Spaced Repetition for Chinese: An SRS System That Sticks

The best SRS method for Chinese vocabulary in 2026 is the simplest one you’ll actually run every day: a small, honest self-rating (something like Flashcard’s Reveal, Pass, Again) applied to vocabulary pulled from text you’re actually reading, reviewed in short daily sessions with a hard cap on new words. The specific scheduling algorithm matters far less than consistency and the quality of the cards going in — Chinese vocabulary from context you understand sticks better than vocabulary from a generic frequency list.

Spaced repetition works because it fights a very specific, very predictable enemy: the forgetting curve. Left alone, your memory of a new Chinese word decays fast — you might recall it an hour later, struggle the next day, and lose it entirely within a week. An SRS interrupts that decay by timing reviews to land right before you’d forget, not on a fixed schedule and not whenever you happen to feel like studying. In Flashcard, that timing is driven by three actions after each card comes up: Reveal shows you the pinyin, definition, and example sentence so you can check yourself; Pass tells the system you knew it, which pushes the card further into the future; Again tells the system you didn’t, which brings the card back soon instead of letting it sit until its next scheduled slot. That loop — reveal, judge yourself honestly, pass or again — is the entire mechanism. Everything else in this post is about using it well.

How should I schedule Chinese reviews?

The scheduling itself is handled for you — that’s the point of an SRS — but you still control two things that determine whether the system actually works: when you show up, and how honestly you rate yourself.

On the “when you show up” side, daily beats any other cadence, even a shorter session done every day beats a longer one done every few days. This isn’t a preference, it’s a consequence of how the underlying math is structured: a card’s next interval assumes you’ll be there to review it inside a certain window. Skip a few days and cards pile up stale, half-forgotten, and bunched together, which makes the session that finally happens feel punishing instead of quick. A five-minute session every day keeps the queue small and the reviews light. A missed week turns that same queue into a 40-minute slog.

Here’s the cadence I’d actually recommend, based on running this workflow on my own Chinese study for a while:

• Morning or commute: one review session, 5-10 minutes, all due cards — no new words yet.
• Whenever you read something new (an article, a chat, a caption): paste it into Flashcard, extract the vocabulary you don’t know, and let those become tomorrow’s new cards rather than today’s, so you’re not reviewing something you saw four minutes ago.
• Evening: a second short pass if the due queue built up during the day, otherwise skip it — there’s no bonus for reviewing a card twice in one day if you already knew it in the morning.
• Weekly, not daily: skim which words keep coming back as “Again” and figure out why — wrong tone, similar-looking character, or a definition that’s too abstract without more context. That’s a five-minute weekly check, not something to do mid-session.

The honesty part matters more than people expect. It’s tempting to hit Pass when you sort of recognized a character, even if you couldn’t have produced the pinyin or meaning cold. Every generous Pass is a card that gets scheduled weeks out before it’s actually secure, and it will show up again as a total stranger when it resurfaces — at which point you’ve lost more time relearning it than you saved by not hitting Again earlier. Rate for recall, not recognition.

How many new words per day is realistic?

This is where most people sabotage their own SRS, and it’s almost always in the same direction: adding too many new words on a good day, then getting buried a week later when all of those cards mature into the review queue at once.

I’d suggest a cap of 10-15 new words a day for most learners working through intermediate material, and I’d pull that down to 5-8 a day once you’re past HSK 4, when the words you’re encountering get more abstract, more easily confused with near-synonyms, and slower to actually stick. The reasoning is simple: new words don’t cost you anything the day you add them, they cost you on day three, day seven, and day twenty-one, when they come due alongside every other day’s new words. A cap isn’t about limiting how much you’re “allowed” to learn — it’s about keeping the future review load something you’ll actually sit down and do.

A rough way to sanity-check your own cap: multiply your daily new-word count by roughly 4, and that’s approximately the size of your due queue once the system stabilizes (accounting for words that graduate to longer intervals and drop out of daily rotation). Fifteen new words a day settles into something like a 50-60 card daily review once it’s running steady. If that number sounds like more than you want to spend ten minutes on each morning, lower the new-word cap now rather than three weeks from now when the queue is already that size.

This is also the strongest argument for extracting vocabulary from real text instead of a frequency list. When Flashcard pulls words from an article you actually read and understood in context, each one already has a mental hook — the sentence it came from, the topic, maybe the reason you were reading that piece at all. Words from a raw frequency list have none of that, so they take more review cycles to stick, which quietly inflates your effective daily cost per word. Fewer, contextual words beat more, context-free ones. I go through the extraction step itself in Turning a Chinese Article Into Flashcards with AI, which is the workflow that feeds this review loop in the first place.

What actually happens when you tap Again

It’s worth being precise about this, because “Again” doing the wrong thing quietly is a common way SRS trust breaks down. When you tap Again, the card doesn’t get punished by being pushed to some distant “review this again in a month” slot — it comes back soon, typically within the session or the next day, so you get another shot at it while the miss is still fresh. That’s the entire design intent: failing a card isn’t a setback to the schedule, it’s the schedule doing its job by noticing the interval was too long and correcting it downward. You should expect to tap Again fairly often on new words in their first week — that’s the system working, not a sign you’re bad at this.

Passing a card too early, on the other hand, is the failure mode that’s invisible until it resurfaces. If you’re not sure whether to Pass or Again, default to Again. The cost of an extra review in three days is a few seconds. The cost of a false Pass is relearning the word from scratch weeks later, at a point where you’d stopped expecting to see it at all.

Pairing SRS with something other than pure recall

Reviewing isolated vocabulary cards is efficient, but it’s not the only thing worth doing with your study time, and it shouldn’t be. Recognizing 记得 in isolation is a different skill from recognizing it inside a sentence with different grammar around it, and pure card review doesn’t build the second skill on its own. If you want a complementary technique that works alongside an SRS rather than replacing it, working through example sentences for the words you’re already reviewing is one of the better uses of the ten minutes you’re not spending on cards — I’ve written about that approach in how example sentences help you learn Chinese faster, which pairs well with whatever vocabulary your SRS queue already has due that day.

Where this fits into the Flashcard workflow

None of this requires configuring an algorithm or tuning intervals by hand — that’s deliberately not something you touch in Flashcard. You paste text, AI extracts the Chinese vocabulary with pinyin, definitions, and example sentences attached, and the extracted words go straight into the Reveal / Pass / Again review loop, with text-to-speech audio on each card so you can hear it as well as read it. The only two decisions that matter are the ones this post is about: showing up daily for a short session, and capping new words so the queue stays a five-minute task instead of a forty-minute one. Everything else — when a card resurfaces, how far out it gets pushed after a Pass — is handled for you. If you want your own vocabulary out of the system at any point, export is one click, via a shareable URL, CSV, or clipboard, and the whole thing is free with no install. You can see the rest of what it does, along with the other tools I’ve built, on the Flashcard project page.