You spent a year drilling Japanese vocabulary. Every morning, twenty minutes, the same routine: flip the card, retrieve the word, watch the interval grow. Then life happened — a busy month, a busier three, and suddenly you could not remember whether 忘れる meant "to forget" or "to forgive." The words you had reviewed a hundred times were gone, or so it felt. This experience is so common that learners have a name for the fear behind it: I'm losing the language.

The scientific name for that process is language attrition — the gradual loss of proficiency in a language you once knew. It is one of the most studied and least discussed phenomena in language learning, and the research contains both bad news and genuinely good news. The bad news: forgetting is real, measurable, and faster than most people expect. The good news: a forgotten language is not erased — it goes dormant, and the reactivation process is dramatically faster than the original learning. This article covers what attrition research has discovered over the past four decades, from Ebbinghaus's forgetting curve to a 50-year study of Spanish learners, and ends with a concrete flashcard protocol for keeping your vocabulary alive.

What Language Attrition Actually Is

Language attrition is defined simply as "the process of decreasing proficiency in or losing a language." It comes in two flavors. First-language (L1) attrition happens when your native language weakens because you live in a new country and a new language takes over daily life — the classic case of immigrants who begin to grope for words in their mother tongue. Second-language (L2) attrition is what most of us fear: the steady decay of a foreign language after learning stops, the moment you close the flashcard app for the last time.

Researchers have reached a strong consensus about what decays first. According to the current state of the field, attrition "manifests itself first and most noticeably in speakers' vocabulary (in their lexical access and their mental lexicon), while grammatical and especially phonological representations appear more stable." In plain language: your grammar and pronunciation survive longer; your words go first. The vocabulary you learned last, used least, and reviewed most rarely is the first to slip — which is exactly what makes a flashcard app the right tool for fighting attrition.

The Forgetting Curve, Replicated 130 Years Later

Every discussion of forgetting starts with Hermann Ebbinghaus, the German psychologist who, in 1885, memorized thousands of nonsense syllables and then tested himself at intervals to map exactly how memory decays. His famous forgetting curve showed that memory falls off steeply within the first hours and days, then levels out into a long, slow tail. Most of the loss happens shockingly fast — in a matter of days, not months.

For over a century, the curve was treated as folklore in some circles. Then in 2015, Jaap Murre and Joeri Dros of the University of Amsterdam published a rigorous replication in the journal PLoS ONE. One subject spent 70 hours learning lists of syllables and relearning them after delays of 20 minutes to 31 days, using Ebbinghaus's own method of savings (measuring how much faster relearning was than original learning). The results matched the 1885 data almost perfectly:

"We conclude that the Ebbinghaus forgetting curve has indeed been replicated and that it is not completely smooth but most probably shows a jump upwards starting at the 24 hour data point." — Murre & Dros (2015), PLoS ONE

Two lessons follow. First, forgetting begins the moment learning ends — the curve does not wait politely for a convenient time. Second, and more importantly, Ebbinghaus's method of savings is itself the good news: the amount of time needed to relearn a list was far smaller than the time needed to learn it the first time. The information was never truly gone; it just needed a nudge back above the surface.

The 50-Year Study: Bahrick's Spanish Learners

The single most important study in attrition research is Harry Bahrick's "Fifty Years of Second Language Attrition," published in the Modern Language Journal in 1984. Bahrick tracked 773 people who had studied Spanish in high school or college — a perfectly ordinary group of classroom learners, not hyperpolyglots — and tested their retention at intervals ranging from 1 to 50 years after their last Spanish course. It remains one of the largest longitudinal studies of language memory ever conducted.

The results surprised the field. Instead of a smooth, endless decline, Bahrick found a two-phase pattern. In the first three to six years after instruction ended, learners lost vocabulary at a noticeable rate. But then the curve flattened dramatically: people tested 25, 35, and even 50 years later performed at roughly the same level as people tested at 8 or 14 years. Bahrick called the stable residue the "permastore" — knowledge that had become so consolidated it seemed permanently retained.

Two further findings mattered enormously for how we study. First, recognition was far more durable than recall — you will recognize a word long after you can produce it, which is why "I know it when I see it" is a real, measurable stage of memory. Second, the level of original training predicted the level of permanent retention: learners who had overlearned material (continued practicing it beyond the point of first mastery) lost far less over the decades. Bahrick's conclusion, in his own words, was that the results offer "a basis for understanding how various procedures of acquisition affect the rate of loss and ultimately minimize" it.

Yesterday's Words: Reactivating a Long-Forgotten Language

If Bahrick proved that something survives, Kees de Bot and Saskia Stoessel (2000) proved that the survivors can be brought back. In their study, published as "In Search of Yesterday's Words: Reactivating a Long-Forgotten Language" in Applied Linguistics, they examined adults who had learned a language in childhood — and then not used it for decades. When these adults were asked to recognize and relearn the dormant language, the results were striking: even after 30 years or more of disuse, learners retained partial knowledge, and reactivation was dramatically faster than fresh learning.

The researchers interpreted this as evidence that "yesterday's words" are not destroyed by time; they are stored, weakened, and waiting. The practical implication for anyone who abandoned a language years ago is almost liberating: you are not starting over. You are waking something up. The relearning advantage — the "savings" that Ebbinghaus measured in minutes and de Bot and Stoessel measured in decades — means the second journey to fluency is cheaper than the first.

Use Matters: The 16-Month Study

How much does continued use really protect a language? A 2020 study by Ulrike Jessner, Kathrin Oberhofer, and Manon Megens, published in Applied Psycholinguistics, tracked 114 multilingual German-speaking young adults who had learned English as their first foreign language and French or Italian as their second. The researchers measured their spoken output shortly before graduation from upper secondary school and again approximately 16 months later.

The finding was exactly what the "use it or lose it" intuition predicts — with a twist. The amount of foreign language use after graduation had a noticeable impact on the measured quality of the second foreign language (the weaker one), but almost no visible connection for the stronger first foreign language, where initial proficiency was much higher. In other words: the more fragile the skill, the more it depends on continued use. The study also found that learners' self-assessments were unreliable in one direction — people who felt their speaking had deteriorated were often wrong; their actual output had not declined as much as they feared.

Even the First Language Can Fade

Attrition does not spare native speakers. Monika Schmid's landmark research on German Jews who emigrated to English-speaking countries in the 1930s — published as First Language Attrition, Use and Maintenance (2002) — documented how the mother tongue erodes when a second language takes over daily life. Her findings showed that even L1 speakers lose lexical access over decades of disuse, and that children are far more vulnerable than adults: a child who emigrates can lose the native language almost completely, while an adult who emigrates after puberty retains a robust, if slowed, L1.

The other side of this coin is that the brain hoards even accidental exposure. A 2020 study in Scientific Reports (Oh, Matchin, and colleagues) showed that non-Māori-speaking New Zealanders, exposed to Māori only through ambient culture — place names, TV, signage — had built a measurable "proto-lexicon": implicit knowledge of which strings of sounds are Māori words, without knowing their meanings. You absorb language even when you are not trying. That is why a language you once studied seriously is never a blank slate.

What This Means for Your Flashcards

The research translates into a concrete maintenance strategy. Here is the protocol the studies above suggest.

1. Never let the first six years go silent

Bahrick's curve drops hardest in the first three to six years after learning stops. A tiny maintenance routine in that window — even five minutes a day of reviews — is disproportionately powerful, because it prevents the steep part of the decline from ever happening.

2. Keep a "permastore" deck, not a "restart" plan

If you abandoned a language, do not buy a beginner course and start from lesson one. Run a reactivation review: pull up your old deck, let the algorithm surface your old cards, and relearn. Thanks to the savings effect (de Bot & Stoessel, 2000), you will move through material at several times the speed of a new learner. Your old intervals can be rebuilt aggressively.

3. Overlearn the words you truly need

Bahrick's clearest actionable finding: overlearning buys permanent retention. For the words you actually use — your core vocabulary — keep reviewing them past the point where you "know" them. In a spaced-repetition system, that means letting intervals grow into months and years rather than deleting mastered cards. The FSRS algorithm in FluentCards does exactly this: as a card's stability rises, reviews space out to months, which is precisely the overlearning-plus-spacing combination Bahrick's data rewards.

4. Recognize more than you recall, and accept it

Bahrick found recognition outlasts recall. When a word is on the tip of your tongue in conversation, your brain is showing you the permastore at work — the trace is intact even when production stalls. Practice both directions: recognition cards (see the word, know the meaning) keep your passive vocabulary alive for reading and listening, while recall cards (see the meaning, produce the word) protect your speaking. A balanced deck does both.

5. Protect your weakest language first

Jessner's 16-month data showed the weaker language is the first to bleed. If you study two languages, spend your maintenance minutes on the newer one — it has the most to lose. The stronger one can coast on accumulated stability.

6. Use it — in any form

Every study above converges on the same variable: use. But use is not only conversation. Reading a page of native content, listening to a podcast, shadowing a dialogue, or reviewing a deck all count as activation. The point is not intensity; it is regularity. A language kept warm with ten minutes a day will outperform a language hit with a three-hour session once a month.

The Honest Summary

Language attrition is real, and it moves faster than optimism. But the research also dismantles the worst version of the fear. You do not lose a language the way a file is deleted from a hard drive; you lose it the way a trail overgrows. The path is still there, and the second clearing is faster than the first. Spaced repetition is the tool for the job because it is, at its core, an anti-attrition machine: it schedules a review at the moment your memory is about to slip, converting the forgetting curve into a maintenance schedule.

So if you are staring at a deck you have not touched in months, the science says: open it. The words you think you lost are still there, waiting to be reactivated — and every card you flip is one more word moved into the permastore.

Further reading: the science of retrieval practice (the testing effect), how FSRS schedules your reviews, why interleaving and spacing build durable memories (desirable difficulties), and how sleep consolidates the vocabulary you reviewed today.