1. The core insight
Spaced repetition rests on Hermann Ebbinghaus's 1885 work on the forgetting curve: memory of new content decays roughly exponentially. But if you review just before the curve hits a critical low, retention bounces back stronger than before.
After several well-timed reviews, intervals grow longer and longer while retention stays high:
| Review # | Interval since previous | Retention after this interval |
|---|---|---|
| 1 | 1 day | ~50% |
| 2 | 3 days | ~60% |
| 3 | 7 days | ~75% |
| 4 | 15 days | ~85% |
| 5 | 30 days | ~92% |
| 6 | 90 days | ~95% |
Numbers are illustrative averages across studies. The shape — intervals growing while retention rises — is the real point.
2. How SM-2 works
The SM-2 algorithm, published by Piotr Wozniak in 1985, manages two numbers per item:
- Interval — days until the next review
- Ease — a multiplier between 1.3 and 2.5
After each review you rate yourself 0–3:
- 0 = no idea
- 1 = vague recall
- 2 = correct but with effort
- 3 = correct, easy
Higher ratings push Ease up and lengthen the next Interval. Low ratings push Ease down and reset the Interval to 1 day. Our learning plan uses this exact algorithm.
3. Where SRS helps English learners most
Vocabulary
The classic use case. Add a word; review by SM-2 schedule. Our tools auto-add to your SRS queue when you get a dictation wrong.
Chunks and collocations
More efficient than single words. Memorising "address concerns" as a unit retains ~30% better than memorising "address" and "concerns" separately.
Personal grammar mistakes
Take the exact sentence in your writing that contained an error. Reverse-side: corrected version. Add to SRS. Reviewing your own errors beats reviewing a textbook's example errors.
Listening difficulties
Specific phonetic patterns that catch you (e.g. /r/ + /l/ across word boundaries) — capture the sentence and review it through SRS. Hearing it every few days normalises the pattern.
4. Five common mistakes
- Too many new items per day. Cap at 15 new cards. More creates a review queue that becomes hateful within two weeks.
- Skipping hard cards. These are the ones that need the most repetition. Trust the algorithm to manage them.
- Chasing zero queue. It's OK to leave 10 reviews undone — 7 days of doing 30 is better than 1 day of doing 50.
- Reviewing too fast. Spend at least 5 seconds per card. Skim-reviewing burns cards without deepening memory.
- Adding cards before initial encoding. First meet a word in context — read it, hear it, use it once. Then add to SRS. Dropping cold definitions onto the queue produces low retention.
5. Using our SRS
- Open the study plan for today's due count
- Click "Start review" for a continuous session
- Rate each card honestly (0–3)
- Finish and return to the plan
Daily 30-minute pattern:
- 10 min SRS clearing due items
- 15 min new content (dictation / listening / cloze)
- 5 min passive listening or free reading
6. What SRS is not good at
SRS handles declarative memory: vocabulary, collocations, grammar rules. It does not directly train:
- Fluency — that's procedural memory, built by using language, not by recalling it
- Listening speed — needs volume of authentic input
- Writing thought — needs deliberate writing with feedback
Treat SRS as a maintenance system, not a complete teacher.
7. The compound effect
20 minutes of SRS daily for 6 months gives you, on average:
- +1500–2000 active vocabulary
- +3000–4000 passive vocabulary
- ~85% retention three months after the last review
This is the kind of gain no cramming session ever delivers. The trick is to make 20 minutes daily non-negotiable.