The Forgetting Curve, Replicated

Every field has one founding measurement it keeps coming back to — the number that made it a science. For learning science, that number is a decay curve. One man traced it by hand, on his own memory, before psychology had labs. Everything else in this library’s memory cluster is, in one way or another, an answer to what he found.

Most workplace training runs on a silent bet: what is taught in January is still there in June. The first person to test that bet with numbers worked alone at a desk in Berlin. He read lists of invented syllables aloud to the beat of a metronome. His name was Hermann Ebbinghaus. In 1885 he found that the bet fails — and fails in a specific, measurable way (Ebbinghaus, 1885). A pair of Amsterdam researchers confirmed it, curve for curve, 130 years later.

This article walks through the original experiment. Then the 2015 replication, which made it one of the best-attested results in psychology. Then what the curve does and does not let you conclude. And last, why a finding older than the car still describes the default outcome of most corporate training.

The experiment Ebbinghaus ran on himself

The design is bare, and the bareness is its power. It repays a careful look — both for what was measured and for the discipline of the measuring. Ebbinghaus wanted to study memory with meaning stripped out, so he built material that had none: thousands of consonant–vowel–consonant “nonsense syllables” — ZOF, DAX, WID — strung into lists. He learned each list until he could recite it perfectly. He waited a set interval. Then he measured what forgetting had cost him (Ebbinghaus, 1885).

His key tool was the savings score. He did not ask “how much can I still recall?” He asked “how much faster can I relearn the list than I learned it the first time?” If a list first took ten minutes and relearning took six, savings were 40%. Savings pick up more than free recall can. They detect memory traces that have sunk below the level of conscious recall.

He tested himself at seven intervals, from about twenty minutes out to thirty-one days. The numbers fell fast, then slowly. Savings were roughly 58% after twenty minutes and about 44% after an hour. After a day, roughly a third. After a month, about 21% (Ebbinghaus, 1885).

Two traits of that curve matter more than any single value on it. First, the loss is front-loaded: the steepest forgetting happens within hours, not weeks. Second, the curve flattens — what survives the first days tends to last. A century later, researchers fit 105 candidate equations to 210 published retention datasets. The same basic form held across materials and methods: steep then flat, well described by log and power functions (Rubin & Wenzel, 1996).

A 130-year-old result, re-run

For most of the twentieth century, the forgetting curve held a strange status. Everyone taught it. Almost no one re-ran it in its original form. Ebbinghaus was his own only subject, his statistics were pre-modern, and his book predates the randomized experiment. When psychology’s replication debates heated up in the 2010s, the curve was an obvious pick for a stress test.

Jaap Murre and Joeri Dros at the University of Amsterdam re-ran the 1885 study as faithfully as modern conditions allowed (Murre & Dros, 2015). One participant. Freshly built nonsense-syllable lists. The same savings measure. And the same seven retention intervals, from twenty minutes to thirty-one days.

The replication succeeded. The 2015 curve tracks the 1885 curve closely — steep early loss, then a flattening tail. The authors also fit equations to the data: a simple exponential fits poorly, while power-family functions do well, in line with the wider retention literature (Murre & Dros, 2015). The replication even reproduced a subtle anomaly. Retention at the 24-hour point sits slightly above what a smooth curve predicts — a “jump” visible in both datasets. Murre and Dros suggest it may reflect memory consolidating during sleep (Murre & Dros, 2015).

The replication’s value goes beyond sentiment. Psychology’s replication era taught the field a hard lesson: famous findings often shrink or vanish under modern re-testing. Several of this library’s uncomfortable-evidence articles document exactly that fate.

Against that backdrop, this result stands out. A Victorian-era, single-subject finding came back curve for curve, anomaly included. No pile of citations could say as much. The shape of forgetting is not an artifact of one era’s methods. It is a stable fact about human memory.

Successful replications of 130-year-old experiments are vanishingly rare, and the authors were struck by how well the original held up (Murre & Dros, 2015). Ebbinghaus’s fixed daily schedule, controlled pacing, and obsessive procedure anticipated standards that would not be formalized for decades.

Beyond nonsense syllables

Ebbinghaus chose meaningless material precisely so nothing could rescue his memory — no associations, no structure, no story. That purity invites a hopeful misreading. Perhaps real knowledge, wrapped in meaning, escapes the curve entirely — and the whole result is an artifact of studying gibberish. In short: nonsense syllables are meaningless by design, and meaningful material behaves better. The objection is half right, and its half-wrongness has been measured across half a century. Meaning, structure, and prior knowledge slow the decay — but they do not repeal it.

The most striking evidence comes from Harry Bahrick. He studied people who had learned Spanish in school, testing them up to fifty years later. Knowledge dropped steeply across the first few years after the last course. Then it settled into a plateau Bahrick called “permastore,” holding roughly level for decades (Bahrick, 1984). Same shape as Ebbinghaus’s curve — steep, then flat — stretched across a lifetime.

Professional schooling repeats the pattern. Reviews of medical training find large losses of basic-science knowledge within the first year or two after study, even among strong learners (Custers, 2010). Meaning changes the slope of the curve, not its existence.

What the curve licenses — and what it doesn’t

The curve’s practical value lies in its lawfulness, so it pays to be precise about what the law covers. It describes unreviewed memory: the path of material learned once and then left alone. It does not say memory must follow that path. The whole spacing and retrieval literature shows that well-timed practice resets the curve at ever shallower slopes (Cepeda et al., 2006), (Roediger & Karpicke, 2006). Read correctly, the forgetting curve is not a prophecy but a baseline. It is the cost of doing nothing, in numbers — the yardstick for judging any program’s value.

Lawful also means predictable, at the level that matters for scheduling. The curve’s exact rates shift with the learner, the material, and how strong the first learning was. That fact is sometimes raised as if it sinks the whole project. It does the opposite. The form of the decay is stable while its rate varies. So each learner’s retrieval history can estimate their personal rates, and each estimate sharpens the forecast of when a given memory will near failure.

That is the entire engineering premise of modern spaced-repetition systems — not one curve for everyone, but one function family, fitted per person and per item, updated with every attempt (Rubin & Wenzel, 1996), (Cepeda et al., 2008).

One boundary deserves respect. Savings scores measure relearning speed, not job performance. A savings figure of 21% at one month does not mean a trained employee performs at 21%. Skills used in daily work review themselves, so they last. The curve bites hardest on exactly the material corporate training most often certifies: rarely used knowledge learned in one sitting. That is why the retention findings from professional schooling track the lab ones so closely (Custers, 2010).

The most expensive ignored fact in training

The costing exercise is worth doing out loud, because the curve prices it. Take a training budget and note the share spent on one-exposure formats. Then apply the decay the replicated data describe. Most of what that spend deposited is out of reach within weeks — before most of it is ever needed. No buyer would accept that depreciation schedule on equipment. Learning escapes the scrutiny only because the loss is invisible: nobody re-tests, so nobody books the write-down.

Now hold the curve against the standard shape of corporate learning: a workshop, a webinar, an e-learning module. One exposure, a completion checkbox, and no scheduled contact with the material ever again. The forgetting curve is a precise description of what happens to that investment next. It is not that one-shot training teaches nothing. It is that the design concedes the steepest part of the curve without a fight.

The irony is that the countermeasures are among the best-documented tools in cognitive science. Dunlosky and colleagues reviewed ten common learning techniques to see which ones work. The techniques that dominate practice — rereading, highlighting, summarizing — earned low-utility ratings. The two rated high utility were exactly the two that attack the curve: distributed practice and practice testing (Dunlosky et al., 2013).

The evidence behind those two ratings is unusually deep. A meta-analysis — a study that pools the results of many studies — covered 839 assessments of distributed practice across 317 experiments. It found that spacing study over time reliably beats massing it in a single session (Cepeda et al., 2006). A follow-up study mapped the schedule itself. The best gap between reviews scales with the retention horizon — the longer you need to remember something, the longer the ideal wait before revisiting it (Cepeda et al., 2008).

And retrieval does more than measure memory; it strengthens it. Students who practiced recalling a passage retained far more a week later than students who spent the same time restudying it (Roediger & Karpicke, 2006). The mechanism is what Bjork and Bjork call a desirable difficulty: conditions that make retrieval feel harder in the moment produce more durable learning than conditions that feel easy and fluent (Bjork & Bjork, 2011).

The two high-utility techniques also compound — with each other, and with the curve itself. Spaced sessions built from retrieval attack the decay twice per visit: the gap supplies the desirable difficulty, and the retrieval resets the slope. That is why the strongest applied designs in this library, from successive relearning to interleaved question streams, are at bottom delivery systems for this one page of 1885 arithmetic.

None of this would matter in practice if the right review moment were easy to schedule by hand. It is not. The fade point differs by learner, by concept, and by how the last retrieval went. That is precisely why a 140-year-old descriptive law went unused for so long. Applying it takes bookkeeping at a per-person, per-concept grain that no calendar, and no instructor, can sustain.

What the evidence doesn’t show

The forgetting curve is robust, but it is routinely over-claimed. Four limits are worth stating plainly:

• The viral numbers are not in the data. The claim that “people forget 70% of training within 24 hours” does not come from Ebbinghaus. His measure was savings on relearning, not percent forgotten, and his material was deliberately meaningless. His numbers do not transfer to meaningful, structured content — which decays more slowly, as Bahrick’s long plateaus show (Bahrick, 1984).
• It is an n-of-1 result, twice over. Both the 1885 original and the 2015 replication used a single subject (Murre & Dros, 2015). The generalization rests on the surrounding retention literature — hundreds of datasets showing the same functional form (Rubin & Wenzel, 1996) — not on these two studies alone.
• The curve is descriptive, not prescriptive. It tells you forgetting is lawful; it does not by itself tell you when to review. Optimal intervals come from the spacing literature and shift with the retention goal (Cepeda et al., 2008); there is no single schedule that is right for all material or all horizons.
• The rate is not a constant. Degree of initial learning, material type, sleep, and individual differences all change the slope. A single “company forgetting curve” applied to every employee and every topic is a convenient fiction; the law is in the shape, not in one universal decay rate.

Where the evidence stops

1. 1The viral numbers are not in the data
2. 2It is an n-of-1 result, twice over
3. 3The curve is descriptive, not prescriptive
4. 4The rate is not a constant

© 2026 FUTURE PROOF™

From description to prescription

Ebbinghaus gave the field a law of loss. The replication confirmed the law was real, not an artifact of one man’s heroic self-testing (Murre & Dros, 2015). The spacing and retrieval literatures then turned that description into a prescription: interrupt the curve with effortful retrieval, timed to the retention horizon (Dunlosky et al., 2013).

What remained missing for most of that history was the machinery to act on it — to track where each learner sits on each concept’s curve and intervene at the right moment. The science has been settled for longer than anyone reading this has been alive. The engineering, finally, is not the hard part anymore.