The forgetting curve
Retention halves in a day. Then again. Then again.
definition
The exponential decay of memory over time in the absence of review.
The forgetting curve is Hermann Ebbinghaus's 1885 finding that human memory decays exponentially — not linearly — in the absence of review. It was one of the first quantitative findings in experimental psychology and remains one of the most-replicated results across the history of the field.
The specific shape: retention drops from ~100% at learning to ~40% within 24 hours, then more slowly to ~15-20% at one week, converging on a low asymptote thereafter. The initial drop is steep; the later decay is shallow. This produces the counter-intuitive property that review timing matters enormously — a review at day 1 buys you much more retention than a review at day 7.
The curve isn't a psychological quirk. It's a biological one. Memory consolidation is a physical process involving synaptic strengthening in the hippocampus and gradual transfer to the neocortex. Without reactivation, the initial trace decays before consolidation completes. With well-timed reactivation, the trace strengthens and the curve resets.
Every card in a Brainback recall queue is scheduled based on this curve. The SM-2 algorithm approximates the timing that catches memory right before decay drops below ~85% retention. Each catch stretches the next decay window. Over months, this produces retention that would be impossible with any other schedule — 90%+ retention on cards you first learned six months ago, with 5-10 minutes of daily maintenance.
What the research says about the forgetting curve.
Three papers that established the finding.
Every design decision Brainback makes about the forgetting curve flows from these papers. Cited so you can verify.
Memory: A Contribution to Experimental Psychology
Original discovery. Ebbinghaus memorized lists of nonsense syllables and tested his own retention at intervals. The exponential decay curve was the result. The methodology (self-experimentation, quantitative retention measurement) became the model for a century of memory research.
Replication and Analysis of Ebbinghaus' Forgetting Curve
Modern replication of Ebbinghaus using digital methods. Confirmed the exponential shape with a decay constant closely matching the 1885 original. The forgetting curve holds 130 years later.
The Fate of First-List Associations in Transfer Theory
Established that the forgetting curve is not just decay but also interference — later learning interferes with earlier learning. This is why review is more effective than initial learning: it distinguishes the target memory from interfering ones.
How it actually works, at the cognitive and neural level.
The mechanism isn’t decorative. Understanding WHY the effect exists lets you tell when it applies and when it doesn’t.
The biological mechanism is synaptic decay. Freshly formed memories exist as strengthened connections between specific neurons in the hippocampus. Without reactivation, these connections gradually weaken back toward baseline over days to weeks — the specific timescale depends on the memory type, the encoding depth, and the presence of interference from other memories.
Consolidation is the parallel process that gradually transfers the memory from hippocampus to distributed neocortical networks. This transfer takes weeks to years depending on the memory. During the consolidation window, retrieval and reactivation are critical — a memory that's retrieved during consolidation ends up more robustly encoded in the neocortex; a memory that isn't gets pruned.
The exponential curve emerges from the interaction of these two processes. Right after learning, hippocampal decay is dominant; retention drops fast. As consolidation progresses, neocortical representation grows and stabilizes; the decay rate slows. Reviews at the right moments accelerate consolidation and interrupt decay — which is why the timing matters and why massed review is far less effective than spaced review.
Same scenario. Two study strategies.
Foreign language · Spanish verb conjugations.
You just learned 20 Spanish verbs and their present-tense conjugations.
You don't review them. Next day: you remember ~40% (Ebbinghaus's number for nonsense syllables; slightly higher for meaningful material). One week later: ~15-20%. Two weeks: ~10%. You've effectively lost the entire lesson.
You review at day 1 (retention: 80% before review → 95% after). Day 3 (75% → 95%). Day 7 (85% → 98%). Day 21 (90% → 99%). Day 60 (95% → 100%). Six months later: still at ~90% retention with a single 5-minute review.
Same 20 verbs. Vastly different retention curves. Without review: gone in 2 weeks. With SM-2 review: fluent for a year and beyond, with 5-10 minutes of daily maintenance total. This is the compounding that makes long-term retention feasible.
The method, made operable.
Every card in your recall queue is scheduled to return just before Ebbinghaus's curve would drop it below ~85% retention. That's the entire logic of SM-2: catch the memory before it decays, and each catch stretches the next decay window.
Three subject areas where the effect is strongest.
The method works everywhere. It works especially well in these three domains — for reasons specific to each.
The forgetting curve is steepest on isolated word-meaning pairs. Spaced review is the difference between a fluent speaker and a course-completer.
200+ drug facts to hold through Step 1. Without the curve-flattening effect of SM-2, cramming is the only alternative, and cramming fails at high volume.
Discrete facts + relationships. Ebbinghaus's original nonsense syllables were closest to this material type; the finding applies most directly.
Three weaker approaches — and why they underperform.
Understanding what the method beats (and by how much) tells you when to reach for it.
why weaker · Fights only the last 24 hours of decay. Everything from 3+ days ago has already fallen off the useful part of the curve.
why weaker · Self-assessment of memory strength is unreliable. Students consistently over-estimate what they know. The algorithm doesn't have that bias.
why weaker · Massed review is less effective than distributed review, even at matched total time. The spacing effect is real.
How to actually do it tomorrow morning.
5-15 minutes daily to maintain a queue of 100-500 cards. The steps below are the sustainable protocol — not a heroic one-off.
Trust the algorithm. Don't try to out-guess what to review — it's calibrated to your specific curve.
Review at approximately the same time daily. Consistency matters more than duration.
Grade honestly. If you had to think, that's Hard, not Good.
Don't skip more than 3 days in a row. Long gaps let cards drift back to shorter intervals.
Don't over-review. If the queue is empty for the day, stop — grinding old cards defeats the spacing effect.
Trying to defeat the curve by studying harder the day before an exam. The curve is exponential; the last-day cramming only fights the last 24 hours of decay. Everything from 3+ days ago has already fallen off the useful part of the curve.
The mechanism, in one paragraph.
Memory is a physical process — synaptic strengthening decays without reactivation. The curve isn't a psychological quirk; it's the neurobiology of memory consolidation. Fighting it requires reactivation, and reactivation is timed.
Everything students ask about the forgetting curve.
Is the exact curve the same for everyone?▾
The shape is universal. The specific decay rate varies by individual + material + emotional salience. SM-2 personalizes to your specific curve via the ease factor.
How much can I flatten the curve?▾
Fully. Consistent SM-2 review keeps well-studied cards at >90% retention indefinitely. The curve doesn't 'break' — it just gets held above the drop-off zone.
Does sleep help?▾
Yes — memory consolidation happens during sleep, especially deep and REM stages. A well-rested review of yesterday's material is far more effective than a same-day repeat.
What about emotional or vivid memories?▾
Emotionally salient memories decay slower. This is why you remember where you were during major events; the amygdala strengthens the memory during encoding. Course material lacks this natural strengthening; scheduling substitutes.
How does the curve interact with sleep deprivation?▾
Badly. Consolidation is impaired without sleep. Cramming through the night trades short-term recall for long-term retention. Almost always a bad trade for an exam more than 48h out.
Three more methods that stack with this one.
the forgetting curve
Try the method in the product.
Spaced recall runs on every account, free tier included.