The Science of Spaced Repetition: Why It Works and How to Use It

As an Amazon Associate we may earn a commission from qualifying purchases made through links on this page, at no extra cost to you.

Spaced repetition is a review schedule. It is not a personality, not a classroom the Language Research Desk sat in, and not a promise that an app streak equals fluency. The desk notes below cover why the schedule works, how to set up a deck you can actually finish, and what to check if you use Anki, Brainscape, or paper cards. Tool comparisons that go beyond this page live in best flashcard apps, best language learning tools, and apps compared.

The Forgetting Curve

In 1885, Hermann Ebbinghaus measured his own recall of nonsense syllables and described a regular drop: a lot of a new item is gone within a day, and most of what remains is gone within a week if it is never reviewed. The exact percentages people quote (often “70% gone in 24 hours”) are classroom shorthand for that curve, not a law that applies to every word in every language. The useful part survived: unreviewed new information decays fast, and a review just before that decay flattens the next curve.

Each successful review at the right moment makes the next gap longer. That is the whole mechanism. Cramming still works for tomorrow’s test. It is a poor way to hold the few thousand word families you need for everyday coverage.

How the Schedule Works

Instead of reviewing every item every day, a spaced-repetition system parks each card on its own calendar:

1. Learn a new word. See it tomorrow.
2. Recall it. See it in about three days.
3. Recall it again. See it in about a week.
4. Then two weeks, then a month.

If you miss the item, the interval shortens. Easy items leave you alone. Hard items come back sooner. That is why a few thousand well-reviewed words outperform a huge shallow list — see how many words you need to be fluent. The algorithm is doing load-balancing, not magic.

Apps that skip a real schedule are just digital stacks. They can still help you cram. They waste time on words you already know.

The Evidence the Desk Will Cite (and No More)

A 2019 meta-analysis in the Journal of Educational Psychology (already cited on the short version of this page) pooled 29 studies and more than 3,000 participants and reported better long-term retention for spaced practice than for massed practice (cramming) across the domains those studies tested.

For very long retention, Bahrick and Phelps (1987) is the usual citation: spaced sessions produced memories that could still be found years later, while massed practice faded. The desk is not adding a new trial and not claiming those participants were language students in every case. The transfer the literature makes is simple: if the thing to remember is a large set of pairs or facts, spacing beats cramming for the long calendar.

What the evidence does not say:

• That Anki is the only valid implementation
• That reviews replace speaking, reading, or grammar
• That 500 new cards a week is “optimized”

Desk Setup: A Deck You Can Finish

This is a listing of setup checks, not a scored catalog and not a claim that the desk ran a lab on your phone.

Card load

• New cards: 10–20 a day is the band that survives a workweek. Reviews of old cards sit on top of that number.
• Daily reviews: do them the day they are due. Skipping three days does not save time; it creates a pile that then gets skipped again.
• Backlog rule: if reviews take more than about 25–30 minutes, stop adding new cards until the queue is honest again.
• Coverage over volume: a stable 2,000–3,000 word-family deck beats a 15,000-card graveyard. Pair the cap with the targets on how many words to be fluent.

Sentence cards (and when a single word is enough)

• Prefer one sentence, one hole (cloze) for verbs, prepositions, and collocations.
• A picture-only concrete noun can be a single word if there is no English on the front.
• Never put five meanings on one card. Split them.
• If the back is only an English gloss, you are training the translation loop described in thinking in a new language. Add a picture, a definition in the target language, or a sentence.

Audio

• Add audio for words you cannot already say.
• Confirm the clip is the language (and variety) you are learning. A European Portuguese clip on a Brazilian deck is a listing mismatch, not a feature.
• Keep the audio matched to the sentence on the card. A random isolated word plus a mismatched example trains two items at once.

What a daily session actually is

A desk-shaped session is reviews first, then new cards, then stop. It is not “clear the whole downloadable deck.” Fifteen focused minutes of due cards beat an hour of browsing shared decks you will never finish.

If you want a method book that walks pronunciation, images, and SRS as one system, Fluent Forever by Gabriel Wyner is the title already used elsewhere on this site for that job. It is a how-to for building cards, not a replacement for input or conversation.

Listing check (The Language Research Desk): Confirm the Wyner Fluent Forever method book (original or revised is fine if pronunciation and SRS are still in the contents). Marketplace search also returns workbooks and app upsells with similar names — check the author line and the table of contents, not only the cover.

Anki, Brainscape, and Paper: Listing Notes, Not a Score

The desk does not run a test fleet and does not mint star ratings. These are checks you do on the listing or the download page before you commit a year of reviews to a tool.

Anki (desktop and Android free; iOS sold separately)

Anki is the flexible spaced-repetition program most serious learners mean when they say “SRS.” It schedules cards, syncs through AnkiWeb, and lets you put text, images, and audio on custom templates.

Listing check (The Language Research Desk):

• Desktop and Android are free; the iOS app is a separate paid download that funds development. Confirm you are on the official Anki page, not an unrelated “Anki” vocabulary clone.
• Shared decks vary wildly. Open the description, card type, and a few samples before you subscribe your next six months to someone else’s 12,000 notes.
• “Anki” on Amazon is usually a book about the method, or unrelated electronics. This page does not treat a random Amazon box as the app.

Desk tips if you already chose Anki: one concept per card; example sentence; image or audio when the word needs it; reviews every day; 10–20 new cards. Those tips are setup, not a claim the desk lived inside the app.

Brainscape (confidence-rated reviews; Pro unlocks the rest)

Brainscape asks for a 1–5 confidence rating instead of a binary pass/fail. Some learners find that more honest than Anki’s default buttons. The trade is less template control and a smaller public library.

Listing check (The Language Research Desk):

• Confirm whether the deck you want is on the free tier or locked behind Pro (monthly vs annual — read the current plan, not an old blog price).
• Confidence ratings only help if you are honest. Rating everything “5” to clear the queue is cramming with extra steps.
• We are not scoring Brainscape against Anki. If you will open Brainscape and you will not open Anki, that is the decision.

Physical cards (Leitner box)

If you prefer paper, the Leitner system is the same schedule with boxes:

• Box 1: daily
• Box 2: about every three days
• Box 3: weekly
• Box 4: every two weeks
• Box 5: monthly

A correct answer moves one box to the right. A miss returns to Box 1. Any index cards and five envelopes work. A branded “Leitner box” is optional.

Listing check (The Language Research Desk): If you buy cards or a box, confirm size (3×5 vs 4×6), ruled vs blank, and that the listing is cards — not a storage crate or a children’s game with the same word in the title. Paper does not sync. That is the whole product note.

For a wider app list (Memrise, Quizlet, Drops), use best flashcard apps for vocabulary. For how flashcards sit next to tutors, textbooks, and audio, use best language learning tools and apps compared.

What SRS Will Not Do

Spaced repetition holds items you decided were worth holding. It will not:

• Give you a speaking partner
• Teach you which 3,000 words to choose (frequency lists and reading do that)
• Replace immersion or grammar in context
• Rescue a deck of isolated translations if you never produce a sentence

Use the algorithm for the memory problem. Use the rest of the site for the language problem.