Spaced repetition for Chinese vocabulary, without the burnout

Spaced repetition is the closest thing language learning has to a free lunch. Review a word just before you would have forgotten it, and each review buys a longer interval. The evidence for it is old and solid.

It also has a brutal dropout rate. The pattern is familiar: enthusiastic first month, 400 cards, a review queue that grows faster than you can clear it, guilt, abandonment.

The technique is not the problem. Deck design is.

Failure 1: adding words you have never met

The most common mistake is downloading a 5,000-word HSK deck and starting at card one.

A word you have never encountered in context has nothing to attach to. You are memorising an arbitrary pairing, which is the hardest possible kind of learning and the fastest to decay. Worse, you have no sense of how the word behaves: what it collocates with, how formal it is, whether it is a verb or a noun in practice.

Add words you have met. Read or listen first, then card the words that got in your way. The card is then a reminder of an experience rather than a cold fact, and it sticks far better.

This also solves prioritisation for free. Words you meet often are the words worth knowing, and you meet them by consuming material rather than by trusting a frequency list.

Failure 2: too many cards per day

Here is the arithmetic nobody does before starting. Every new card generates roughly ten reviews over the following year. If you add 30 new cards a day, you are signing up for a steady state of several hundred reviews per day, forever.

That is a part-time job. It is why people quit.

A sustainable number for most people with other commitments is 5 to 15 new cards a day. That feels insultingly slow in week one. In month twelve it is 2,000 to 5,000 words with a review load you can actually clear, which beats 400 words and a deleted app.

Choose your daily review budget first, in minutes. Then set the new-card limit to whatever keeps you inside it. Not the other way around.

Failure 3: cards that test the wrong direction

There are two very different skills, and they need different cards:

• Recognition: see 医院, know it means hospital. Needed for reading and listening.
• Production: think "hospital", produce 医院 with correct tones. Needed for speaking and writing.

Production cards are roughly three times harder and take much longer to mature. Most learners make only recognition cards, then wonder why they cannot speak.

A reasonable compromise: recognition cards for everything, production cards only for words you actually want to use. You do not need to produce every technical term you can read.

Failure 4: fighting leeches

Every deck develops leeches: cards you fail again and again while the rest of the deck behaves. They are demoralising out of all proportion to their number.

A leech almost always means the card is badly formed, not that you are stupid. Usually one of:

• The card has two answers. A word with several meanings tested as one card fails forever. Split it, or specify which sense.
• It is confusable with another card. 找 and 我, 天 and 夫. Interference is real. Put both on one card explicitly contrasting them, or suspend one until the other is solid.
• You have no context. Back to failure 1. Delete it and re-add it when you meet it in the wild.

Deleting a leech is a legitimate move. A card you have failed twelve times is costing more than it returns.

What goes on a good card

For Chinese, a card that works:

• Front: the characters alone. Not pinyin, or you will read the pinyin and skip the characters.
• Back: pinyin with tone marks, a short meaning, and one example sentence from where you met it.
• Audio if you have it. Recognition without sound builds a reading-only vocabulary.

The example sentence is the part people skip and the part that does the most work. It carries grammar, register and collocation at no extra review cost.

Sentence cards

An alternative worth knowing: instead of word cards, make cards from whole sentences where you know everything except one item.

Advantages: you rehearse grammar and word order alongside vocabulary, and the word is never stored in isolation. Disadvantages: slower per card, and harder to build without a source of good sentences.

Many strong learners run both, with word cards for volume and sentence cards for things where usage matters more than translation.

The habit is the whole game

Spaced repetition only works if the reviews happen. That makes consistency worth more than optimisation.

• Review daily, at a fixed time. The algorithm assumes you show up. Skipping three days creates a pile that makes skipping a fourth easier.
• Clear reviews before adding new cards. Reviews are the debt you already owe.
• Cap the session. Twenty minutes you actually do beats an hour you dread.
• Take the backlog hit when it happens. After a break, do not try to clear 900 cards in a day. Lower the new-card limit to zero and let the queue drain over a week.

Where flashcards stop

Worth saying plainly: flashcards build recall of items. They do not build fluency, and a deck is not a substitute for reading and listening.

The right role for a deck is to make the words you meet while reading stick, so that the next text is slightly easier. Reading is the engine. Cards are the ratchet that stops you sliding back.

KoiSpeak lets you turn any lookup into a card for exactly that reason: the moment a word blocks you while reading is the moment it is worth remembering, and it is also the moment you have the context that makes it stick.