NotesNudge the notes you never go back to

What Ebbinghaus Actually Showed

The forgetting curve is the most reproduced graph in writing about learning: a line dropping steeply in the first hours, then flattening. It appears on product pages, in course marketing, and in almost every article about spaced repetition.

The curve is real and it has been replicated. What it measured is not what the usual axis label says, and the material it measured is as far from your notes as it is possible to get.

For a workplace-oriented comparison with this kind of workflow, the reference provides additional context.

Reviewed August 9, 2026.

What he actually did

Hermann Ebbinghaus ran the experiments on himself between 1880 and 1885, publishing Memory: A Contribution to Experimental Psychology. Sample size: one.

He learned lists of nonsense syllables — three-letter consonant-vowel-consonant units, chosen deliberately because they carried no meaning. He had tried tones, digits and stanzas of Byron and rejected them as unsuitable. Meaning was the thing he wanted removed, because meaning is what makes forgetting irregular.

For an independent external reference related to this topic, see Encyclopaedia Britannica.

Then he relearned the lists after intervals from twenty minutes to a month.

The part almost everyone gets wrong

He did not measure how much he could recall. He measured savings — the reduction in the time needed to relearn the list to the original standard.

If a list took twenty minutes to learn and twelve to relearn, savings is 40%. If relearning took just as long, savings is zero. If no relearning was needed, savings is 100%.

That is a different quantity from "percentage remembered", and it is more sensitive: savings can be positive when conscious recall is nil. His methodological insight was precisely that memory can be measured when nothing can be consciously recalled.

So a graph labelled "% of information retained" is mislabelled. His own figures were roughly 58% savings twenty minutes after learning, about 44% after an hour, about 34% after a day, and about 21% after a month.

It replicated

Jaap Murre and Joeri Dros published a replication in PLOS ONE in 2015. One subject, seventy hours of learning and relearning, intervals of twenty minutes, one hour, nine hours, one day, two days and thirty-one days — the first non-German replication.

The results closely matched the original, which is a genuinely impressive outcome for a self-experiment from 1880.

They also found something the textbook version omits: the curve is not smooth. There is a jump upwards starting around the 24-hour point, generally attributed to sleep consolidation. Ebbinghaus saw the same anomaly and fitted it away.

Note the sample size of the replication. One subject, twice, a century and a third apart. Robust in shape and narrow in population.

What it does not tell you about your notes

This is why the page exists.

Nonsense syllables were chosen to have no meaning. Your notes are the most meaningful material you own — written by you, about things you cared about, connected to everything else you know. Meaningful material forgets far more slowly, and Ebbinghaus stripped meaning out precisely because it makes forgetting irregular and hard to model.

Applying his decay rates to your notes extrapolates from the least meaningful material possible to the most.

Savings is not availability. A curve showing 34% savings after a day does not mean you retain 34% of an idea. It means relearning would go a third faster.

And it says nothing about what to do. The curve describes decay. It does not prescribe an interval, a schedule, or a tool. The intervention research is separate, and citing Ebbinghaus to justify a review schedule is borrowing a description to license a prescription.

How to spot the misuse

An axis labelled "memory retained." It is savings.

A specific percentage attached to your material. "You forget 70% within 24 hours" is a nonsense-syllable figure applied to meaningful content.

The curve used to justify a product's schedule. A description of decay does not imply any particular intervention. When a note-resurfacing tool cites this, the citation is decorative.

And a smooth curve. The best replication says it is not — and the same care applies to every confident figure in this area.

What is genuinely useful in it

Forgetting is rapid at first and then slows. The shape is real, replicated, and worth knowing.

Something survives beyond conscious recall. The savings method demonstrates that material you cannot produce is not gone, which is quietly encouraging for anyone with an archive they cannot remember writing.

And sleep appears to matter, which is the most interesting thing in the replication and the least discussed.

That is a real contribution from a self-experiment in 1880. It is just not a schedule, and not about your notes.

The short version