Neuroscience Fundamentals
Human memory does not work like a computer hard drive; it operates via synaptic plasticity and the spacing effect. By testing recall at the precise mathematical moment when memory is about to fade, you consolidate short-term synaptic connections into durable, permanent neural pathways.
1. The Forgetting Curve and the Spacing Effect
In 1885, German psychologist Hermann Ebbinghaus published his landmark monograph on memory. By testing his ability to memorize thousands of nonsensical syllables, Ebbinghaus discovered what we now revere as the Forgetting Curve: within forty-eight hours of learning a new piece of information, the average human brain forgets approximately 70% to 80% of the material if no active review takes place.
However, Ebbinghaus also made a revolutionary discovery: each time you successfully retrieve a memory from active recall, the rate of forgetting drastically flattens. A concept reviewed at day 1, day 4, day 12, day 35, and day 120 becomes practically permanent. This is known as the Spacing Effect.
Figure 1: Mathematical graph of memory decay: Standard Forgetting Curve vs. the flattened trajectory of Spaced Repetition.
2. The Algorithmic Leap: SM-2 vs. Modern FSRS-5
For over thirty years, spaced repetition software—including early versions of Anki—relied on the SM-2 algorithm, developed by Polish researcher Piotr Wozniak in 1987. While SM-2 was revolutionary for its era, it treated memory difficulty with a crude static multiplier (Ease Factor) that frequently caused Ease Hell: cards marked 'Hard' became trapped in an endless cycle of frequent, exhausting reviews.
Enter the Free Spaced Repetition Scheduler (FSRS). Developed through machine learning trained on over one hundred million user review logs, FSRS-5 models memory using the Three-Component Model of Memory:
- Stability (S): The time in days required for memory retention to drop from 100% to 90%.
- Difficulty (D): The inherent cognitive complexity of the item (rated from 1 to 10).
- Retrievability (R): The instantaneous probability of successfully recalling the item right now.
Algorithm Showdown: SM-2 vs SuperMemo-18 vs FSRS-5 vs Leitner Box
| Algorithm |
Model Origin |
Efficiency Gain |
Custom Parameter Optimization |
Risk of 'Ease Hell' |
| Leitner Box |
Physical flashcard boxes (1972) |
Baseline |
None (Fixed box intervals) |
High (manual friction) |
| Anki SM-2 |
Heuristic Formula (1987) |
+15% vs Leitner |
Manual interval modifiers |
Very High (punishing 'Hard' button) |
| FSRS-5 (Anki 23.10+) |
Machine Learning / DSR Model |
+35% to +45% review efficiency |
Automatic per-user optimization |
Eliminated completely |
Figure 2: Card Architecture Teardown: Why atomic cloze deletions result in 95% retention with half the daily study time.
3. The Golden Rule: The Minimum Information Principle
The most common failure mode among medical students, law scholars, and language learners is crafting bloated flashcards. When a card asks five separate questions or contains a dense paragraph of text, your brain takes mental shortcuts, recognizing superficial cues rather than testing deep recall.
Follow Piotr Wozniak's classic Minimum Information Principle:
- One Atomic Fact Per Card: If a concept has three components, create three distinct cloze deletion cards.
- Understand Before Memorizing: Never make flashcards for material you do not thoroughly comprehend conceptually.
- Use Imagery & Context: Anchor abstract formulas to real-world visual demonstrations or historical mnemonics.
External Video Resource: Cognitive Neuroscience of Memory Retention
Frequently Asked Questions (FAQ)
What is the optimal target retention percentage in FSRS?
For most learners, a target retention of 85% to 90% provides the best balance between retention and workload. Setting retention to 95% or higher exponentially increases daily reviews for minimal marginal gain.
How many new cards should I introduce each day?
We recommend starting with 10 to 20 new cards per day. Remember that each new card generates review obligations across subsequent weeks; introducing 50 new cards a day will quickly generate 300+ reviews daily.
Related Learning & Education Guides