Does comprehensible input work? What the research says

Jibber Jabber publishes this site and makes an input-based app. That is a commercial interest, not evidence. The sources below concern teaching and language learning. They do not test our app. This is a short research guide, not a systematic review of the whole field.

What "work" should mean

"Does comprehensible input work?" is really several questions. It helps to separate them, because the evidence is different for each:

• Does it build understanding? Listening and reading comprehension.
• Does it build vocabulary and grammar? The knowledge underneath comprehension and speech.
• Does it build speaking and writing? Fluent, accurate production.
• Is it enough on its own? Or do learners also need output, feedback and study?

The evidence is strongest for the first two questions and most disputed for the last one.

What the evidence supports

Input is necessary

Learners need contact with language to learn its words and patterns. Krashen, Long, Swain, and Schmidt all give input an important role, but they explain learning differently. This shared starting point does not establish Krashen's stronger claim about the only cause of acquisition. Also, a message need not be understood perfectly before any learning can occur.

Extensive reading helps

One of the most cited studies is the Fiji "book flood" (Elley and Mangubhai, 1983). Rural schools received 250 high-interest illustrated story books in English for each class. After eight months, pupils who read and shared many stories progressed in reading and listening comprehension at about twice the normal rate of matched classes that followed the usual structured English programme. After 20 months the gains had increased and spread to other language skills. This study supports story-rich reading in that setting. It does not isolate reading from teacher support or establish an adult input-only programme.

Comprehension-first teaching can work well

In the 1960s James Asher developed Total Physical Response (TPR), in which beginners respond to spoken commands with actions before they are asked to speak. Asher (1969) and later studies reported good listening results from this kind of teaching. In Principles and Practice (1982), Krashen reviewed method-comparison studies and argued that methods which supply more comprehensible input, such as TPR and the Natural Approach, did better than grammar-heavy methods. These comparisons do not isolate a single cause. Methods can differ in teacher support, time, activities, and assessment. They do not prove an input-only rule.

Understanding depends on knowing most of the words

Hu and Nation (2000) estimated that most learners needed to know about 98% of the running words in a fiction text to understand it well without help. A larger 2023 replication (Kremmel and colleagues) could not fully confirm that exact figure. These findings concern reading tasks, not a universal threshold for videos or conversation. Pictures and context can support messages that contain unknown words. Check understanding of the message, not only the word count.

What critics argued

From the 1980s, several researchers accepted that input is necessary but argued that it is not sufficient. These are the most important positions.

The output hypothesis (Swain, 1985)

Merrill Swain studied children in Canadian French immersion programmes. They had received years of comprehensible input through school subjects taught in French, and their comprehension was strong. But their speaking and writing remained noticeably less accurate than native speakers'. Swain proposed that producing language, "comprehensible output", plays its own role and can push learners towards more precise language. Later output research developed the roles of noticing gaps, testing ideas, and reflecting on language. These are not three separate findings established by the 1985 chapter alone. The source link below verifies that chapter's publication details through a later article's reference list.

The interaction hypothesis (Long, 1981 and 1996)

Michael Long agreed that comprehensible input is crucial. He focused on how input becomes comprehensible in conversation. When a learner does not understand, speakers repeat, rephrase and check. This "negotiation of meaning" makes input easier to understand and gives feedback at the moment it is useful. In his 1996 chapter, Long argued directly that comprehensible input alone is insufficient for some parts of a language.

The noticing hypothesis (Schmidt, 1990)

Richard Schmidt argued that learners need to consciously notice a feature of the input for it to become "intake". On this view, hearing a form many times is not always enough. Attention matters, and teaching can help learners notice forms they would otherwise miss.

Skill acquisition theory (DeKeyser, 2007)

Robert DeKeyser described language learning as similar to learning other complex skills. Explicit knowledge, such as a rule, can become faster and more automatic through practice. This directly contradicts Krashen's claim that learned knowledge never turns into acquired knowledge.

Evidence for instruction (Norris and Ortega, 2000)

This large research review combined the results of 49 studies published between 1980 and 1998. It found that focused instruction produced substantial gains, and that explicit teaching tended to be more effective than implicit teaching in those studies. The authors noted limits: many studies used tests that favour explicit knowledge, and long-term effects on free speech are harder to measure.

The online "input-only" debate

Online, comprehensible input is often presented as a complete method: watch and listen for hundreds of hours, do not study grammar, and do not speak until much later. Some learners follow this strictly and report good results. Others report a plateau, or find that understanding and speaking feel like two separate skills.

Three points help to make sense of this debate.

1. Krashen himself was not "input-only" in the strict online sense. In Principles and Practice he gave conscious grammar a limited role, mainly for checking writing. He said that speaking helps indirectly by inviting more input, and that taking part in conversation is probably necessary in practice for truly successful acquisition. He also recommended teaching learners simple tools for managing conversations, such as asking someone to repeat or explain.
2. Learner reports are not controlled evidence. Hour-count progress reports are useful and motivating, but the people who post are a self-selected group. The reports do not show what the same learners would have achieved with a different plan.
3. Different goals need different activities. If your goal is to understand films, input is the main activity. If you have a spoken exam in six months, you also need speaking practice and some exam-focused study.

ALG: a separate, stronger idea

Automatic Language Growth (ALG) is often mixed up with comprehensible input, but it is a separate approach with stronger claims. It was developed by the linguist J. Marvin Brown, who taught Thai at the AUA Language Center in Bangkok. Brown first used Krashen and Terrell's Natural Approach there in the 1980s and then developed his own version. With Adrian Palmer he wrote The Listening Approach (1988), and he described the history of ALG in From the Outside In (2003).

ALG goes further than Krashen. It claims that adults can reach near-native ability if they acquire a language through a long period of understanding real experiences, and that trying to speak early, or studying rules consciously, can damage long-term pronunciation and fluency. These claims come mainly from Brown's observations of students in his own programme. ERIC describes his 2003 account as a descriptive report. It is not a controlled comparison. That source does not establish that early speech damages adult learners. ALG practice must not be presented as the established research position.

A Reddit discussion about Brown's book asks whether its theory is reliable. We cite it only as an example of a learner question, not as evidence for or against ALG. The online term crosstalk, a conversation where each person can use a different language, is associated with this community.

A practical combined plan

Paul Nation (2007) proposed a framework that includes input, output, deliberate learning, and fluency. He suggested that a well-designed course gives roughly equal time to four "strands":

This is a course-design proposal, not a universal daily prescription. Our practical advice is to choose a mix for your goals:

1. Start with lots of easy, comprehensible input. This builds listening, vocabulary and intuition.
2. Use light study as support. Look up words that block meaning. Read short explanations of patterns that confuse you.
3. Add speaking when you can follow simple conversations. Begin with low-pressure formats, such as a patient tutor or crosstalk.
4. Keep input as the base. Most of your time can stay on listening and reading, with output and study growing over time.

Jibber Jabber's stage-by-stage roadmap and its overview of language learning methodologies take a similar combined approach. For a beginner's plan, see Comprehensible input for beginners.

What we still do not know

• How many hours you need. Hour-based level guides from online programmes are estimates, not research results. Your first language, the target language, and the type of input all change the answer.
• The best mix for each person. Some learners clearly benefit from explicit grammar; others make good progress with very little. Research describes averages, not individuals.
• Whether delaying speech helps. There is little strong evidence either way for long delays. A short silent period is natural for many learners. A very long one is a personal choice, not a proven rule.
• How modern tools change the picture. Most studies predate interactive subtitles, word lookup and large online libraries of learner video. These tools make more input comprehensible, but they have not been studied as much as books and classrooms.

For a comparison of input with immersion and classroom methods, read comprehensible input vs immersion. For short answers to common questions, see the FAQ.