The Science of Spaced Repetition in 2026: What's New

Spaced Repetition: A Quick Refresher

Spaced repetition is a learning technique where you review information at progressively increasing intervals. Instead of cramming everything in one session, you spread reviews over time — hitting each item just before you're about to forget it. This approach is backed by over a century of memory research and powers every serious flashcard app today.

But the science hasn't stood still. Recent advances in cognitive psychology, neuroscience, and machine learning have refined our understanding of how memory works — and how to exploit it for faster, more durable learning.

The Ebbinghaus Forgetting Curve: Updated for 2026

Hermann Ebbinghaus established the forgetting curve in 1885, showing that memory decays exponentially after learning. For over a century, this model remained largely unchanged. Recent studies have added important nuances:

• Individual variability: Forgetting curves differ significantly between people. A one-size-fits-all interval schedule leaves some learners reviewing too early (wasted time) and others too late (forgotten material).
• Material-dependent decay: Complex, meaningful material (like kanji with mnemonic stories) decays slower than arbitrary associations. This is why contextual learning outperforms rote memorization.
• Emotional encoding boost: Items learned with emotional engagement (surprise, humor, personal connection) show flatter forgetting curves — they stick longer.
• Sleep-dependent consolidation: The first sleep cycle after learning dramatically reshapes the forgetting curve. Items reviewed before sleep show up to 20% better next-day retention.

Desirable Difficulty Theory

Robert Bjork's "desirable difficulty" framework has become increasingly influential. The core principle: learning that feels harder in the moment leads to stronger long-term retention.

What does this mean for kanji study?

• Struggling to recall is good. When you pause for 3-5 seconds trying to remember a kanji reading, that retrieval effort strengthens the memory trace far more than instant recognition.
• Spacing is a desirable difficulty. Reviewing at longer intervals feels harder (you've forgotten more), but each successful retrieval builds stronger memory.
• Interleaving is a desirable difficulty. Mixing N5 and N4 kanji in reviews (rather than studying one level at a time) makes each review harder but improves discrimination and long-term recall.

Modern SRS algorithms are designed to keep you in the "desirably difficult" zone — not so easy that you're wasting time, not so hard that you're failing constantly.

Interleaving vs Blocking

Traditional study often uses blocking: study all N5 kanji, then all N4, then N3. It feels productive because each individual session is easier. But research consistently shows that interleaving (mixing categories) produces 20-40% better long-term retention.

Why? Interleaving forces your brain to:

• Distinguish between similar items (critical for kanji that look alike)
• Practice retrieval from long-term memory, not just short-term
• Build flexible, transferable knowledge rather than rigid category-specific recall

Kanjijo's SRS naturally interleaves your reviews — mature N5 kanji appear alongside newly learned N3 kanji, keeping your brain in discrimination mode.

Retrieval Practice: The Strongest Learning Strategy

Retrieval practice — the act of pulling information from memory rather than re-reading or re-watching — is now considered the single most effective learning strategy by cognitive scientists. The "testing effect" has been replicated in hundreds of studies.

Key findings for 2026:

• Retrieval strength vs storage strength: You can "know" something (high storage strength) but be slow to access it (low retrieval strength). SRS specifically trains retrieval strength.
• Failed retrievals still help: Even when you can't recall a kanji, the attempt strengthens the memory pathway. This is why "hard" ratings matter more than "again" in SRS apps.
• Multi-modal retrieval is superior: Recalling meaning, reading, and stroke order activates different memory pathways. Apps that test multiple aspects of each kanji produce stronger overall encoding.

Sleep and Memory Consolidation

Memory research in the 2020s has deepened our understanding of sleep's role in learning:

• Sleep spindles (Stage 2 NREM) replay recently learned information, transferring it from hippocampal short-term storage to cortical long-term memory.
• Slow-wave sleep strengthens declarative memories — facts, vocabulary, and character recognition (exactly what kanji study requires).
• REM sleep integrates new knowledge with existing memories, helping you recognize patterns (like shared radicals across kanji).
• Pre-sleep review window: Studying 30-60 minutes before sleep consistently shows enhanced next-day retention compared to morning study of the same material.

Optimal Spacing: Expanding vs Fixed Intervals

A long-standing debate in SRS research: should intervals expand (1 day → 3 days → 7 days → 14 days) or stay fixed (review every 3 days)?

The current consensus in 2026:

• Expanding intervals work best for initial learning. When you first encounter a kanji, gradually increasing spacing builds long-term retention efficiently.
• Fixed intervals work better for maintenance. Once a kanji is well-learned (mature card), reviewing at consistent long intervals prevents decay without overloading your schedule.
• Hybrid approaches win: The best algorithms use expanding intervals for new/unstable items and stabilized intervals for mature items. This is exactly what modern adaptive SRS implementations do.

AI-Adaptive Algorithms: SM-2 vs FSRS vs Kanjijo

The SRS algorithm landscape has evolved dramatically. Here's how the major approaches compare:

SM-2 (SuperMemo 2, 1987)

The algorithm that started it all. Used by Anki and many early SRS apps. It assigns each card an "ease factor" that increases or decreases based on your rating. Simple, transparent, and effective — but treats all learners identically and doesn't model memory states accurately.

FSRS (Free Spaced Repetition Scheduler, 2023+)

A breakthrough open-source algorithm developed by Jarrett Ye. FSRS models memory with two variables: stability (how long until you forget) and retrievability (probability you remember right now). It uses machine learning to fit parameters to your individual review history, achieving better retention with fewer reviews than SM-2.

Kanjijo's Adaptive Algorithm

Kanjijo builds on insights from both SM-2 and FSRS research, with optimizations specific to kanji learning:

• Individual card modeling: Each kanji has its own difficulty profile based on your performance history with similar characters
• Radical-aware scheduling: Kanji sharing radicals are spaced to leverage pattern recognition without creating interference
• Passive exposure integration: Widget views are factored into the scheduling model — if you've seen a kanji on your lock screen multiple times, the algorithm knows your retrievability is higher
• Adaptive difficulty calibration: The algorithm continuously adjusts its model of your personal forgetting curve as it collects more data

Metacognitive Monitoring

One of the most exciting developments in learning science is the focus on metacognition — your ability to accurately judge what you know and don't know. Research shows that learners are generally poor at this: we overestimate how well we know material we've just reviewed (the "fluency illusion").

SRS systems help by making metacognitive judgments explicit — you must rate how well you knew each card. But there's a catch: learners tend to over-rate their performance, leading to intervals that are too long.

Modern approaches address this by:

• Using response time (not just correctness) as a signal — a 10-second correct recall indicates weaker memory than a 1-second recall
• Implementing "hard" vs "good" vs "easy" gradations that capture confidence levels
• Tracking patterns of self-rating accuracy and calibrating accordingly

The Future: Brain-Computer Interfaces and SRS

Looking ahead, the frontier of SRS research intersects with neurotechnology:

• EEG-guided scheduling: Early research (2024-2026) uses consumer-grade EEG headbands to detect when memory consolidation is occurring, potentially timing reviews to biological readiness.
• Attention detection: Eye-tracking and engagement metrics could tell SRS algorithms whether you truly processed a review or just mindlessly tapped through it.
• Personalized forgetting curve estimation: Rather than inferring forgetting from behavioral data (reviews), direct neural measurement could provide ground-truth forgetting rates.

These technologies are still experimental, but they point toward a future where SRS algorithms have access to far richer data about your learning state. For now, behavioral-based algorithms like Kanjijo's represent the state of the art — and they're remarkably effective.

What This All Means for Your Kanji Study

The science boils down to a few actionable principles:

• Show up daily. Consistency matters more than session length. SRS works because of repeated, spaced encounters over time.
• Embrace difficulty. If reviews feel challenging, that's the algorithm working correctly. Easy reviews are wasted reviews.
• Rate honestly. When the SRS asks how well you knew a card, be truthful. The algorithm can only optimize your schedule with accurate data.
• Leverage passive exposure. Use Kanjijo's widgets to multiply your daily kanji encounters without adding study time.
• Sleep on it. A pre-bed review session isn't lazy — it's scientifically optimal.
• Trust the algorithm. Modern SRS scheduling is based on decades of research. Let it manage your intervals instead of second-guessing with manual cramming.