Forgetting Curve: Why Ebbinghaus Tested Himself
The Forgetting Curve is Hermann Ebbinghaus’s record of how fast he forgot lists of meaningless syllables — using himself as the only subject for years. The famous percentages describe that experiment, not how fast you forget a lecture, a statute, or a language. He published in 1885; classrooms still treat those numbers as a stopwatch.
The Ebbinghaus Memory Experiment Had a Sample Size of One
Forgetting curve research did not start with students, classrooms, or meaningful material. It started with one German psychologist, sitting alone with lists he invented so they would mean nothing.
Between roughly 1879 and 1884, Hermann Ebbinghaus memorized thousands of CVC nonsense syllables — strings such as “WID,” “ZOF,” “KAF” — then retested himself after delays from 20 minutes to 31 days. He chose nonsense on purpose: prior knowledge would have contaminated the clock. Meaning was the variable he wanted out.
That is the Ebbinghaus forgetting curve in its original form: n = 1, material with no semantics, measured as “savings” — how much less time he needed to relearn a list, not how many items he could freely recall. Most popular charts skip that last point. Savings and recall are not the same test.
The precision came from the isolation. So did the limit. One highly practiced adult, using a method no student uses, is not a population.
What the Forgetting Curve Actually Plotted
The Forgetting Curve plotted the speed of losing unused, meaningless traces — and the drop was front-loaded. After about 20 minutes, Ebbinghaus retained roughly 58 percent of a list by his savings measure. After one hour, about 44 percent. After one day, about 33 percent. After six days, about 25 percent. After 31 days, about 21 percent.
Those figures appear in his 1885 monograph Über das Gedächtnis, later published in English as Memory: A Contribution to Experimental Psychology. They are often rounded, sometimes misquoted as “you forget 70 percent in 24 hours,” which is a loose reading of the one-day point, not a classroom law.
Two caveats sit inside the data. First, he was measuring relearning speed, not a quiz score. Second, he was already an expert at the task. Practice effects on the experimenter himself are a known problem in this literature, and later replications with more subjects have recovered the same general shape without matching every percentage.
The curve is a description of unused nonsense under one protocol. It is not a diagnosis of your Tuesday lecture.
Why the Forgetting Curve Does Not Predict Exam Scores
Forgetting curve explained as a 24-hour collapse is a slogan, not Ebbinghaus. If the material has meaning, structure, or use, decay is slower than his tables. If you never retrieve it, even meaningful material still drops fast. The useful claim is the shape — not the 33 percent.
Spaced retrieval works because each successful recall flattens the next drop. Ebbinghaus already saw this: relearning was cheaper than first learning, and distributed practice beat massed practice in his own logs. That finding has held up far better than the exact percentages.
How fast do we forget a name, a formula, or a procedure? Faster without retrieval, slower with it, and never on a single published timetable. A medical student who explains a pathway aloud at 24 hours is not in the same experiment as a man reciting WID-ZOF-KAF in 1880.
Treat the numbers as a warning about delay, not a grade prediction.
Use the Forgetting Curve as a Clock, Not a Verdict
The practical move is to schedule the first retrieval before the steepest part of the drop — the same day, then the next day — and to test yourself, not reread. Rereading feels like knowledge. It is closer to Ebbinghaus’s first encoding than to his savings test.
Do not wait a week to find out you have lost the lecture. That is the one result from a sample of one that still earns its keep.
The Forgetting Curve became a poster because it is simple. Ebbinghaus made it simple by stripping meaning out. Put meaning back in, and you keep more — but only if you ask for it again before the unused trace is gone.