How Spaced Repetition Makes Korean Vocabulary Stick

If you have studied a list of Korean words, felt confident that night, and then struggled to recall them later, you have seen a normal feature of memory: access to newly learned information becomes less reliable when it is not used.

Spaced repetition does not stop forgetting. It gives you repeated opportunities to retrieve information over increasing delays, so review time is concentrated on material that still needs work.

What the forgetting curve actually tells us

Hermann Ebbinghaus's early memory experiments showed a general pattern that later became known as the forgetting curve: forgetting can be relatively rapid after learning and then slow over time. There is not one universal curve saying that every learner forgets a fixed percentage after 24 hours or one week. The shape depends on the material, the learner, prior knowledge, how well something was learned, and the delay before it is tested again.

Modern research gives us a more useful practical conclusion: spacing study across time and practicing retrieval can improve later retention compared with repeatedly studying the same material in one massed session. The best spacing is not a single timetable for everyone; it also depends on how long you want the information to remain accessible.

What is spaced repetition?

A spaced-repetition system schedules another review after you study or retrieve an item. If recall remains successful, the delay can usually grow. If you fail to recall the item, the scheduler can bring it back sooner.

A simplified lifecycle looks like this:

1. Learn a Korean word or pattern.
2. Try to retrieve it after a delay instead of immediately rereading it.
3. Record whether recall succeeded and how difficult it felt.
4. Let the scheduler choose another review time from that history.
5. Shorten the future delay after a lapse and lengthen it as the memory becomes more stable.

Those steps are the important idea. A sequence such as “review on Day 2, then Day 6, then Day 16” is only an example schedule, not a rule your memory should follow.

Retrieval matters, not just seeing the card again

Spacing is most useful when review makes you attempt an answer before revealing it. Research on retrieval practice consistently finds that trying to recall learned information can improve later access to it compared with additional passive study, especially when errors can be corrected with feedback.

For Korean, that means a useful card should ask you to do something: recall a meaning, produce a Korean form, recognize a sound, complete a sentence, or distinguish two patterns. Simply rereading the front and back is a much weaker test of whether you can retrieve the information later.

From SM-2 to FSRS

Early digital spaced-repetition systems helped popularize algorithms based on the SuperMemo family, and Anki historically used a scheduler derived from SM-2 ideas. Newer schedulers can model memory more directly.

Chamelingo's course review uses FSRS-5. FSRS tracks memory state from review history and schedules around estimated retrievability and a retention target rather than applying one fixed interval ladder to every card.

That distinction matters because review workload is a tradeoff. A higher target retention means cards generally need to return sooner and more often; a lower target can reduce scheduled reviews but allows more forgetting. The right workload therefore depends on the learner and the material rather than a universal number of reviews per day.

Why this is useful for Korean

1. Korean vocabulary accumulates quickly

A learner eventually needs vocabulary across many situations: greetings, food, transport, work, relationships, abstract topics, and whatever they personally read or listen to. TOPIK also rewards broad reading and listening vocabulary, but there is no official rule saying that a particular TOPIK level requires exactly N memorized words.

Instead of chasing a magic vocabulary count, use your course level, real comprehension gaps, and official TOPIK materials to decide what deserves attention next.

2. Word families can reinforce each other

Many Korean words contain Sino-Korean morphemes associated with Hanja. For example, 학 (學) is related to learning or study in words such as:

• 학교 — school
• 학생 — student
• 학습 — learning / study

Seeing related words across separate reviews can help you notice recurring structure. The shared syllable is a clue, not a license to guess the meaning of every unfamiliar word, so confirm new vocabulary in context.

3. Your own mistakes reveal what needs more review

The most valuable “confusable words” are not the same for every learner. Your review history can reveal that you repeatedly mix up two meanings, forget a particular ending, recognize a word but cannot produce it, or understand it in text but miss it in audio.

A scheduler can then return those weak items more often instead of forcing every word through the same amount of practice.

Building a sustainable review routine

Choose a new-card load from your actual workload

There is no evidence-based universal quota such as “beginners should learn 10–15 cards per day.” A sustainable number depends on the difficulty of the material, how much time you have, and how many reviews your existing cards already generate.

A better rule is:

• start with a deliberately small stream of new material
• complete the reviews already due
• watch whether the queue and study time remain comfortable over several days
• reduce new material if reviews begin to pile up
• increase gradually only when the routine is still sustainable

Consistency is usually more useful than repeatedly creating large backlogs and then trying to clear them in one session.

Give each card a clear retrieval target

A card does not have to contain only one visible word, but it should make the question clear. For example:

Recognition card
Front: 감사합니다
Back: thank you / gratitude expression + audio and a short example

Production card
Front: “Thank you.” (polite/formal context)
Back: 감사합니다 + audio

These test different skills. Keeping the target clear makes review results easier to interpret than placing several unrelated facts on one card and deciding that the whole card was “correct.”

Add context when context changes the answer

Korean vocabulary and grammar are sensitive to speech level, collocation, and situation. A short example can prevent you from memorizing an English gloss as if it were a perfect one-to-one translation.

Audio is particularly useful when the retrieval target includes listening or pronunciation, because romanization cannot represent every Korean contrast reliably.

Rate the recall you actually had

A scheduler can only learn from the feedback you give it. If you could not retrieve the answer, record a failure rather than choosing a passing rating because the answer “looked familiar” after you revealed it. If recall succeeded but was difficult, use the rating semantics the product provides for that state.

If you miss a day, resume the due reviews instead of treating the backlog as a reason to abandon the system. Adjust the flow of new material if necessary.

Spaced repetition in Chamelingo

Chamelingo integrates review with the learning path instead of treating flashcards as a separate pile of vocabulary. Course material can return through scheduled review, and the app uses FSRS-5 for course review scheduling.

Depending on the study surface, review can include vocabulary, grammar, recognition, production, and audio rather than only English-to-Korean word pairs. The useful metric is not how many cards you can click through; it is whether previously learned Korean remains retrievable when you need it again.

You can preview the material behind that review system in the free vocabulary browser and grammar reference.

Further reading

• Anki manual: FSRS, desired retention, and workload
• Cepeda et al. review of distributed practice
• McDermott review of retrieval practice

The research supports spacing and retrieval as useful learning principles. It does not support one magic interval sequence, one required daily card count, or one forgetting percentage that applies to every learner.