Most people have experienced this: you study a word, feel confident you know it, and then completely blank on it three days later. You're not forgetful by nature. You're experiencing one of the most well-documented phenomena in cognitive science β the Ebbinghaus Forgetting Curve.
Understanding why you forget, and how modern spaced repetition algorithms counteract it, will fundamentally change how you approach vocabulary learning.
The Ebbinghaus Forgetting Curve
In the 1880s, German psychologist Hermann Ebbinghaus conducted a remarkable series of experiments on his own memory. He memorized hundreds of nonsense syllables and tested how quickly he forgot them. His findings, published in 1885, remain among the most cited results in all of psychology.
Ebbinghaus discovered that forgetting is not gradual β it is exponential. Within 20 minutes of learning something new, you've typically forgotten 42% of it. Within an hour, 56%. Within a day, 74%. After a month with no review, the information is almost entirely gone.
This curve isn't a personal failing. It's how human memory works by design. Your brain continuously evaluates what information is worth retaining and discards what it hasn't used recently. It's an energy-saving mechanism β and it works against language learners.
The Key Insight: Retrieval Strengthens Memory
Ebbinghaus also discovered the solution. Each time you successfully retrieve a memory β each time you see a word and recall its meaning β you reset the forgetting curve and it decays more slowly next time. This is called the testing effect or retrieval practice effect.
The implication: reviewing a word at exactly the right moment β just before you would have forgotten it β is dramatically more effective than reviewing it randomly or too frequently. This is the principle behind spaced repetition.
Why Traditional Flashcard Apps Fall Short
The original spaced repetition system, SM-2 (developed in the 1980s), was a major advance over random review. It assigned cards to fixed review intervals β 1 day, 6 days, 15 days, and so on β based on how you rated your recall.
Apps like early versions of Anki used SM-2. And it worked β much better than no system at all.
But SM-2 has significant limitations:
1. Fixed intervals ignore individual memory profiles. SM-2 uses the same schedule for everyone. But people forget different things at different rates. A word you've seen in 20 different contexts will stick far longer than an obscure technical term you've encountered once.
2. It doesn't model stability accurately. SM-2 treats each correct recall as roughly equal, regardless of how many times you've reviewed the card or how much time has passed. In reality, a correct recall after 30 days provides much stronger evidence of memory consolidation than a correct recall after 1 day.
3. It can't easily adapt. If you miss a week of study, SM-2-based apps flood you with overdue cards using schedules that no longer match your current retention state.
The result: you either over-review cards you actually know well (wasting time) or under-review cards your memory needs to see again (leading to forgotten words).
How FSRS Solves This
FSRS (Free Spaced Repetition Scheduler) is a modern algorithm developed by Jarrett Ye and published in 2022. It represents a fundamental rethinking of spaced repetition based on contemporary memory research.
FSRS models memory using two core parameters for each card:
Stability (S)
Stability represents how long you can remember a card before your retention drops to a target threshold (typically 90%). A card with high stability might not need review for 60 days. A card with low stability might need review in 3 days. Stability increases with each successful review and grows faster after longer review intervals β a phenomenon called memory consolidation.
Difficulty (D)
Difficulty captures how intrinsically hard a particular piece of information is to learn and retain for you personally. FSRS learns each card's difficulty over time by analyzing your review history. A word you keep getting wrong accumulates a high difficulty score, and FSRS schedules it more frequently as a result.
How FSRS Calculates Review Intervals
For each card, FSRS computes the exact interval that gives approximately a 90% probability of recall at review time. This is called the desired retention target. The calculation uses both stability and difficulty, producing intervals tailored to that specific card for that specific learner.
The math behind FSRS is grounded in the Three-Component Model of Memory (DSR model), which describes forgetting in terms of:
- Desired retention (R): your target recall probability
- Stability (S): how stable the memory is
- Difficulty (D): how hard the card is intrinsically
By computing personalized intervals from these three values, FSRS escapes the fixed-schedule limitation of SM-2.
The Numbers: FSRS vs. Traditional Spaced Repetition
Research comparing FSRS to SM-2 and similar algorithms consistently shows significant efficiency gains. Published studies and community benchmarks suggest FSRS requires 10β20% fewer total reviews to achieve the same retention rate as SM-2.
To put that in concrete terms: if SM-2 requires 1,000 total reviews to bring a 500-word vocabulary to 90% retention, FSRS achieves the same result with approximately 800β900 reviews. Over months of study, this compounds into hours saved β without sacrificing what you know.
The efficiency gain is larger for learners with heterogeneous vocabularies (some very easy words, some very hard ones), because FSRS can allocate review time much more precisely. Easy words get long intervals; hard words get frequent review. SM-2 is blunter.
How Voccle Uses FSRS
Voccle integrates FSRS as its default scheduling algorithm. Every card you create β whether typed manually, imported from a library deck, or generated from a pasted article using the AI extraction tool β is automatically managed by FSRS.
You don't need to configure anything. The algorithm runs in the background:
- When you create a card, it enters the "new" state.
- When you study it for the first time, FSRS records your response (Again / Hard / Good / Easy) and calculates the initial stability estimate.
- Each subsequent review updates both the stability and difficulty estimates based on whether you recalled it correctly and how confident you were.
- Each morning, Voccle shows you exactly the cards due for review β the ones whose predicted retention has dropped to the target threshold. No more, no less.
The result is a learning system that works with your brain's natural forgetting curve rather than against it.
The Bottom Line
Forgetting is not a bug β it's how memory works. The Ebbinghaus Forgetting Curve is unavoidable. But by reviewing words at precisely the right moment, using an algorithm that models your individual memory profile, you can build long-term vocabulary retention with dramatically less wasted effort.
FSRS is the best available implementation of this science. And it's built into Voccle β completely free.