance, highlights saved words on any web page you visit, so you notice them in context. Its own Recall mode is a simple flashcard review with a mastery score, not a full SM-2 or FSRS scheduler, so pair it with Anki (via CSV export) if you want algorithmic intervals.

Learn more about getting browser-captured words into a deck in how to import words into Anki.

Troubleshooting

The bottom line

Start with a fixed schedule (1, 3, 7, 14, 30, 90 days), make one-fact cards from real sentences, and set your new-card rate from the time you actually have. When you move to software, let an adaptive algorithm like FSRS handle intervals at around 90% retention. And keep reading: spaced repetition stops you forgetting words, while reading teaches you what they really mean. For more on the memory side, see how to remember new words.

Sources: Ebbinghaus, H. (1885), Über das Gedächtnis; Cepeda, N.J. et al. (2006), "Distributed practice in verbal recall tasks", Psychological Bulletin; Cepeda, N.J. et al. (2008), "Spacing effects in learning", Psychological Science; Roediger, H.L. & Karpicke, J.D. (2006), Psychological Science; Bjork, R.A. (1994), "Memory and metamemory considerations in the training of human beings"; Leitner, S. (1972), So lernt man lernen; Wozniak, P. (1987), SM-2 algorithm.