Spend a week in any language-learning forum and you’ll meet the input orthodoxy. Stop studying grammar. Consume enormous quantities of material you mostly understand. Acquisition takes care of itself. It gets stated with the confidence of settled science.
It isn’t settled.
The core insight holds up well. The strong version has been under attack from within the field for forty years, and the argument is more useful than either slogan. Sources are linked inline so you can check every claim yourself.
Where the idea comes from
It all traces back to Stephen Krashen, whose theory of second-language acquisition, developed from the late 1970s, is still the most influential and most argued-over framework in the field. Two of his claims matter here.
The first is the acquisition–learning distinction: Krashen argued that acquiring a language subconsciously, through exposure, is a fundamentally different process from consciously learning rules, and that only acquisition produces fluent, spontaneous use.
The second is the input hypothesis: we acquire language by understanding input slightly beyond our current level, the famous "i+1." On this account comprehensible input is not merely helpful but the sole causal mechanism.
Anyone who has learned a language mostly by reading and listening will recognise something true in that. The trouble is the word "only."
The objections, which are substantial
Krashen has drawn forty years of criticism from inside the field. The main lines:
- It is difficult to falsify. Acquisition is defined as subconscious, and the "affective filter" can explain away any case where plenty of input produced nothing. A theory that accommodates every possible result isn’t doing much work, whatever its merits.
- Input alone appears insufficient. The best evidence here comes from Canadian French immersion, where English-speaking children spent years soaking in comprehensible input. They came out with excellent comprehension and fluent, confident speech, and grammatical accuracy that plateaued well short of native norms, with the same errors persisting for years. Merrill Swain’s output hypothesis came directly out of watching that happen.
- Noticing seems to matter. Richard Schmidt argued that learners largely acquire the features they consciously notice, which reopens exactly the door the strong input position tries to close.
- Some explicit instruction helps. Reviews of form-focused instruction generally find measurable gains from drawing attention to grammatical form, especially for things easy to miss in the rush of speech.
The immersion programs that generated the best evidence for input also generated the best evidence for its limits. Years of comprehensible input produced fluent speakers who kept making the same grammatical errors.
What about moving abroad?
Living in the country is supposed to be the ultimate immersion. The outcomes are stubbornly uneven.
The broad pattern is clear gains in fluency, confidence and listening, the things constant interaction gives you. Grammatical accuracy improves less reliably, and a certain number of people come home having barely moved.
What predicts the difference isn’t months in the country. It’s how much real interaction you actually had. Live with local speakers and make local friends and you improve. Spend the year in an English-speaking expat bubble and you won’t, however good the view was.
What the evidence does support
Set the theory war aside and a few things are reasonably safe:
- Large amounts of comprehensible input are necessary. Nobody serious disputes this. No one in history has reached fluency through grammar exercises alone.
- It is probably not sufficient by itself. The immersion-programme evidence points fairly clearly at accuracy gaps that input alone doesn’t close.
- Output does work input does not. Producing language forces you to commit to structures you can happily skate past while listening. Both matter and they aren’t interchangeable.
- Attention to form accelerates things. A little grammar, at the moment you’ve noticed something in real input and want it explained, pays off well.
- Comprehensibility is the constraint. Input far above your level does very little, which is the real failure mode of "just watch TV." Unintelligible audio is mostly wasted evenings.
The practical version
What falls out of all this is unglamorous and belongs to neither camp.
Make input the bulk of your time, and make sure it’s actually comprehensible, meaning you following the thread and missing details, not sound washing over you. Add output early, because it trains something listening can’t reach. Treat grammar as a reference you consult when real input has confused you, not a syllabus to march through first.
And drop the idea that being there is the same as immersion. What counts is engaged contact with the language. You can manufacture that at home and fail to get it in Madrid.
If output is the side you are neglecting, as most self-directed learners are, our AI conversation practice exists to make speaking low-stakes enough to do daily, long before you feel ready for it.
Input is the engine. Output and attention are the steering. Anyone insisting you can do without one of them is selling a tidier story than the evidence supports.
