Comprehensible input: what it is and how to use it

Comprehensible input is language you can follow even though you do not know every word in it. The idea is that you acquire a language mainly by understanding messages in it, in large quantities, and that drills and word lists play a supporting role at most.

As a daily practice it is hard to argue with: read and listen a lot, at a level you can follow, to things you care about. As a complete theory of language learning it has been argued with for forty years, and some of the objections matter a great deal if your target language has tones or characters.

What comprehensible input is

The term comes from Stephen Krashen, an American linguist and professor emeritus at the University of Southern California, who set out his theory of second language acquisition in the late 1970s and 1980s. It consists of five linked hypotheses:

• Acquisition versus learning. Acquisition is subconscious and produces real ability. Learning is conscious knowledge of rules. For Krashen, only acquisition lets you speak fluently.
• The monitor. Consciously learned rules can only check and edit what you say. They cannot generate it.
• Natural order. Grammatical features are acquired in a fairly predictable order that teaching does not change.
• Input. You acquire by understanding input that is slightly beyond your current level.
• The affective filter. Anxiety and boredom block input from doing its work.

The input hypothesis is the centre of it, and it is where i+1 comes from. Here i is what you have already acquired, and +1 is the next small step. You understand the new piece not by studying it but from context: the situation, the pictures, what you know about the world, the rest of the sentence.

In Krashen's account, speaking is the result of acquisition and not its cause. Given enough understood input, it emerges on its own.

What counts as comprehensible

i+1 cannot be measured, so in practice learners use rougher tests.

• You can follow the thread. You know who is doing what and why, without stopping every sentence.
• Unknown words are occasional. For comfortable reading, a common rule of thumb is one or two new words in a hundred. More than a handful per paragraph and you are decoding, not reading.
• Support counts. Video, gestures, a familiar topic, a story you already know in your own language, or target-language subtitles all make harder material comprehensible. Rereading and relistening do the same.
• Understanding the gist is enough. You are not required to be able to translate it.

The old advice from Kungfucius still stands here: if a word is important, it will come back. Frequent words repeat across everything you read and hear, and each meeting adds a little. A word that appears once in a podcast and never again was not worth a flashcard. Forgetting it is your brain being efficient, not lazy.

How to find comprehensible input for Chinese, Japanese and Korean

The awkward stage is the beginning, when nothing made for native speakers is comprehensible. For Chinese and Japanese the writing system stretches that stage out. The answer is material written for learners, used in bulk.

Stage	Chinese	Japanese	Korean
Beginner	Graded stories in Du Chinese, graded readers such as the Mandarin Companion series, learner video filmed with pictures and gestures	Free graded readers from the Tadoku group, learner channels that tell simple stories with drawings	Dialogues and story lessons from Talk to Me In Korean, webtoons with simple everyday speech
Intermediate	Learner podcasts entirely in Chinese, dubbed cartoons, dramas with Chinese subtitles	Simplified news for learners, slice-of-life anime and drama with Japanese subtitles, manga with furigana	Variety shows and dramas with Korean subtitles, children's novels
Advanced	Novels, talk shows, podcasts on subjects you already know	The same	The same

Some principles apply to all three:

• Read and listen to the same thing. Text with audio lets your ear and eye confirm each other. This matters most in Chinese, where the characters give no reliable sound.
• Go narrow. Ten episodes of one show or five articles on one topic repeat vocabulary far more than ten unrelated ones.
• Drop anything boring. Interest is what keeps you understanding for an hour instead of ten minutes.
• Use reading tools without guilt. Pop-up dictionaries turn i+3 into something close to i+1.

Longer lists are on the podcasts page.

Does comprehensible input really work? The criticism

Almost no researcher disputes that large amounts of understandable input are necessary. The dispute is over whether input is sufficient, and whether the theory is precise enough to test.

The output hypothesis. In 1985 Merrill Swain studied children in Canadian French immersion schools. After years of comprehensible input in class they understood French very well, yet their speaking and writing still showed persistent grammatical errors. Her conclusion was that learners also need to produce language. Being pushed to say something precisely makes you notice what you cannot yet say, lets you test your guesses, and makes you think about form. Swain did not claim output matters more than input, only that it adds something input does not. Krashen's reply is that such output is too rare to explain acquisition.

The noticing hypothesis. Richard Schmidt argued in 1990, partly from his own diary of learning Portuguese in Brazil, that learners must consciously notice a feature in the input before it becomes part of their language. Later versions are softer: noticing helps, even if it is not always required. Either way it contradicts the idea that acquisition is wholly subconscious. Critics such as John Truscott have in turn called the noticing hypothesis vague.

The vagueness of i+1. Critics from Barry McLaughlin and Kevin Gregg onwards have pointed out that neither i nor +1 can be defined or measured for a given learner, which makes the hypothesis very hard to test or to prove wrong.

Explicit instruction. Krashen gives conscious grammar study almost no role. The research does not support that strong position. A much-cited meta-analysis by John Norris and Lourdes Ortega in 2000 found that explicit instruction produced larger gains than implicit approaches, and that the gains lasted, although the tests in those studies tended to favour explicit knowledge.

Interaction and speaking. Michael Long's interaction hypothesis holds that conversation helps precisely because it makes input comprehensible: you ask, the other person rephrases, and you get feedback at the moment you need it.

Why input alone is slow for tones and characters

This is Kungfucius's own observation, and it concerns the languages on this site. Adult learners can listen to Mandarin for hundreds of hours and still not hear tones as part of the word, because their ear files pitch under emotion and emphasis. A short period of explicit tone training with feedback, plus imitation drills such as shadowing, fixes this far faster than waiting.

Characters are the other case. Children in China, Taiwan and Japan spend years of school learning them deliberately, after they already speak the language. An adult foreigner will not absorb 2,000 characters from exposure. Deliberate study with spaced repetition gets you to the point where reading can take over. Input is the main course, but these two side dishes are not optional.

Steve Kaufmann and LingQ

Steve Kaufmann is the best-known living advertisement for input-based learning. Born in Sweden in 1945 and raised in Montreal, he joined the Canadian diplomatic service and learned Mandarin full-time in Hong Kong in the late 1960s, then Japanese during a posting to Tokyo. He later spent years in business in Japan. His book The Linguist appeared in 2003, and he has run a YouTube channel on language learning since 2007.

His method is input with very few rules. He listens to and reads large amounts of content he finds interesting, looks up words as he goes, and does not worry when he forgets them, on the grounds that they will return. He uses grammar explanations as an occasional reference, treats flashcards as optional, and starts speaking in earnest once he has enough words to say something. He keeps starting new languages in his seventies and eighties, which is his answer to the claim that adults cannot learn.

Two details are worth noting. Krashen has pointed to Kaufmann as support for the input hypothesis. And Kaufmann's own Mandarin, by his account, came from six to seven hours a day that included three hours with teachers and at least an hour of deliberate character study. Even the most famous input learner did not learn Chinese from input alone.

LingQ is the platform he co-founded with his son Mark to make this routine practical. You import or choose a text with audio, words you have not met are highlighted, and each word you look up is saved as a "LingQ" and tracked until you mark it as known. A browser extension pulls in web articles and video captions. The app counts your known words, which is motivating and should be taken about as seriously as a step counter.

How to start this week

1. Choose one source at your level from the table and commit to 20 to 30 minutes a day. Listening on a commute counts.
2. Do not make a flashcard for every unknown word. Look it up, move on, and let the frequent ones return.
3. If you want cards, take a few sentences from what you read, as described under sentence mining.
4. Add what input does not supply: a little tone or pronunciation drill, some characters, and one conversation or written message a week where someone corrects you.
5. After a month, raise the difficulty one notch. If the new material leaves you lost, go back down. Nobody is checking.