The forgetting curve: why you forget new words (and how to stop)

What is the forgetting curve?

The forgetting curve describes how quickly new information fades from memory once you stop reinforcing it. You learn a new word, close the book, and — without any further review — most of what you just learned decays within the first day or two. After that initial drop, the rate of loss slows down: whatever's left fades more gradually. It's not a flaw in your memory. It's the default behavior of memory itself, and it's the reason a study technique like spaced repetition exists at all.

What did Ebbinghaus actually discover, and how?

The forgetting curve traces back to Hermann Ebbinghaus, a German psychologist who spent the 1880s running one of the first systematic experiments on human memory — on himself. He memorized lists of nonsense syllables (meaningless three-letter combinations like "WUX" or "ZOF," chosen specifically because they had no existing associations to lean on) and then tested his own recall at various delays afterward, tracking how much he could still reproduce.

Worth being honest about the method: this was one person, testing himself, on invented syllables rather than real vocabulary, with the record-keeping tools of the 1880s. Modern memory research has run far larger and more varied studies since. But the basic shape Ebbinghaus found — a steep early drop followed by a long, shallow tail — has held up remarkably well across a century-plus of follow-up work, and it's still the standard mental model researchers and teachers reach for. Treat it as a useful, well-supported picture of how forgetting behaves in general, not a precise formula that predicts exactly how much of today's vocabulary list you'll remember on Thursday. Nobody has ever produced a universal percentage that applies to every learner, every word, and every context — treat any claim that specific should raise an eyebrow.

Why is forgetting so steep right after you learn something?

The steep early drop isn't your memory malfunctioning — it's your brain doing exactly what it's built to do. You're exposed to enormous amounts of information every day, and most of it genuinely doesn't matter later. Without some signal that a particular piece of information is worth holding onto, the default is to let it fade, because storing everything indefinitely isn't how brains work efficiently.

A single exposure to a new word — reading it once, hearing it once — doesn't send that signal. Nothing about seeing "dwell" one time tells your brain "you'll need this again." The information was never wrong to forget; it just never got flagged as important. What does flag it as important is what happens next: whether you encounter it again, and specifically whether you have to work to retrieve it again.

What actually flattens the curve?

Each successful retrieval, timed close to the point where you'd otherwise have forgotten, resets the decay and — this is the part that matters — makes the next decay slower. Review a word right after learning it and again a day later, and the curve for that word gets shallower. Review it again a week later, and it flattens further still. Enough of these well-timed reviews, and a word stops decaying on any timescale you'd notice day to day.

That's the entire mechanism behind spaced repetition: instead of reviewing everything on a fixed schedule, or leaving review to whenever you happen to think of it, a system tracks each item individually and brings it back right before the forgetting curve would otherwise win. Spaced repetition explained goes into how that scheduling actually works — Budling's Learn mode runs on FSRS, an algorithm that adjusts each card's next review date based on how you actually rated it, rather than a one-size-fits-all interval. This post is about why that's necessary in the first place; that one is about the fix.

Why does cramming feel like it works, even though it doesn't?

Cramming produces a real, measurable effect: you'll probably do fine on a quiz tomorrow. The problem is what that quiz is actually measuring. Rereading a word list repeatedly the night before a test builds recognition — you'll see the word and it'll feel familiar, correct, obviously known. Recognition is a low bar. It's much easier than recall, which is producing the word or its meaning from scratch, unprompted, with nothing in front of you to jog your memory.

A multiple-choice quiz mostly tests recognition. An actual conversation, or writing a sentence, or filling in a blank with no options listed, tests recall — and recall is exactly what collapses fastest along the forgetting curve. That gap between "I recognized it on the quiz" and "I could produce it three days later" is why cramming feels productive in the moment and evaporates almost as fast as it was built. Active recall is the technique built specifically to close that gap — testing yourself by trying to produce the answer, not just reviewing it, so what you're practicing during study time matches what you'll actually need later.

What does review timing actually look like for vocabulary?

For a single new word, a workable pattern looks roughly like: review it again within a day of first learning it, then a few days later, then roughly a week out, then longer gaps after that — each successful recall pushing the next review further away, each miss pulling it back in. The specific intervals matter less than the principle: review just before you'd forget, not on a fixed calendar and not only when you happen to remember to.

Tracking that by hand for more than a handful of words quickly becomes unmanageable, which is the actual reason spaced repetition software exists — it's not a shortcut around the forgetting curve, it's bookkeeping for it. Budling runs this automatically: every flashcard set uses FSRS to schedule reviews per card based on how you rate your own recall, so words you know cold get pushed weeks out while shaky ones stay in a tighter loop until they stick.

How fast do you forget new vocabulary?

Faster than most people expect, and unevenly — the sharpest drop happens in the first day or so after learning something new, then the rate of loss slows considerably. Exactly how much sticks depends on the word (a concrete, frequently-used word tends to stick better than an abstract, rare one), how it was learned (in a sentence versus as an isolated definition), and the learner. There's no single number that applies across the board, which is precisely why a fixed review schedule undersells easy words and oversells hard ones — an adaptive system that tracks each word individually handles that variation better than a flat calendar can.

Does the forgetting curve apply to all types of memory?

Broadly, yes — the same shape (fast initial decay, then a slower tail) shows up across many kinds of learned information, not just Ebbinghaus's nonsense syllables. But not all memory decays at the same rate or responds to review the same way. Meaningful, connected information (a word learned in context, tied to something you already know) tends to resist forgetting better than isolated facts, which is part of why learning vocabulary inside example sentences works better than memorizing bare word lists. The curve is a general pattern, not a promise that every fact in your head is decaying identically.

Stop losing words to the curve

The forgetting curve isn't a reason to give up on vocabulary — it's just what happens by default when review is left to chance. The fix isn't more studying; it's better-timed studying. Take Budling's placement quiz to find your starting level, then sign up free and let spaced repetition handle the timing your memory needs — so the words you learn today are still there next month.