Everyone who has studied a language knows the pattern: you learn twenty words on Sunday, and by Friday you can recall three of them. The problem is not your memory — it is your schedule.
The forgetting curve
In the 1880s, Hermann Ebbinghaus measured how quickly he forgot nonsense syllables and found that memory decays roughly exponentially: steep loss in the first hours and days, then a long, slow tail. The crucial discovery was what happens when you review just before you would have forgotten: the curve resets, and it decays more slowly each time. Review a word after one day, then three, then a week, then a month, and it becomes effectively permanent — for a total of four or five reviews.
Spaced repetition systems (SRS) simply automate this arithmetic. Anki is the best-known tool and the one I use; every serious flashcard app now has some version of the algorithm inside. The software’s only job is to answer one question, thousands of times: which cards are about to fall off the cliff today? You review those, and only those. The efficiency is startling when you feel it working — ten minutes a day maintains a vocabulary of thousands of words, because at any given moment only a tiny fraction of them are actually at risk.
What actually goes wrong
The algorithm is not the hard part. In my experience, spaced repetition habits die for three predictable reasons:
- The deck grows faster than the habit. Adding fifty new cards a day feels productive and creates a review avalanche a month later — because every new card is a promise of four or five future reviews, and the promises all come due at once. Twenty new cards a day is plenty; ten is sustainable forever.
- Cards test recognition, not production. Seeing a word and remembering its meaning is far easier than producing the word from meaning. If you only train recognition, that is all you get: a reading vocabulary that abandons you the moment you open your mouth. Make at least some cards in the production direction.
- Cards lack context. A bare word pair (“mitigate = giảm nhẹ”) is brittle. A short sentence with the target word blanked out gives your memory something to grip — collocations, grammar, register, all for free.
There is a fourth, quieter failure mode: treating the SRS as the whole method. Flashcards maintain memories; they do not create understanding. The words have to come from somewhere real — reading, listening, conversation — and they have to be spent somewhere real, or the deck becomes a well-tended museum of words you never use.
A minimal setup that works
Make cards from words you actually met — in a book, a show, a conversation — rather than from frequency lists. The encounter itself is half the memory; a word you needed once is a word you will plausibly need again. One sentence per card, target word hidden. Ten new cards a day, reviews every morning with coffee, and be ruthless about deleting cards you keep failing: some words are simply not ready to be learned yet, and they will come back around when your level catches up to them.
Two settings-level tips for Anki specifically: turn the daily new-card limit down before motivation forces you to, and never let the review backlog exceed what you can clear in one sitting — a three-day backlog cleared honestly beats a ten-day backlog abandoned entirely. Missed days happen; the algorithm tolerates them better than your motivation does.
Expect a plateau around the six-month mark, when the easy, common words are done and every new card is rarer and drier. That is the point where most decks die, and the fix is not discipline but material: go back to reading and watching things you enjoy, and let the deck follow your curiosity instead of leading it.
The system is boring. That is precisely why it works.