Input vs Output in Language Learning: What Krashen and Swain Actually Claimed
A careful comparison of Krashen’s Input Hypothesis and Swain’s Output Hypothesis, the evidence limits, and a practical French input-output learning loop.
Materially updated
Input and output do different work, and the research does not supply a universal daily ratio. Krashen hypothesized that acquisition develops through understanding input and that production emerges; Swain later proposed that producing language can, under some conditions, facilitate learning by making a learner notice a gap, test a hypothesis, or reflect on form. Swain did not claim output creates all—or even most—language competence, and one output study cannot prove Krashen wholly right or wrong.
Online arguments often turn this into a team sport: input believers versus speaking believers. That framing erases the actual claims and produces bad advice. Reading and listening supply language to interpret. Speaking and writing create a need to select and organize language. Feedback and another encounter with input can connect the two.
What counts as input and output?
| Process | Examples | Immediate question |
|---|---|---|
| Input | Listening to a conversation, reading an article, following a story, inspecting a model answer | What does this message mean, and how is that meaning expressed? |
| Output | Explaining an idea, asking a question, writing a message, retelling a story | How can I express this intended meaning with the French available to me? |
| Interaction and feedback | Clarification, reformulation, correction, comparison with a model, collaborative problem-solving | Was the message understood, and what form or choice needs revision? |
Activities can change category during one session. A conversation includes incoming language, outgoing language, and responses to each. Reconstructing a text begins with input, creates output, then returns attention to input.
Krashen's Input Hypothesis
In his 1982 formulation, Stephen Krashen proposed that a learner moves from current competence i toward the next stage i+1 by understanding input containing that next structure while focusing on meaning. Context, world knowledge, and non-linguistic information help make the message comprehensible.
Krashen also proposed that:
- when enough communication is understood, i+1 will occur without a teacher precisely sequencing one next grammar item;
- input should be roughly rather than perfectly tuned;
- production ability emerges over time rather than being directly taught into existence.
These are theoretical claims of the Input Hypothesis. “Understand meaningful language” is not the same instruction as “sit passively near French audio.” Attention to the message is central to Krashen's own wording.
For a calibrated reading and listening routine, use the companion comprehensible input for French guide. For the broader history of the model, see the Krashen i+1 overview.
Why Swain proposed an Output Hypothesis
Merrill Swain developed her argument in the context of research on Canadian French immersion. Students could receive extensive meaningful French input and develop strong comprehension, yet their productive language did not necessarily become fully target-like in every respect. In a 1985 book chapter, she argued for a role for comprehensible output in the development of communicative competence.
The careful version of the later claim is not “output replaces input” or “speaking is half of acquisition.” Swain and Lapkin wrote that output may, sometimes and under some conditions, facilitate second-language learning in ways that differ from or enhance input.
Three proposed learning functions of output
1. Noticing a gap
Understanding a message does not always require processing every grammatical form. Production can expose a missing resource: “I understand what she regrets, but I cannot construct that relationship in French.” Swain and Lapkin's 1995 paper examined moments when learners producing a second language became aware of a linguistic problem and modified their output.
2. Testing a hypothesis
When a learner tries a form, the response can provide information. The interlocutor may understand, ask for clarification, reformulate the phrase, or correct it. The learner's first version is a hypothesis about how French expresses the meaning; the next version tests a revision.
3. Metalinguistic reflection
Learners can use language to discuss language: compare two forms, explain a choice, or negotiate a correction. This reflection may help consolidate a form-function relationship. Merely naming a grammar term is not the goal; the useful question is why one choice expresses the intended meaning better.
These functions are possibilities, not automatic consequences of opening your mouth. Repeating an error without attention, relevant input, or feedback may not create the same process.
What did empirical studies find?
The results are informative but conditional.
- Swain and Lapkin (1995) analyzed cognitive processes around language problems during output and presented evidence that producing language could trigger noticing and modified output.
- Izumi and colleagues (1999) tested output's proposed noticing function across two task phases. One phase produced noticing and immediate incorporation without a post-test effect; another produced improvement on a later post-test. The mixed pattern matters.
- Izumi (2002) studied adult English learners working on relative clauses. In that experiment, output-plus-input activities produced greater learning gains than input used only for comprehension, while visual input enhancement did not produce comparable measurable gains.
These studies concern particular learners, target forms, tasks, and outcome measures. They support investigating mechanisms such as noticing; they do not establish that every minute of output beats every minute of input or that every skill requires the same schedule.
Krashen vs Swain: the real difference
| Question | Krashen's Input Hypothesis | Swain's Output Hypothesis |
|---|---|---|
| What drives the proposed process? | Understanding messages containing language beyond current competence | Producing language that may push attention toward gaps, alternatives, and form |
| What is the learner primarily doing? | Going for meaning with contextual support | Trying to express a meaning precisely enough for the task or interlocutor |
| What is claimed about production? | Production ability emerges; it is not the source of acquired competence in the model | Output can facilitate learning under some conditions and does more than display existing knowledge |
| What does the theory prescribe per day? | No universal minute ratio | No universal minute ratio |
Swain's proposal challenges a strong claim that comprehensible input alone accounts for everything learners need, but it does not prove that Krashen's emphasis on understandable input was useless. The hypotheses focus attention on different processes.
A practical input → output → input loop
A good loop gives output something meaningful to express and gives the next input pass a concrete question.
- Understand: read or listen to a short French message and recover its main point.
- Close the source: summarize or respond without copying.
- Mark the gap: note where you could not express the intended time, relationship, detail, or stance.
- Return to input: inspect how the source or a reliable model expresses that function.
- Revise: produce a second version that solves one specific gap.
- Transfer later: use the form in a new message or conversation rather than memorizing the original.
This loop operationalizes the possibility Swain highlighted without pretending that every gap is immediately acquired.
Original French example
Use this short passage as input:
La bibliothèque du quartier fermera plus tôt vendredi en raison de travaux. Les livres pourront être rendus dans la boîte extérieure. En revanche, les documents réservés ne seront disponibles qu'à partir de lundi.
- Meaning check: Why is the library closing early? What remains possible? What is delayed?
- Output: Leave a 30-second voice message explaining the change to a friend.
- Gap check: Could you express cause, contrast, and the time restriction?
- Input return: compare your version with en raison de, en revanche, and ne… que in the source.
- Transfer: explain a different service disruption using one cause and one contrast.
The important event is not simply “speaking happened.” It is the sequence of intended meaning, attempted form, noticed gap, relevant model, revision, and later reuse.
Choose the balance by the bottleneck
If very little French is understandable
Increase supported input: shorter messages, visual context, familiar subjects, graded material, and French transcripts. Output can remain brief and low-pressure—pointing, choosing, completing a known frame, or giving a one-sentence summary.
If comprehension is much stronger than speaking
Add tasks that force a message beyond memorized phrases: retell, ask for missing information, compare two options, or defend a choice. Record the exact gap, find a model, and try again. The guide to practising French speaking alone contains repeatable drills.
If writing contains recurring errors
Use short output, targeted comparison, and delayed reuse. The French error log template helps preserve the mistaken choice, contrasting form, and retrieval prompt without logging every typo.
If the goal is a four-skill exam
Prepare the assessed behaviours. Reading and listening require broad, unfamiliar input under constraints; speaking and writing require complete output under time. Test results by component can reveal which process needs more training, but a practice percentage is not a language-acquisition theory.
Four common myths
“Input is passive”
Input can be processed shallowly or deeply. Following purpose, inference, stance, and discourse relationships is active meaning work even when no speech is produced.
“Any speaking automatically teaches you”
Output may create useful noticing or hypothesis testing, but repetition without attention or a relevant model does not guarantee a change.
“Swain proved Krashen wrong”
Swain proposed an additional facilitative role for output and developed evidence around particular processes. That is not a single experiment falsifying every claim in Krashen's framework.
“Research says spend 80% on input and 20% on output”
Neither hypothesis supplies that universal ratio. Any schedule should reflect current comprehensibility, the learner's bottleneck, the task, and the goal.
Build an evidence loop. Use the InfinLume diagnostic to locate fragile grammar, the InfinLume app for structured French input and practice, and your own recordings or writing to test what you can produce. A diagnostic result should guide a task, not become a complete claim about proficiency.
Frequently asked questions
Is input or output more important for language learning?
The question is too broad for one evidence-based answer. Input provides language to understand; output may facilitate noticing, hypothesis testing, reflection, and production practice under some conditions. The useful balance depends on the learner and goal.
What is Swain's Output Hypothesis?
It is the proposal that producing language can sometimes facilitate second-language learning in ways different from or complementary to input, including by triggering noticing, hypothesis testing, and metalinguistic reflection.
Did Swain say output causes most language acquisition?
No. The careful claim associated with Swain and Lapkin is facilitation sometimes, under some conditions—not responsibility for all or most competence.
Can I learn French only by watching videos?
Videos can provide valuable input when the French message is being processed. Whether that alone meets your goal depends on the goal. If you need to speak or write, use tasks that require those behaviours and give you a chance to compare, revise, and reuse language.
Should I correct every output error?
No universal rule requires that. Prioritize errors that block meaning, recur, or concern the target of the task. Too many simultaneous corrections can make the next revision unfocused.
Primary sources
- Krashen (1982): Principles and Practice in Second Language Acquisition
- National Diet Library bibliographic record for Swain (1985), “Communicative competence: Some roles of comprehensible input and comprehensible output in its development”
- Swain & Lapkin (1995): “Problems in Output and the Cognitive Processes They Generate”
- Izumi, Bigelow, Fujiwara & Fearnow (1999): “Testing the Output Hypothesis”
- Izumi (2002): “Output, Input Enhancement, and the Noticing Hypothesis”