I've seen the Ebbinghaus forgetting curve in at least fifty productivity blog posts, twenty YouTube videos, and every other pitch deck for learning apps. The graph is always the same: a smooth exponential decay showing you forget 50% of new information within an hour, 70% within a day, and 90% within a week.
Here's the problem: Hermann Ebbinghaus never published that curve.
What he did publish in 1885 was far more interesting—and more useful for anyone building a spaced repetition practice. Let me show you what the data actually says.
TL;DR
Ebbinghaus tested himself on nonsense syllables and found memory decay follows a power law, not the exponential curve in most infographics. His real findings: forgetting is steepest in the first hour, individual differences matter enormously, and meaningful material behaves differently than random syllables. Modern research confirms the basic shape but shows huge variation based on what you're learning and how you encode it.
What Ebbinghaus Actually Did
In 1885, Hermann Ebbinghaus published Über das Gedächtnis (Memory: A Contribution to Experimental Psychology). He was 35, working alone in Berlin, and determined to measure memory with the same rigor physicists applied to motion and heat.
His method was brutal in its simplicity. He created 2,300 nonsense syllables—three-letter combinations like "WID," "ZOF," "KAJ"—specifically designed to have no meaning. He'd memorize lists of 13 syllables until he could recite them perfectly twice in a row, then test himself at fixed intervals: 20 minutes, 1 hour, 9 hours, 1 day, 2 days, 6 days, 31 days.
The metric wasn't "how many syllables can you recall?" It was "how many repetitions does it take to relearn the list perfectly?" If the original list took 30 repetitions to learn and relearning took 18 repetitions after one day, he'd saved 12 repetitions—a 40% savings.
This "savings method" is clever because it captures partial memory. Even if you can't consciously recall something, relearning it faster proves some trace remains.
The Real Curve: Power Law, Not Exponential
When Ebbinghaus plotted his data, he found that memory retention followed a power function:
b = 100k / (log t)^c + k
Where:
- b = savings (percent of effort saved on relearning)
- t = time since learning
- k and c = constants fitted to the data
In plain English: forgetting is fastest right after learning, then slows down. The curve is steep in the first hour, then flattens. After a month, you're still forgetting, but much more slowly.
Here are his actual numbers for a 13-syllable list:
- 20 minutes: 58% savings
- 1 hour: 44% savings
- 9 hours: 36% savings
- 1 day: 34% savings
- 2 days: 28% savings
- 6 days: 25% savings
- 31 days: 21% savings
Notice: you lose more in the first hour (14 percentage points) than you do between day 2 and day 31 (7 points). This is the core insight. The popular "you forget 70% in 24 hours" claim doesn't appear anywhere in Ebbinghaus's work.
Why the Myth Persists
The smooth exponential curve you see everywhere likely comes from a 1913 reinterpretation by psychologist Arthur Gates, who simplified Ebbinghaus's findings for teachers. The exponential form is easier to draw and explain, so it stuck.
But the difference matters. An exponential decay suggests forgetting happens at a constant rate—you lose the same percentage per unit time. A power law means forgetting slows down—the rate of loss decreases over time. That second model is what Ebbinghaus actually found, and it's what modern research continues to confirm.
What Modern Research Adds
Ebbinghaus's 1885 experiment had one subject (himself) learning one type of material (meaningless syllables). A century of follow-up research has tested thousands of people learning everything from Swahili vocabulary to organic chemistry mechanisms. Here's what we know now:
1. Meaningful Material Decays Slower
When Ebbinghaus tested himself on meaningful poetry instead of nonsense syllables, retention was dramatically higher. A 1979 replication by Bahrick and colleagues tracked people's memory for Spanish vocabulary over 50 years. They found:
- 3 years: ~60% retention
- 25 years: ~50% retention
- 50 years: ~30% retention
Compare that to Ebbinghaus's 21% savings after just 31 days with nonsense syllables. Meaning acts as a preservative.
This is why medical students using SmartRecall to learn pharmacology see better retention than Ebbinghaus's curve would predict—drug mechanisms connect to physiology, side effects connect to patient cases. Every connection is a retrieval path.
2. Individual Differences Are Enormous
Ebbinghaus was his own subject, so he couldn't measure variation between people. Modern studies show forgetting rates vary by a factor of 3-5x depending on:
- Prior knowledge: If you already know 2,000 Mandarin words, the 2,001st sticks faster than the first did.
- Encoding depth: Passive reading produces steeper forgetting than active recall or elaborative rehearsal.
- Sleep: A 2006 study by Stickgold found that memory consolidation during sleep can reduce forgetting by 30-50% compared to staying awake for the same interval.
When I look at SmartRecall's anonymized retention data across 40,000 users, the spread is massive. Some users retain 80% of new cards after 30 days; others are at 40%. The algorithm (FSRS) adapts to this, but the point is: there's no single forgetting curve.
3. Retrieval Practice Reshapes the Curve
Ebbinghaus measured passive decay—what happens when you learn something once and never think about it again. But every time you successfully retrieve a memory, you reset the curve at a higher starting point.
This is the foundation of spaced repetition. If you review a card on day 1, day 3, day 7, and day 21, you're not fighting the original forgetting curve four times. You're creating a new, shallower curve after each review. By the fourth review, the curve might be so flat that your next review isn't due for six months.
Bjork and Bjork (1992) called this "desirable difficulty." The harder you have to work to retrieve something (without failing), the more you strengthen it. This is why SmartRecall and other SRS tools deliberately wait until you're about to forget before showing you a card again.
Practical Takeaways for SRS Users
If you're using spaced repetition to study for the USMLE, learn Japanese for the JLPT N3, or master React hooks, here's what Ebbinghaus's real findings mean for you:
Review Within 24 Hours
The steepest drop happens in the first day. If you learn 20 new anatomy terms on Monday morning, review them Monday evening or Tuesday morning—not Thursday. Most SRS algorithms (SM-2, FSRS) default to a 1-day first interval for exactly this reason.
In SmartRecall, we let you adjust this if you're cramming for an exam next week, but the default is evidence-based: catch cards while they're still fresh.
Meaningful Encoding Beats Rote Repetition
Ebbinghaus used nonsense syllables specifically to isolate pure memory. You're not doing that. You're learning things that mean something. Exploit that.
For a pharmacology card on metformin, don't just memorize "metformin → decreases hepatic glucose output." Add:
- Why: activates AMPK, inhibits complex I in mitochondria
- Clinical context: first-line for T2DM, watch for lactic acidosis in renal failure
- Mnemonic: "met-FORM-in" → forms less glucose
Every connection makes the memory more resistant to decay.
Expect Variability
If you're retaining 60% of cards after a month and your study partner is at 75%, that doesn't mean you're doing it wrong. Forgetting rates vary. What matters is that your retention is improving over time as the algorithm learns your personal curve.
SmartRecall's FSRS algorithm tracks four parameters per user: initial stability, difficulty, retrievability decay rate, and review success impact. After 50-100 reviews, it has a decent model of your forgetting curve, not Ebbinghaus's.
Don't Panic About Early Forgetting
Losing 40% of new information in the first day feels bad, but it's normal. The curve flattens. If you stick with spaced repetition, that card you barely remembered on day 2 will be rock-solid by day 60.
I've seen med students quit Anki after a week because they felt like they were forgetting too much. They were comparing themselves to an imaginary ideal, not to the actual science. Forgetting is part of learning. The goal isn't to prevent it—it's to time your reviews so you catch cards just before they slip away.
The Curve Ebbinghaus Didn't Study
One thing Ebbinghaus didn't measure: what happens when you keep using the information in real contexts. His experiment isolated memory from application. But if you're learning Spanish and you use "tener" in conversation three times this week, that's not just spaced repetition—it's active integration.
This is why language learners who combine SRS (for vocabulary) with immersion (for usage) see better long-term retention than either method alone. The forgetting curve for "words I've only seen in flashcards" is steeper than the curve for "words I've used in conversation."
SmartRecall can't simulate conversation, but it can remind you to use what you're learning. That's why we added the "apply this week" tag—a nudge to take three cards from your reviews and actually use them in a sentence, a practice problem, or a teaching moment.
Why This Still Matters in 2026
Ebbinghaus published his findings 141 years ago using a sample size of one and materials nobody would ever want to remember. Why does it still matter?
Because he proved that memory is measurable. Before 1885, memory was philosophy and introspection. After 1885, it was data. That opened the door for everything that followed: Leitner boxes in the 1970s, SuperMemo's SM-2 algorithm in 1988, Anki in 2006, and modern FSRS-based tools like SmartRecall today.
If you're choosing between algorithms, which spaced repetition algorithm actually wins compares SM-2, FSRS, and Leitner directly.
The specific numbers from his experiment don't generalize perfectly to your MCAT prep or your Mandarin HSK 5 study. But the shape of the curve—steep early drop, flattening over time, reset by retrieval—holds up. That's the foundation every spaced repetition algorithm builds on.
The Forgetting Curve You Should Actually Use
If you want a forgetting curve that applies to your learning, don't trust the infographic. Trust your data.
After 100 reviews in any decent SRS tool, you'll have enough history to see your personal retention curve. In SmartRecall, the stats page shows you:
- Retention rate by card age (1 day, 1 week, 1 month, 3 months)
- Average interval length (how long between reviews)
- Lapse rate (how often you forget cards)
That's your Ebbinghaus curve. It won't match his, and it won't match mine. But it's the one that matters.
Ebbinghaus gave us the method. Modern algorithms give us the personalization. Your job is to show up, do the reviews, and let the curve flatten over time.
The forgetting curve isn't something to fear. It's something to work with.
