You Learn a Language Mostly by Understanding Things

Stephen Krashen's input hypothesis, formulated in the 1980s, proposed that language is acquired by understanding messages rather than by studying rules — and that the useful input is material just beyond your current level, often written as i+1.

The strong version of that claim has not held up. The practical version has, and it is probably the most useful single principle available to a self-directed learner.

What the evidence supports

Well supportedContested
Large volumes of input are necessaryThat input alone is sufficient
Understanding matters more than decodingThat explicit grammar study is useless
Input drives vocabulary growthThat output contributes nothing
Anxiety impairs performanceThe precise mechanism proposed for it

Nobody seriously disputes that you need a great deal of input. The argument is about what else is required, and the practical consensus is that explicit instruction speeds things up considerably while input does the bulk of the work.

Comprehensible is the operative word. Audio you cannot follow is not input in any useful sense — it is noise, and passive exposure to noise produces very little. The common failure is assuming that having a podcast playing in an unfamiliar language constitutes learning.

Making input comprehensible

The difficulty is that genuinely comprehensible material is scarce at the beginning, which is the point at which you need it most.

TechniqueHow it helps
Visual supportpictures and video carry meaning the words do not
Known contenta familiar story supplies the meaning
Graded materialvocabulary controlled to a level
Repetitionthe fourth pass is comprehensible when the first was not
Transcriptslets you resolve what you missed
Parallel textstranslation alongside supplies meaning directly

Content you already know is the most underused of these. Watching a film you have seen a dozen times, dubbed into the target language, gives you the meaning for free and lets your attention go entirely to the language.

Why it works

Words acquired from context arrive with information a definition omits — register, collocation, the grammatical frame they sit in. And the pattern of a structure is absorbed across many encounters rather than memorised once.

This is why learners with heavy input often have accurate intuitions they cannot explain, while learners who studied rules can state the rule and still produce unnatural sentences. The two kinds of knowledge are different, and both are useful.

Where input alone falls short

GapWhat is needed
Production speedspeaking and writing practice
Noticing fine distinctionsexplicit explanation
Features that carry little meaningtargeted attention — endings, articles
Accuracy in outputfeedback and correction

Low-salience features are the clearest case. Grammatical gender, article choice and agreement endings carry almost no information, so a learner can understand thousands of hours of input without ever noticing them reliably. A five-minute explanation fixes what years of input might not.

A workable synthesis

Treat input as the engine and study as the steering. Spend most of your time understanding things — reading, listening, watching — at a level where you follow the meaning. Use explicit grammar study in short, targeted doses to notice what input alone will not surface. Produce language regularly so that retrieval keeps pace with recognition.

The one thing to avoid is the arrangement most courses default to: a great deal of study about the language and very little contact with it. That produces learners who know the rules, can pass a test and cannot follow a conversation — which is the specific failure the input hypothesis was formulated against, and the part of it that was right.

Lower the cost of every lookup

Input works when you can follow it. NDT Studio's free offline dictionaries make the occasional check instant, across 45 pairs.

Open the dictionary app

Frequently asked questions

What is comprehensible input?

Language you can understand the meaning of, even if you do not know every word — often described as input slightly above your current level. The claim is that acquiring language happens primarily through understanding messages rather than through studying rules.

Is the input hypothesis accepted by researchers?

Its strong form, that input alone is sufficient and explicit study contributes nothing, is not generally accepted. The weaker claim, that large amounts of comprehensible input are necessary and highly effective, is widely supported and uncontroversial in practice.

Does output matter at all?

Yes. Producing language forces you to notice gaps, tests hypotheses about how the language works, and builds the retrieval speed that comprehension alone does not train. Most practitioners treat input as the foundation and output as necessary alongside it.