Spaced Repetition: The Complete Guide

The Forgetting Curve

In the 1880s, Hermann Ebbinghaus memorized nonsense syllables and measured his own forgetting. The result — the forgetting curve — showed memory decays steeply within hours, then progressively slower. Modern replications confirm the shape. The practical takeaway is not the decay itself but its cure: every retrieval at the point of near-forgetting strengthens the memory far more than a retrieval while the material is fresh.

This is the spacing effect, and it may be the most replicated finding in all of cognitive psychology. Cepeda and colleagues' 2006 meta-analysis aggregated 254 experiments and 14,000+ participants: spaced practice beat massed practice (cramming) in essentially every condition studied.

Why It Works

Two mechanisms, both well-supported:

Study-phase retrieval. Each spaced revisit forces your brain to retrieve rather than recognize. Retrieval is the exercise; recognition is watching someone else lift.
Encoding variability. Sessions spread across different days, moods, and contexts create multiple retrieval cues — more doors into the same memory.

Cramming feels effective because the material stays in working memory the whole session. The test comes a week later, after the steep part of the curve. Spaced learners are playing a different game: deliberately letting a little forgetting happen, then retrieving through it.

The Optimal Intervals

Research gives ranges, not exact numbers — they depend on material difficulty and your retention target. A practical, evidence-aligned schedule:

| Review | When | What it does | |---|---|---| | 1st | ~24 hours after learning | Catches the steepest decay | | 2nd | ~3 days | First true spaced retrieval | | 3rd | ~1 week | Consolidation phase | | 4th | ~2 weeks | Long-term strengthening | | 5th+ | ~monthly | Maintenance |

The rule of thumb: expand the interval each time you succeed; contract it when you fail. This is the engine behind Anki's SM-2 algorithm and every modern SRS app.

What Spaced Repetition Is For (and Not For)

For: facts, definitions, syntax, formulas, vocabulary — anything you must produce from memory. Language learning is the canonical case; medical school is the extreme one.

Not sufficient alone for: skills. You cannot spaced-repetition your way to writing React components — skills need deliberate practice with feedback, not recall of flashcards. But skills sit on a substrate of facts and patterns, and spacing that substrate makes the skill practice faster. The combination — spaced recall of fundamentals + project-based application — is the complete system (we cover the skill side in how to learn any skill fast).

Building Your System

Atomic facts only. One fact per card: "What does dict.get(k, default) return when k is absent?" — not "explain dictionaries."
Write cards yourself. Formulating the question is itself an encoding event; borrowed decks skip it.
Always attempt retrieval. Flipping the card without a genuine attempt trains nothing.
Daily, small. 15-20 minutes daily beats 2 hours weekly — that is the whole point.
Prune ruthlessly. A card you've nailed five times is retired; keep the deck about your current weak edges.
Pair with application. Every week, use the fact-space in a real project; application is the most realistic retrieval practice there is.

When a tool automates this loop — quizzing you on exactly the concepts your answers show you're weakest in — the interval management disappears into software. That is what adaptive quizzes inside our learning roadmaps do: the schedule expands and contracts per concept, per learner, without a deck to maintain.

The Evidence, Briefly

Ebbinghaus (1885): the forgetting curve
Cepeda et al. (2006), meta-analysis of 254 studies: spacing effect robust across all conditions
Roediger & Karpicke (2006): retrieval practice beats restudy substantially at a one-week delay
Karpicke & Bauernschmidt (2011): repeated retrieval, not repeated exposure, drives retention

One honest caveat: the effect sizes come largely from lab studies of verbal material; applied skill-learning inherits the principles but with looser numbers. The direction of the evidence, however, is unanimous: space it, retrieve it, or lose it.