Spaced repetition is the foundation of long-term vocabulary acquisition. For nearly four decades, the language learning community has relied on variations of the SM-2 algorithm, originally developed by Piotr Woźniak in 1987 and popularized by Anki.
In recent years, cognitive science and data modeling have produced a major breakthrough: the Free Spaced Repetition Scheduler (FSRS). Here is why FSRS represents a generational leap over SM-2, and why it is uniquely suited for vocabulary acquired through real video immersion.
The Problem with SM-2: “Ease Factor Hell”
The classic SM-2 algorithm models memory using a single dynamic multiplier called the Ease Factor (typically starting at 250%):
- When you answer a card correctly, the next interval is multiplied by the Ease Factor.
- When you fail a card (“Again”), the interval resets to day 1, and the Ease Factor decreases by 15-20 percentage points.
- If you repeatedly struggle with a difficult phrase, the Ease Factor drops to its minimum (130%).
This creates a well-known flaw known among Anki power users as Ease Factor Hell: once a card becomes difficult, its multiplier remains permanently depressed. Even after you master the word months later, the algorithm continues scheduling it every few days, creating a mountain of redundant reviews that leads to study burnout.
| Dimension | Legacy Anki SM-2 (1987) | Modern FSRS (2020s) |
|---|---|---|
| Theoretical Foundation | Heuristic multipliers | DSR (Difficulty, Stability, Retrievability) model |
| Memory States | Binary: learned or unlearned | Continuous probability of recall curve |
| Failure Penalty | Permanently drops Ease Factor | Recalculates stability without ruining future growth |
| Interval Precision | Fixed mathematical intervals | Dynamically tuned to your target retention (e.g., 90%) |
| Overdue Handling | Rigid penalties for late reviews | Dynamically adjusts for elapsed time without card corruption |
| Review Efficiency | High review fatigue over time | 20% to 30% fewer reviews for the same retention rate |
How FSRS Models Human Memory
FSRS does not use an arbitrary multiplier. Instead, it measures memory across three distinct psychological variables known as the DSR Model:
- Difficulty ($D$): How inherently difficult the information is to grasp (from 1 to 10). A short cognate like “coffee” has low difficulty; an idiomatic compound like “一期一会” has higher initial difficulty.
- Stability ($S$): The time (in days) required for the probability of recall to drop from 100% to 90%. Stability increases with each successful retrieval.
- Retrievability ($R$): The real-time probability that you can recall the item right now. As time passes since your last review, retrievability declines along an exponential forgetting curve.
Instead of guessing intervals, FSRS calculates the exact day when retrievability drops to your target threshold (typically 90%). If you successfully recall a card when its retrievability has dropped to 85%, FSRS recognizes that your memory performed under pressure and awards a significant boost to its stability.
Why Video Vocabulary Requires FSRS
When you learn words from a vocabulary list, cards are isolated and uniform. But when you acquire vocabulary from authentic YouTube videos—such as Japanese vlogs or English interviews—the words arrive in rich, chaotic contexts:
- Varying Initial Exposure: Some words you hear five times in one vlog; others appear once in passing. FSRS accurately models this variance through its adaptive difficulty parameter.
- Natural Contextual Anchors: A word attached to a facial expression, a voice tone, and a storyline forms multiple neural pathways. With FSRS, you do not need hundreds of reviews to cement a vivid video moment into memory.
- Irregular Study Habits: Real life involves missed days. With SM-2, missing a weekend can flood your queue with hundreds of overdue cards. FSRS recognizes that extra time elapsed since the last review actually provides stronger memory reinforcement when recalled, preventing overdue card panic.
How LingoDew Implements FSRS
In LingoDew, every looked-up word flows directly into a native mobile FSRS scheduler:
- When you tap a word in a subtitle, LingoDew captures the headword, canonical reading, translated sentence, and exact audio timestamp.
- During review, you grade cards across four intuitive buttons: Forgot, Hard, Remembered, or Easy.
- Words advance through structured mastery phases—from Unseen and Queried into Learning, and ultimately into Mastered.
- When a word reaches Mastered, its active slot is permanently released on the free tier, allowing you to continually absorb new vocabulary.
By combining authentic video input with the statistical rigor of FSRS, you eliminate both textbook boredom and flashcard burnout.