Most writing systems are accidents of history. Korean's was engineered on purpose — by a king who wanted everyone to be able to read.
Swedes need subtitles to watch Danish television. The two languages are close enough on the page to share a newspaper — but the moment a Dane opens their mouth, the Swede is lost.
Polli generates stories at your level in 21 languages. Every word is tracked, every pattern reinforced.
If you spend any time in language learning communities, you'll encounter the gospel of spaced repetition. It goes like this:
"Use Anki. Make flashcards. Review them daily. The algorithm will schedule reviews at optimal intervals based on the forgetting curve. This is the most scientifically-backed method of memorization in existence."
Every word of this is true. And it leads people spectacularly astray.
Spaced repetition systems work. Hermann Ebbinghaus demonstrated the forgetting curve in 1885. The original numbers come from his self-experiments memorizing nonsense syllables, but the general shape has held up: memory decays steeply at first, then more slowly, and review resets the curve. The exact percentages get repeated more rigidly than the research warrants, but the principle stands. SRS schedules reviews at expanding intervals to catch memories before they fade.
The key research comes from Cepeda, Pashler, Vul, Wixted, and Rohrer (2006), who conducted a meta-analysis of spacing effects across 254 studies. Their finding: distributed practice (spacing reviews over time) consistently outperforms massed practice (cramming) for long-term retention. The optimal spacing interval depends on the retention interval, but the principle is universal. Spacing works.
For vocabulary recognition, seeing a word and knowing what it means, SRS is excellent. If your goal is to build a passive vocabulary as fast as possible, spaced repetition flashcards are one of the most efficient tools available.
The problem starts when people confuse vocabulary recognition with language acquisition.
Knowing what a word means when you see it on a flashcard is not the same as using that word in a sentence. Production uses different muscles from recognition. You can build cards that target production, cloze deletion, audio prompts, or full sentences. Most vocabulary SRS, though, is still the familiar "word on the front, definition on the back" format. That's recognition practice with extra steps.
Understanding a word in context is different too. A word on a flashcard has one meaning and zero context. In a sentence, words are modified by grammar, colored by register, and constrained by collocation. Knowing correr means "to run" doesn't prepare you for se me corre el maquillaje ("my makeup is running").
Grammar intuition doesn't come from flashcards either. Grammar is a set of patterns your brain extracts from thousands of examples. Reviewing "the dative is used for indirect objects" doesn't help you feel when a sentence needs the dative. That intuition comes from encountering datives in context, hundreds of times, until the pattern becomes automatic.
Flashcards are silent and static. Real speech is fast, connected, full of reductions and elisions. Fluency isn't knowing a lot of words; it's accessing the right words quickly, in the right order, with the right morphology, under time pressure. SRS trains retrieval of isolated items. Fluency requires retrieval of items in combination, in real time.
That doesn't make SRS bad. It makes it narrow. SRS is a tool for retaining isolated vocabulary items, and it does that task brilliantly. The error is treating it as a complete method.
SRS is addictive because it provides visible, measurable progress.
Every day, Anki tells you exactly how many cards you reviewed, your retention rate, your streak, your mature card count. These numbers go up over time. They feel like progress. They are progress, of a specific kind.
But the metrics SRS measures (cards reviewed, retention rate) correlate weakly with the thing learners actually want (communicative competence). A learner with 10,000 mature Anki cards and no reading or listening practice will have an impressive passive vocabulary and almost no ability to use it.
The danger isn't that the learner is wasting time. They're building vocabulary knowledge. The danger is opportunity cost. Every hour spent reviewing flashcards is an hour not spent reading, listening, speaking, or writing. And for overall language acquisition, those activities are more impactful per hour than flashcard review.
Research by Paul Nation (2013) shows that vocabulary acquired through extensive reading produces deeper word knowledge than vocabulary learned through decontextualized study. It's slower per word, but the word arrives with grammar and meaning attached. You learn fewer words per hour, but you learn them better, and you learn grammar, collocations, and reading fluency at the same time.
SRS optimizes for retention: keeping individual items in memory. Language learning requires acquisition: internalizing a system of interrelated patterns so deeply that you can use them automatically.
Retention and acquisition aren't the same thing. You can retain 5,000 flashcard answers (retention) without being able to construct a single paragraph (acquisition). Conversely, a learner who reads extensively may not "know" a word on a flashcard but can use it correctly in context, because they've acquired it as part of a network of associations, not as an isolated fact.
Stephen Krashen draws a similar distinction between "learning" (conscious knowledge of rules) and "acquisition" (unconscious competence). SRS enhances learning. Input-based activities drive acquisition. Both are useful, but acquisition is what produces fluency.
There's one category where SRS earns its reputation completely: closed, finite systems of atomic items.
Japanese kanji is the canonical case. There are roughly 2,000 jōyō kanji forming the backbone of general literacy. Each character is a discrete unit — a visual form mapped to readings and meanings. The core set is fixed. Compounds and context-sensitive readings exist (which is why SRS alone won't make you a fluent reader), but the underlying recognition load is finite in a way that productive vocabulary isn't. Grinding through the set with Anki dramatically accelerates the path to functional reading; you still need vocabulary, grammar, and context, but the character bottleneck moves out of the way.
The same logic applies to medical and legal vocabulary, anatomical terms, and chemical formulas. SRS is strong anywhere the content is itemized, bounded, and primarily about recognition rather than generation. It's the right tool for memorizing the Linnaean classification of beetles or the names of muscles. It's not the right tool for learning to talk about beetles.
The lesson isn't that SRS is bad. SRS is brilliant at its native problem: finite recognition. It gets less useful as the task moves toward open-ended generative competence. Language vocabulary sits awkwardly in between. Each word is atomic, but using a word well requires non-atomic context. Kanji is closer to "atomic" than "Spanish vocab," which is why SRS feels different at the two ends.
The strongest position, supported by both research and practical experience, is SRS as a supplement to extensive input, not a substitute for it.
A productive daily routine might look like:
In this model, reading does the heavy lifting. SRS fills in the gaps by catching words you've encountered once or twice in reading and ensuring you see them again before you forget. It's a support tool, not the main event.
This is closer to how SRS was originally conceived. Paul Pimsleur's graduated-interval recall system, developed in the 1960s, was designed as a component of a broader instructional program, not as a standalone method. The Anki community's elevation of SRS to a complete learning system is a case of tool-worship: the tool is so elegant that people forget it's only doing one job.
There's another problem with SRS-as-method: it's psychologically punishing at scale.
Anki decks grow. Every new card you add increases your future review load. Heavy users routinely end up doing hundreds of reviews per day, sometimes deep into the multi-hundreds, depending on how aggressively they add new material. The daily count becomes an obligation. Missing a day creates a review backlog that takes hours to clear. The tool that once felt empowering starts feeling like a second job.
This isn't a discipline failure. It's a systems design problem. SRS creates compounding obligations without a natural stopping point. The solution is not more willpower. It's recognizing when the tool has outgrown its usefulness and redirecting that time toward input.
If your Anki reviews feel like work, that's not a sign you need more willpower. It's a sign the balance has shifted and reading would serve you better.
Spaced repetition is one of the most effective memorization techniques ever developed. It is not, by itself, a language learning method.
Use it for vocabulary support. Don't use it as your primary learning activity. And when someone in a language learning forum tells you that Anki is "all you need," remember what Ebbinghaus actually studied: the retention of meaningless syllables. SRS was born from research on memorizing nonsense. That's a real achievement. It just isn't the same thing as language acquisition.
If your Anki reviews are starting to feel like a second job, here's where to put that hour instead: Reading Your First Book in Another Language →