How language acquisition works (in plain English)

Most arguments about language methods are really arguments about one question: how does a language get into an adult head and stay there? Research on second language acquisition has worked on that question for about sixty years. It has not produced a single agreed theory, but it has produced a useful map.

This page walks through the map in plain words: the difference between learning and acquisition, then input, output, interaction, noticing, memory and automaticity, and finally what it all means if your language has tones, characters or particles.

Learning and acquisition are not the same thing

The distinction comes from Stephen Krashen, who made it the first of his five hypotheses in the late 1970s and early 1980s. It is the most useful single idea in the field, and also the most disputed.

Learning, in the classroom sense, is conscious. You study a rule, memorise a word list, fill in a worksheet. The result is knowledge about the language: you can explain when Japanese uses を, or state that Mandarin has four tones.

Acquisition is what happens when you use the language for meaning: following a story, arguing about dinner, reading something because you want to know how it ends. It is largely unconscious. The result is the ability to understand and speak without consulting any rule, and often without being able to state one.

	Learning	Acquisition
How it happens	Deliberate study of rules and items	Understanding and using the language for real messages
What you get	Knowledge you can state and check	Intuition: what "sounds right"
Speed of access	Slow, needs attention	Fast, automatic
Where it shows	Tests, careful writing, self-correction	Conversation, listening at natural speed
Typical weakness	Falls apart under time pressure	Slow to build, hard to see growing

Anyone who has passed a grammar test and then frozen at a hotel desk has met the difference in person. So has anyone who speaks their mother tongue perfectly and cannot explain a single rule of it.

In Krashen's account, learned knowledge has only one job. It works as a monitor, an editor that checks what the acquired system produces, and only when you have time and are thinking about correctness.

Can learned knowledge become acquired? The interface debate

This is where researchers part company. The question is whether there is a connection, an interface, between the two kinds of knowledge.

• No interface. Krashen's position: learning never turns into acquisition. They are separate systems, and only understood input feeds the one that matters.
• Strong interface. Robert DeKeyser and others, drawing on general skill learning: a rule you know consciously can become automatic through large amounts of meaningful practice, the way a learner driver stops thinking about the clutch.
• Weak interface. Rod Ellis and others: explicit knowledge does not convert directly, but it helps indirectly. Knowing a rule makes you notice the pattern in what you hear, and noticing is what feeds acquisition.

The middle position is the most widely held today. Studies that compare taught and untaught learners generally find that explicit teaching helps, with the caveat that such studies often measure with tests that favour explicit knowledge. Krashen's distinction survives in a renamed form: researchers now speak of explicit and implicit knowledge, and nearly all agree these are different things, even where they disagree about the traffic between them.

Kungfucius sides with the weak interface, for a practical reason. A rule read once is a signpost. It does not carry you anywhere, but you find the road sooner.

Is input or output more important?

Input comes first, and nobody serious disputes that. You cannot acquire what you have never heard or read. Krashen's claim is stronger: that understood input is not only necessary but sufficient. The details are on the page about comprehensible input.

Output, speaking and writing, adds things input does not. Merrill Swain observed that pupils in Canadian immersion schools understood French very well after years of input and still made persistent errors when speaking. Her explanation: having to produce a precise sentence shows you the hole in your knowledge. You can understand a sentence while ignoring half its grammar. You cannot say it that way.

Interaction combines both. In Michael Long's account, conversation is valuable because it adjusts itself to you. You look puzzled, the other person rephrases, and you receive input tuned to your level at exactly the moment you need it, often with a correction built in.

So the working answer is: mostly input, by a wide margin in hours, with regular output and conversation doing a job that input cannot.

Noticing and interlanguage

Richard Schmidt argued that learners have to notice a feature before it can become part of their language. Hearing a particle ten thousand times is of little use if your attention slides past it every time. The strong form of this claim is disputed. The weaker form, that noticing helps a great deal, is widely accepted.

What you build as you go is called interlanguage, a term from Larry Selinker in 1972. It is your current internal version of the language: a system of its own, with rules, some borrowed from your first language, some overgeneralised, some correct. Your errors are not random. They are the visible rules of the present version.

Interlanguage develops in stages and tends to follow its own order, whatever the textbook's chapter sequence. It can also stop developing in places, usually once communication works well enough. That is fossilisation, and it is why correction matters more at intermediate level than beginners expect.

Memory and automaticity

Two kinds of memory are involved. Declarative memory holds facts: this word means that, this particle marks the object. Procedural memory holds skills: producing the sentence at speed without thinking. Vocabulary starts as the first kind. Fluent grammar and pronunciation live in the second.

Declarative memory fades fast unless refreshed, which is why spaced repetition works: it brings a word back shortly before you would lose it. But a word you can recall on a flashcard in three seconds is not yet a word you can use. Conversation gives you a fraction of that time.

Automaticity is the bridge. It means the language runs without using up your attention, so that your limited working memory is free for what you want to say and not for how to conjugate it. It comes from one thing only: meeting and using the same patterns many times, in real time, in real contexts. This is why hours matter so much, and why knowing about a language and being able to use it are such different conditions.

Reading about push-ups builds no muscle.

What this means for an Asian language

The general picture holds for any language. Several things weigh more heavily for Chinese, Japanese, Korean, Thai or Vietnamese.

• You need more input, not less. A speaker of English learning Spanish gets thousands of familiar-looking words for free. In Chinese or Korean almost nothing looks familiar, so the hours of understood input needed are far higher. The estimates are on the page about how long it takes to learn a language.
• Tones and unfamiliar sounds need deliberate noticing. Adult ears file pitch under emotion. In Kungfucius's experience, a short period of explicit tone training with feedback does what hundreds of hours of plain listening do not.
• Scripts are learned, in the classroom sense. Children in East Asia spend years of school on characters. No adult absorbs them from exposure. This is deliberate study that then makes acquisition through reading possible.
• Small grammatical words are easy to understand past. Particles, measure words and sentence endings rarely block comprehension, so input alone lets you ignore them. Output and correction bring them into focus.
• Politeness levels are acquired socially. Korean and Japanese speech levels can be explained in an afternoon and take years of real situations to get right.

What to do with this

1. Give most of your hours to acquisition: listening and reading you understand, about things you care about.
2. Use learning as the signpost: short explanations, the script, tones, a modest flashcard habit.
3. Speak and write every week with someone who corrects you, so that you notice what input lets you skip.
4. Expect a delay. What you learn today becomes automatic months later, after you have met it many more times.
5. Judge methods by this map. A method that offers only rules, or only input, or only speaking, is selling one part of the picture.