Why You Forget Vocabulary Right After Learning It (and How Spaced Repetition Fixes It)

You forget most new vocabulary within a day because that's how memory is built to work by default, not because you're bad at learning languages. A German psychologist mapped this exact pattern more than a century ago, and it's called the forgetting curve.

What is the forgetting curve, exactly?

In the 1880s, Hermann Ebbinghaus ran a now-famous series of experiments on himself, memorizing lists of meaningless syllables and testing his own recall at increasing intervals afterward. What he found became one of the most replicated results in memory research: forgetting follows a curve rather than a straight line: steep at first, then flattening out over time.

Ebbinghaus documented that without any reinforcement, a large share of newly learned information disappears within the first day, and the rate of loss slows sharply after that. The first few hours after learning something new are where you lose the most ground, not week three or month two. This is why a word you looked up yesterday can feel completely gone today, while a word you reviewed several times over the past month still feels solid.

The shape of the curve also explains something counterintuitive: the fix isn't to study harder in that first session. Cramming a word ten times in one sitting barely changes where it sits on the curve a day later, because the curve is about time elapsed, not repetitions crammed into a single block. What changes the curve is coming back to the word again, later, after some forgetting has already started.

Why does a single exposure to a new word almost never stick?

When you meet a word for the first time, your brain doesn't yet know if it's worth keeping. Memory formation is expensive, and the brain is selective about what it consolidates into longer-term storage. A word seen once, with no repetition and no particular emotional or contextual weight, looks to your brain like noise rather than a signal worth preserving.

This is separate from the vocabulary-in-context research on why words met in a real sentence stick better than words on an isolated list, covered in more depth in vocabulary in context vs word lists. Context helps build a stronger initial trace, but even a well-anchored first exposure still decays fast without a second and third encounter. Context improves what you retain from one exposure; it doesn't cancel the forgetting curve itself.

The practical consequence: feeling confident about a word right after you look it up is not a reliable signal that you'll remember it tomorrow. That confidence reflects the word sitting in short-term working memory, fast but shallow, rather than in the more durable long-term storage that survives days or weeks without contact.

Why does spaced repetition actually work against this curve?

The core insight behind spaced repetition, first formalized by researchers building on Ebbinghaus's original work, is that reviewing a word right before you'd naturally forget it resets and strengthens the curve more efficiently than reviewing it either too early or too late. Review too early, while the word is still fresh, and the review does almost nothing, since you weren't at risk of forgetting yet. Review too late, after you've already forgotten it, and you're relearning from scratch instead of reinforcing an existing trace.

Each successful review doesn't just delay forgetting, it changes the shape of the curve itself. A word reviewed at the right moment decays more slowly the next time, which is why well-spaced review intervals stretch out over weeks and then months, while a word you keep failing stays on a short, frequent cycle. This is the entire mechanism spaced repetition software is built to exploit: finding that right moment for each individual word, for each individual learner, rather than applying one fixed schedule to everyone.

Fixed-interval systems, review this word again in three days, then seven, then fourteen, for everyone regardless of how well they actually knew it, are a rough approximation of this idea from the 1980s. Modern algorithms go further by modeling the curve individually per word and per person, discussed in detail in how FSRS spaced repetition works in Lira.

How does FSRS improve on that basic idea?

FSRS (Free Spaced Repetition Scheduler) starts from the same Ebbinghaus-derived premise, that forgetting follows a predictable curve, but treats that curve as something to measure per word and per person rather than assume as a fixed average. It tracks a value researchers call memory stability, essentially how resistant a specific word is to forgetting for you specifically, based on your actual review history with that word.

A word you've rated confidently correct five times in a row gets a longer interval before its next review, because its estimated stability is high, the curve for that word has flattened. A word you keep getting wrong stays on a short cycle, because its stability is still low and the curve for that word is still steep. This is a direct, practical application of Ebbinghaus's century-old observation, just computed individually instead of applied as one generic rule to every learner and every word.

The result is fewer wasted reviews on words you've already secured, and better-timed reminders on the words actually at risk of slipping, right before they would have been forgotten rather than well after.

What does this mean for how you should review vocabulary?

Reviewing a word once, right after you learn it, and never again is close to the least efficient way to spend that study time, since it sits precisely where the forgetting curve is steepest and gives you no second encounter to flatten it. A single review a few days later, timed to catch the word just as it starts to fade, does more for long-term retention than five reviews crammed into the same afternoon it was first learned.

This also means the words most worth reviewing aren't necessarily the ones that feel hardest right now, they're the ones approaching their natural forgetting point, whether that's tomorrow or in six weeks. That's a moving target for every word individually, which is exactly the kind of tracking a spaced repetition system handles better than manual review scheduling.

In Lira, every word you tap while reading gets queued into FSRS-based review automatically, with the original sentence attached so the review reconnects you to the context where you first met the word, not just a bare translation. You don't need to build a deck or guess at review timing yourself, the schedule adjusts to your actual recall on each word as you rate it.

FAQ

How much vocabulary do you actually lose in the first 24 hours?

Ebbinghaus's original experiments, and the research that has replicated the pattern since, point to a steep initial drop within the first day without any reinforcement, with the rate of loss slowing considerably afterward. The exact percentage varies by study and by how the material was first encoded, but the shape of the curve, fast then slow, holds up consistently.

Does reviewing a word more times in one session help it stick better?

Not much, once you're past a couple of repetitions. Massed repetition (reviewing the same word many times back to back) produces a short-term illusion of mastery that fades almost as fast as a single exposure would, because the curve is driven by time elapsed and spaced re-encounters, not by total repetitions within one sitting.

Is the forgetting curve the same for every type of word?

No. Concrete, frequent, emotionally anchored words tend to resist forgetting better than abstract or rare ones, which is part of why vocabulary met in an engaging story tends to stick better than the same word drilled on an isolated flashcard. The underlying curve shape still applies to both, but its steepness varies by word.

More on the method behind Lira's approach to vocabulary is covered in vocabulary in context vs word lists and how FSRS spaced repetition works in Lira. Questions about the app are answered in our FAQ, or contact us directly.