Use spaced repetition to retain Chinese vocabulary long-term: card design, Anki setup, daily habits, and mistakes that waste your reviews.
The Forgetting Curve And Why Cramming Fails
Memory research over a century old shows that we forget new information on a predictable curve: most of a new word evaporates within days unless it is reviewed at expanding intervals. Cramming fifty words the night before a test produces a brief spike and near-total loss within weeks. Spaced repetition, or 间隔重复 (jiàngé chóngfù — spaced repetition), fights the curve by scheduling each review just before you would forget: after one day, then three, then a week, then a month. Each successful recall strengthens and slows the next decay. For a language with thousands of unfamiliar sound-meaning pairs like Mandarin, this is not optional. Learners who review on schedule retain over ninety percent of vocabulary; learners who study in bursts relearn the same words endlessly. The algorithm turns forgetting from an enemy into the trigger for learning.
How SRS Algorithms Schedule Reviews
Modern spaced repetition apps assign each card a stability value that grows with successful recalls and shrinks with failures. When you press again on a forgotten card, its interval resets short; when you recall 朋友 (péngyou — friend) instantly after a month, the next interval might stretch to three months. The math is invisible but the effect is profound: your daily review load concentrates on exactly the words you are about to lose. Newer algorithms adapt to your personal speed, so two learners with the same deck see different schedules. What matters practically is trust: answer honestly, never press good on a guess, and let the system do its job. Cheating the algorithm only defers forgetting to a moment when you need the word in real conversation.
Designing Cards That Teach
Card design determines half your results. The best Chinese card is minimal: front shows the word 图书馆 (túshūguǎn — library) with audio, back shows pinyin, meaning, and one short example sentence. Test recognition direction for most words — see Chinese, recall sound and meaning — because that is what reading and listening demand. Add production cards sparingly for core vocabulary you must speak: front shows English to go, back shows 去 (qù). One card, one fact: never test two meanings or two readings on a single card. Include audio whenever possible so pronunciation is trained alongside meaning. Avoid English-heavy fronts that let you guess from context; the card should force genuine retrieval. Twenty well-made cards beat a hundred lazy ones, because each review is a real memory event rather than a recognition illusion.
What Never Belongs On A Card
Most bloated decks fail for the same reasons. Do not add full grammar explanations, long word lists, or characters you have never met in context — cards review memory, they cannot create understanding from nothing. Do not add rare words from HSK lists that you have never seen used; you will memorize them, forget them, and never miss them. Skip sentences longer than fifteen syllables, because they test patience rather than recall. And never add words you do not actually want to learn; deck-building is curation. A useful filter before adding anything: have I seen this word twice in real content, and will I plausibly need it this year? If not, let it go. A lean deck of three thousand beloved words outperforms a graveyard of ten thousand strangers every single time.
Setting Up Anki For Chinese
Anki is the standard tool and its defaults mostly work for language learning, with a few adjustments. Set new cards to ten or fifteen per day and review limit high enough to clear the queue. Install a Chinese note type that includes fields for characters, pinyin, audio, and example sentence; community decks share good templates. Enable audio autoplay so every review trains listening. Keep learning steps short so forgotten cards return quickly. Sync to your phone, because the best SRS habit is the one you carry everywhere. Before adding bulk imports, look words up in a reliable dictionary to confirm meanings — community decks contain errors, and memorizing a wrong gloss costs double effort to fix later. Spend one evening on setup, then never fiddle again; consistency beats configuration.
The Daily Review Habit
SRS only works if reviews happen daily, because the algorithm schedules cards for specific days. Attach reviews to an existing anchor: morning coffee, commute, or the first ten minutes at your desk. Do reviews before new cards — clearing the queue protects the memories you already own. If the pile exceeds your time, suspend new cards and grind the queue down; an overdue mountain is the normal state of a paused deck, and it shrinks steadily with a few honest days. Expect resistance around day forty, when early enthusiasm fades and daily reviews peak. That is the moment the system is doing its most valuable work. Learners who survive month two with an unbroken streak almost never quit, because visible retention — recalling words from months ago instantly — becomes its own reward.
Balancing New Cards Against Reviews
The arithmetic of SRS is unforgiving: each new card creates roughly ten reviews over its lifetime, spread across months. Adding thirty new cards daily sounds ambitious until reviews hit three hundred a day in month two and the deck collapses. A sustainable rate for most learners is ten to fifteen new cards daily, which stabilizes around a hundred reviews — twenty to thirty minutes. Watch your numbers: if review time creeps over forty minutes, pause new cards until it settles. During intensive periods, like before a trip or exam, drop new cards entirely and let reviews shrink. Think of your deck as a garden with a fixed water supply; new cards are new plants, and planting faster than you can water kills everything. Slow planting, daily watering, endless harvest.
When SRS Stops Working
Spaced repetition fails in predictable ways. Leeches — cards failed eight or more times — signal a deeper problem: the word may be too similar to another, lack context, or carry a confusing gloss. Suspend leeches and relearn them through reading or mnemonics rather than more grinding. Another failure mode is recognizing cards without knowing words: you memorize the card's visual layout instead of the vocabulary. Fix this by editing problem cards, adding a fresh example sentence, or testing in the other direction. The deepest failure is a deck disconnected from real input; words that never appear in your reading feel abstract and fade fast. If reviews feel meaningless for weeks, the cure is not more reviews but more real Chinese — stories, shows, conversations — that give your cards a reason to exist.
Pairing SRS With Real Input
SRS is the cement, but input is the bricks. The strongest workflow sends vocabulary in both directions: mine words from your reading and listening into the deck, and let deck words reappear in fresh content. When 满意 (mǎnyì — satisfied) comes up in review and then shows up in a story that evening, the memory locks permanently. Graded materials make this circulation easy; read daily from the graded library and add only the words that repeat across texts, because frequency signals usefulness. A helpful ratio is two hours of input for every hour of review. Learners who invert that ratio build decks full of strangers. The goal was never to be good at flashcards; it is to read, listen, and speak, with the deck quietly ensuring nothing you meet is ever lost.
A Sustainable Long-Term Routine
Here is a routine that survives years. Morning, fifteen minutes: clear due reviews, audio on, honest grading. Evening, fifteen minutes: read one graded text and add up to five recurring words as new cards. Weekly, ten minutes: check statistics, suspend leeches, and adjust the new-card limit to keep reviews near twenty minutes. Monthly: export a backup and reflect on whether deck words are appearing in your real life; if not, shift mining toward content you actually consume. Never chase zero — a hundred pending cards after a busy weekend is normal, not failure. Protect the streak above all: even five minutes of reviews on your worst day keeps the algorithm honest. Over twelve months this humble routine quietly builds four thousand words with ninety percent retention, which is the entire foundation of fluency.