The Forgetting Curve: Why New Words Vanish in 24 Hours — and How Reading Interrupts It

Twenty new words on Monday. By Thursday you can recall six of them, and two of those only with a hint. Your memory isn't failing. It's doing exactly what it was built to do — discard information that hasn't proved it matters.

That discard rate was measured almost 140 years ago, and it hasn't improved since. What has changed is that we now know precisely what convinces a brain a word is worth keeping. Most study methods don't do it. Reading, done right, does it automatically.

Try the method first: Generate your first bilingual story

What the Forgetting Curve Actually Shows

In 1885, a German psychologist named Hermann Ebbinghaus shut himself in a room and memorized lists of invented syllables — WID, ZOF, KEB — then tested how much survived after twenty minutes, an hour, a day, a week. The results were brutal. Roughly half was gone within the hour. After a day, about two-thirds. After a week, he was holding on to a quarter of what he'd learned.

Plot those numbers and you get the forgetting curve: a cliff, then a long shallow slide. The cliff is the first 24 hours. Whatever survives that window decays slowly for weeks. Whatever doesn't is gone by the weekend.

The experiment has held up. A careful 2015 replication, run with modern controls, reproduced the same basic curve. This isn't a quirk of one Victorian psychologist's brain. It's the default setting of yours.

But there's a detail in the original experiment that most summaries skip, and it changes what the curve means for you. Ebbinghaus memorized nonsense on purpose. Syllables with no meaning, no context, no connection to anything he knew. He wanted to measure pure memory, uncontaminated by association.

So the famous curve isn't a picture of how you forget language. It's a picture of how you forget language stripped of everything that makes it language.

Why Flashcard Words Sit on the Steep Part of the Curve

Here's the uncomfortable part: a flashcard is closer to WID and ZOF than you'd like. "Sin embargo = however" is an isolated pair — no sentence around it, no voice saying it, no reason it appeared. Your brain files it in the same drawer as Wi-Fi passwords and the confirmation code from Tuesday's parcel. Things that decay on schedule.

Now put the same word somewhere. Elena is awake at four in the morning in a Valencia apartment, listening to her neighbor's television through the wall. She should be furious. Sin embargo, she finds the muffled voices strangely comforting. The word now has a room, an hour, a small contradiction attached to it. Forgetting it means unpicking all of that — and brains are bad at unpicking.

Context doesn't exempt a word from the curve. Meaningful material decays too — just from higher up, and more slowly. What context buys you is time.

Time for what, though? A slower slide is still a slide.

How Repeat Encounters Bend the Curve

Only one thing flattens the forgetting curve: meeting the word again, ideally around the moment it starts to slip. Each successful retrieval resets the decay at a shallower angle. After a handful of well-spaced encounters, the slope goes nearly flat — the word stops being something you studied and becomes something you know.

This is the spacing effect, one of the most consistently replicated findings in memory research, and it's the engine inside every spaced repetition app. So let's be fair to flashcards for a moment: the scheduling behind Anki works. The research it's built on is solid.

The failure is human, not algorithmic. Reviewing decontextualized word pairs is joyless enough that most people quit within weeks — and a perfect review schedule you abandon flattens nothing. The apps that track this see the same shape over and over: a burst of enthusiasm, a growing pile of overdue cards, silence.

There's a subtler problem too. The card teaches you the card. Recognizing "sin embargo" on a white rectangle, with three seconds and no pressure, is not the same skill as catching it mid-sentence while the rest of the paragraph keeps moving.

What you'd actually want is a system with the timing of spaced repetition and none of its texture. No decks. No due counts. No white rectangles.

The Review Schedule Hidden Inside Every Story

Language has a statistical property that works quietly in your favor: a small number of words do most of the work. In ordinary text, the hundred most frequent words cover roughly half of everything on the page. The next few thousand cover most of the rest.

Think about what that means if you read a short story today and another one on Thursday. "Pero" shows up eleven times in the first story and nine in the second. "Después" appears in both. "Aunque" waits four days, then reappears exactly when it was about to fall off the cliff. Nobody scheduled these reviews. The structure of the language scheduled them.

And every encounter arrives in a new disguise. On Monday, "después" came after a breakup. On Thursday it comes after a train departure. Retrieving a word from a fresh context is harder than recognizing it on a familiar card — and that difficulty is precisely what makes the memory stronger and more flexible. You're not learning that one card. You're learning the word.

Research on vocabulary picked up through reading suggests a word needs somewhere between eight and twelve encounters before it reliably sticks. That sounds discouraging until you notice that any reasonably common word hits that number within a couple of weeks of daily reading — without you tracking anything.

The rare words won't hit it on their own. They need help.

What to Do With the Words That Keep Slipping

Three moves, all of them cheap.

Reread before you move on. A second pass through yesterday's story catches its vocabulary two or three days after the first encounter — close to the ideal interval, right as the steep part of the curve is doing its damage. This is why rereading a story you already understand isn't wasted time. It's the review session, disguised as the easy option.

Read tomorrow, not on Sunday. The curve does its worst work in the first 24 hours. A Sunday study session leaves six days of unopposed decay; fifteen minutes tomorrow re-encounters this week's words while they're still catchable. Frequency of contact beats size of session, and it isn't close.

Let some words go. If "nutria" — otter — slips away after one story about a river, that's the system working. The curve is a filter, not a flaw. Words that matter recur; words that recur survive. Spend your attention on the ones the language keeps handing back to you.

None of this requires willpower or software. It requires stories at a level where you meet new words a few at a time — not twenty per paragraph, all standing on the cliff edge together.

The forgetting curve never goes away. You'll still lose most of what you meet once — everyone does, in every language, including their own. But you don't beat the curve with effort. You beat it with encounters, and a story at your level is a few hundred well-timed encounters wearing a plot. Generate your first bilingual story at your level, read it today, and reread it Thursday. The schedule takes care of itself.

Not sure of your reading level? Take the free CEFR test →