What is comprehensible input? The research, the critics and how to use it

Here is a sentence in Spanish:

Cuando llegamos a la estación, el tren ya se había ido, así que esperamos una hora en el andén.

If you know some Spanish, you probably understood it, even if andén was new to you. A station, a train that had already left, an hour of waiting: an andén must be a platform. In that moment you did something that the linguist Stephen Krashen argued is the engine of language learning. You understood a message that was slightly beyond what you knew, and the meaning carried you to the new word.

Krashen called this comprehensible input. It’s one of the most influential ideas in language teaching, and one of the most argued about. This article explains what Krashen actually proposed, what the research supports, what his critics say, and how to use the idea without falling for the slogans.

Where the idea comes from

Krashen set out his theory in Principles and Practice in Second Language Acquisition (Krashen 1982), which he has since made freely available on his website, and developed it in The Input Hypothesis: Issues and Implications (Krashen 1985). The theory rests on five hypotheses:

1. Acquisition and learning are different. Acquisition is the subconscious process by which children pick up their first language. Learning is conscious knowledge about a language, such as grammar rules. Krashen argued that acquisition is what produces fluent use.
2. Natural order. Learners tend to acquire grammatical features in a predictable order, regardless of the order in which they’re taught.
3. The Monitor. Learned rules work mainly as an editor, letting you correct what you say or write when you have time and are thinking about form.
4. The input hypothesis. We acquire language by understanding input slightly beyond our current level. Krashen wrote this as i + 1, where i is what you can already handle.
5. The affective filter. Anxiety, low motivation and low self-confidence can block input from being acquired, even when it’s understood.

Two details are often lost in retellings. First, Krashen described understood input as necessary but not sufficient on its own; in his model, the other condition is a learner who is relaxed and open to it. Second, he became a strong advocate of reading in particular. In The Power of Reading (Krashen 2004, first published 1993), he argued that free voluntary reading, reading what you choose for pleasure, is one of the most powerful sources of vocabulary, grammar and literacy development.

What the research supports

Krashen’s theory as a whole has been heavily criticized, as we’ll see. But its central practical claim, that large amounts of understood input drive progress, has substantial support, especially for reading.

The Fiji “book flood.” In a study often cited in this debate, Elley and Mangubhai (1983, “The impact of reading on second language learning”) gave 250 high-interest storybooks in English to each participating class in eight rural Fijian schools that had few books. After eight months, the pupils in these classes had progressed in reading and listening comprehension at about twice the normal rate of matched pupils in the regular program. After twenty months, the gains had grown and spread to other language skills.

Reluctant readers in Japan. Mason and Krashen (1997, “Extensive reading in English as a foreign language”) reported on Japanese university students, among them a class of students who had previously failed English. After a semester of reading graded readers, the class that had failed nearly caught up with a traditionally taught class on a cloze test, and their attitudes toward English improved markedly.

Meta-analyses. Studies that pool many experiments point the same way. Nakanishi (2015, “A meta-analysis of extensive reading research”) combined 34 studies and found a medium-sized positive effect of extensive reading on reading proficiency. Jeon and Day (2016, “The effectiveness of ER on reading proficiency: A meta-analysis”) combined 49 studies and found positive effects of small to medium size, larger for adults than for children.

Comprehension depends on known words. Research on vocabulary coverage gives “comprehensible” a practical meaning: most learners need to know about 98% of the words in a text to understand it adequately without help (Hu & Nation 2000, “Unknown vocabulary density and reading comprehension”). More on that in the 98% rule.

What the critics say

Most of the criticism isn’t that input doesn’t matter. It’s that Krashen’s account of how it matters is too vague, and that input alone isn’t the whole story.

It’s hard to test. Barry McLaughlin (1987, Theories of Second-Language Learning) and Kevin Gregg (1984, “Krashen’s monitor and Occam’s razor”) argued that key terms were never defined precisely enough to be tested. How would you tell whether a learner had “acquired” or “learned” a structure? How do you measure someone’s i, let alone i + 1?

Input without output leaves gaps. Merrill Swain (1985, “Communicative competence: Some roles of comprehensible input and comprehensible output in its development”) studied grade 6 students in Canadian French immersion programs. After years of rich, understandable French, they performed like native speakers on some measures, but their grammar was still not native-like. Swain argued that producing language, which she called comprehensible output, pushes learners to process grammar in ways that understanding alone doesn’t.

Conversation and feedback help. Michael Long’s interaction hypothesis (Long 1996, “The role of the linguistic environment in second language acquisition”) holds that input is especially useful when learners negotiate meaning in conversation, asking what something means or being misunderstood, and that feedback may help, at least for vocabulary and some grammar.

You have to notice things. Richard Schmidt (1990, “The role of consciousness in second language learning”) argued that learners must consciously notice a feature of the input for it to be learned. Input that flows past unnoticed may never become knowledge.

Teaching works too. A large review by Norris and Ortega (2000, “Effectiveness of L2 instruction: A research synthesis and quantitative meta-analysis”) found that focused instruction produced substantial gains on the features taught, that explicit instruction was more effective than implicit instruction, and that the effects of instruction lasted. The authors cautioned that the way gains are tested affects the results.

Read together, these studies suggest a balanced picture. Understood input is the foundation, and most learners get far too little of it. But speaking, writing, conversation and some deliberate attention to form help too, especially for accuracy.

What “comprehensible” means in practice

You can’t measure i + 1, but you can feel it. Here’s the same moment written three ways in Spanish. Imagine you’re an intermediate learner.

Too easy:

Jim abre la puerta. Entra un hombre viejo. Tiene una cicatriz en la cara.

(Jim opens the door. An old man comes in. He has a scar on his face.)

About right:

Cuando abrí la puerta, entró un marinero viejo con una cicatriz que le cruzaba la mejilla.

(When I opened the door, an old sailor came in, with a scar across his cheek.)

Too hard:

Apenas hube descorrido el cerrojo, irrumpió en la posada un lobo de mar curtido, con la mejilla surcada por un costurón lívido.

(No sooner had I drawn back the bolt than a weathered old sea dog burst into the inn, his cheek furrowed by a livid scar.)

The first version is perfectly comprehensible, but it has little new to offer. It’s useful for building speed, not for meeting new language. The third is literary Spanish, with a rare tense (hube descorrido) and a run of words most intermediate learners won’t know: cerrojo, irrumpir, curtido, surcada, costurón, lívido. Without a dictionary at every turn, it isn’t input at all, just noise. The middle version is the sweet spot: one or two new words, surrounded by enough that’s familiar to work them out.

Notice also that “comprehensible” depends on more than the words. Audio, a story you already know and a gloss for a tricky word can all turn a slightly-too-hard sentence into one you understand.

How to use comprehensible input

1. Read a lot of text you mostly understand. Aim for pages where you know nearly every word. See how to choose your reading level.
2. Choose stories you want to finish. Interest keeps you reading, and reading volume is what moves you forward. Classic novels are a good place to start.
3. Add audio. Reading along with a recording links spelling to sound and keeps you moving. See reading while listening.
4. Look up selectively. Check only the words that block the meaning, or that keep coming back. See dictionary or context.
5. Reread and read narrowly. Rereading a chapter, or reading several books by the same author or on the same topic, brings the same words and structures back in new combinations. Krashen calls the second habit narrow reading (Krashen 2004, “The case for narrow reading”).
6. Notice a little. After a session, pick one structure you met several times and look it up. Schmidt’s work suggests that attention helps.
7. Use what you read. Retell a chapter aloud, write three sentences about it, or talk it over with a friend or a language partner. Swain’s work suggests that output does work input can’t.
8. Be patient. In the Fiji study, the gains built over months, not days. Input works by accumulation.

Common misunderstandings

“Just watch TV in the language and you’ll absorb it.” Only if you understand it. For a beginner, most native television is noise, not input. Start with material made for your level.

“Grammar study is a waste of time.” The research doesn’t support that. Instruction has measurable effects; the case for input is that it’s essential, not that it’s the only thing that works.

“Comprehensible means easy.” It means understood. A text at the edge of your level can be comprehensible with the right support, and a very easy text, while useful for fluency, adds little that’s new.

How Lexo helps

Lexo is an iPhone and iPad app for getting large amounts of comprehensible input from great stories. Classic novels such as Treasure Island, Pride and Prejudice and The Count of Monte Cristo are rewritten at CEFR levels A2 to C2 in the language you’re learning, from Spanish to Japanese, and you choose the level for each book.

When a sentence sits just beyond your level, Lexo helps make it comprehensible. Tap any word to see what it means in that sentence, the meaning of the phrase it belongs to, and a translation of the whole sentence into your own language. Every book is narrated, with the words highlighted as they’re read, and a sentence-by-sentence mode slows things down for difficult passages.

Lexo is a reading app: it gives you input. Pair it with conversation, writing or a class for the rest. It’s in beta and launching on the App Store soon. Get notified at launch, or see how Lexo’s levels work.