How to Increase Spaced Repetition Efficiency

The short answer

To increase spaced repetition efficiency, upgrade in this order. Use a modern scheduler such as FSRS. Review native language to target language so each card is real retrieval, not recognition. Put words in sentences (mined sentences or cloze) instead of isolated pairs. Then stop repeating the same sentence: use a new one every review so you practice building language, not memorizing a card. After that, grade each word on its own and pack several due words into one sentence so one attempt clears more of the queue.

Level 1: Choose a good algorithm

This is the easiest win because it is mostly a setting. Older algorithms such as SM-2 use crude ease factors that drift. FSRS tracks how hard each item is and how stable your memory of it is, then brings the card back near the point you would forget it.¹ Use FSRS, and use the newest version (right now that is FSRS-6).

Level 2: Native language to target (retrieve, do not recognize)

Show the meaning in your native language. Produce the word in the language you are learning. That is retrieval: you pull the form out of memory.

The other direction (see the foreign word, recall the native gloss) is mostly recognition. So are multiple-choice and tap-the-translation cards. Recognition is easier, so sessions feel productive. It is a weaker memory event than retrieval, which is what you want each review to be.²³

If the answer is already on the screen, you are not testing the skill you think you are.

Level 3: Use sentences, not word lists

A native-to-target word pair only links a form to a gloss. It does not show how the word usually shows up: what sits next to it, which verbs or prepositions it takes, or what kind of sentence it belongs in. Context is useful because you see the situations the word actually appears in, not just a dictionary line.⁴

Two common upgrades from word lists:

• Sentence mining: you save a real sentence and review that sentence on a schedule.
• Cloze cards: you see a sentence with one blank and fill in the missing word.

Both are better than a naked word list. Both still have a construction problem.

With a mined sentence, you review the same sentence over and over. You memorize that sentence: that word order, those neighbors. You are not practicing how to construct a new sentence with those words. And because the whole sentence is one card, one wrong word marks the whole review wrong, even if the other words were right. For more on that trap, see why sentence mining fails.

With a cloze card, the sentence is already built. The grammar, word order, and most of the words are given. You only fill in the blank. You never construct the sentence yourself. That is why cloze can feel easy in review and still leave you unable to write the sentence on your own. See cloze card drawbacks for a fuller breakdown.

Level 4: Use a new sentence each time

Keep the word (or grammar point) you are reviewing. Change the sentence every review.

If the prompt never changes, you start recognizing the card. You remember what this card always says, not how to use the word. A new sentence forces you to use the same item in a new context. That is closer to real use, and it stops the “I knew it in Anki” trap.⁵

Level 5: Grade each word, not the whole sentence

Once you write a full sentence, do not mark the entire card right or wrong. If you missed one word and got four other words right, only the missed word should come back soon. If you mark the whole sentence wrong, you waste reviews on words you already know. If you mark the whole sentence right, the weak word disappears for too long.

Level 6: Put several due words in one sentence

Suppose three words are due today. Three separate cards means three exercises. One sentence that uses all three lets you practice them together in a single attempt. You still retrieve each word. You get the same reviews done with fewer exercises.

Example: quiero, café, and con are due. Instead of three cards, you translate something like “I want coffee with milk.” You produce all three in one go.

How to take advantage of these improvements

Levels 1 and 2 are just settings. Levels 3 and 4 can be done in existing apps like Anki with a little extra work. Levels 5 and 6 are painful to run by hand.

LinGoat solves all of this for you. It uses FSRS-6, native-to-target sentence writing, a new sentence each review, per-word grades, and packs multiple due words into each exercise. See how LinGoat works or start practicing.

References

1. Ye, J., Su, J., & Cao, Y. (2022). A stochastic shortest path algorithm for optimizing spaced repetition scheduling. Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 4381-4390. https://doi.org/10.1145/3534678.3539081
2. Carpenter, S. K., Pan, S. C., & Butler, A. C. (2022). The science of effective learning with spacing and retrieval practice. Nature Reviews Psychology, 1, 496-511. https://doi.org/10.1038/s44159-022-00089-1
3. Roediger, H. L., & Marsh, E. J. (2005). The positive and negative consequences of multiple-choice testing. Journal of Experimental Psychology: Learning, Memory, and Cognition, 31(5), 1155-1159. https://doi.org/10.1037/0278-7393.31.5.1155
4. Nation, I. S. P. (2013). Learning Vocabulary in Another Language (2nd ed.). Cambridge University Press. https://doi.org/10.1017/CBO9781139524759
5. Nielsen, E., et al. (2024). Sentence-based spaced repetition for vocabulary learning. Proceedings of the 19th Workshop on Innovative Use of NLP for Building Educational Applications (BEA). https://aclanthology.org/2024.bea-1.29/