kin

What comprehensible input is

By Jacob Pradels ·

Comprehensible input is material you can understand even though some of it is new — roughly 95 to 98 percent of the words are known, and the rest are guessable from context. The idea comes from Stephen Krashen (1982), and the claim is that a language is acquired by understanding messages in it, not by studying it. The research on vocabulary thresholds supports the practical version: you can follow real text once you know about 98 percent of its words (Hu & Nation, 2000), and real-world coverage takes several thousand words (Nation, 2006). In practice it means reading and listening to things you actually want to read, at a level where you understand nearly all of it, and meeting the new words again and again rather than looking each one up once.

The idea

Stephen Krashen's input hypothesis (1982) puts it plainly: you acquire a language by understanding messages in it, just above your current level. The word that matters is understanding. Reading a page where every third word has to be looked up is not comprehensible, however much grammar is on the page. Reading a page where you get the meaning of almost all of it, and can work out the rest from context, is.

The practical claim is less dramatic than the theory sometimes sounds. Comprehensible input is not a magic technique that makes language learning effortless. It is a description of the condition under which reading and listening actually work, and it explains something most learners notice: the hours that stick are the hours spent understanding something, and the hours spent staring at drills mostly don't.

Where the threshold sits

The most-cited number in this area comes from Hu and Nation (2000), who looked at how much of a text native English speakers understood, and found that comprehension holds up when about 98 percent of the words are known. Later work suggests the line is softer than that — readers do better with more help from context — but not that it's much lower. Under 90 percent, understanding breaks down.

The other half of the picture is how many words that requires. Nation (2006) put the figure for everyday coverage at several thousand words, not a thousand. This is the part that surprises people: to read a newspaper without a dictionary, you're looking at roughly 3,000 to 5,000 word families, and that's for a language as close to English as Spanish. For a language unrelated to yours, the numbers are far higher.

What each level of coverage buys you, roughly
Known wordsWhat you can do
~500Simple exchanges about familiar things; the first few minutes of a conversation
~1,000Greetings, shopping, directions, an introduction
~2,000Simple conversations about everyday life and work
~3,000–5,000Most everyday spoken Spanish; headlines and simple articles
~8,000+Newspapers, films, fiction, most of what a native speaker reads

That last row is why reading and listening in the target language beat translating everything into your own. Translated material is fully understandable but teaches almost nothing: the words never arrive in the shape you'll meet them in. The number of words you need is high, and only the target language gets you there.

What makes a text comprehensible

Waring and Takaki (2003) measured the last one directly: texts about unfamiliar topics needed roughly twice as many glossed words as texts on familiar topics to reach the same comprehension. In plain terms, a story about something you already know is dramatically easier to understand than one about something you don't — and that is why the same story, written for a different reader, is not equally hard for everyone.

Why drilling is so much less efficient

A flashcard presents a word stripped of context, on a schedule the app chose. It teaches you that the word means something, in isolation, once. The research on spaced repetition (Ebbinghaus's own experiments, and the modern review by Cepeda et al., 2008) shows that reviewing at expanding intervals does improve retention of what it shows you. What it doesn't show is that the word transfers into reading. Recognition in isolation and recognition in a sentence are different skills, and the second is the one you need.

Spaced repetition is also, on its own, close to the worst possible use of your attention, because it presents the same small item over and over. It feels like work. Comprehensible input feels like reading, and it can be about anything you like — your partner's hometown, a recipe, the football. The hours happen because you would have spent them anyway.

The part people get wrong

Comprehensible input is not "easy reading". If it's too easy, you learn very little. The new words have to be there, and they have to arrive in a sentence you can mostly follow. The goal is roughly one unknown word for every twenty you already know, which sounds like a lot until you realise that at 3,000 words you already have around 150 such encounters in a typical article — each one a chance to meet a word in the shape you'll meet it again.

And the new words do not stay unknown. Here's the part that matters: a word met once is nearly worthless. A word met six times over six weeks, in six different sentences, in texts you enjoyed, becomes yours. That's why looking it up feels productive and usually isn't — it moves the word out of a context you were about to learn it in, and into a dictionary, once.

What this looks like day to day

If you want to see what this looks like in practice rather than a description of it, how long it takes to learn Spanish has the hour estimates, and the numbers guide has the vocabulary you'll meet first.

Sources

Questions

What percentage of words do I need to know to understand a text?

About 98 percent, according to Hu and Nation (2000). Later research suggests the line is softer — context helps more than that study assumed — but not that it's dramatically lower. Below roughly 90 percent, understanding falls apart.

Does comprehensible input mean I should only read easy material?

No, and this is the most common misreading. If the text is too easy you learn almost nothing new. The target is around 95 to 98 percent known words: mostly understandable, with real new words arriving in sentences you can follow, where context helps you work them out.

Is comprehensible input better than spaced repetition?

They do different jobs, and the best results usually come from both. Spaced repetition is very effective at keeping a specific item retrievable (Cepeda et al., 2008). The gap is that it teaches recognition in isolation, which doesn't necessarily transfer into reading. Comprehensible input is what puts words into sentences and contexts you can actually use them in.

How many words do I need to speak Spanish?

For an English speaker, roughly 350 to 400 guided hours gets to everyday conversation (B1), and 600 to 750 hours gets to professional working proficiency, per the Foreign Service Institute. Those hours only count if they're hours of understanding — which is what makes the method matter more than the number.

Does it work for languages that aren't related to English?

The principle holds, but the numbers change. Comprehensible input still needs enough known words to carry the meaning, so you have to build a larger base before real material becomes readable. The Foreign Service Institute puts Spanish and French at 24 to 30 weeks; for languages unrelated to English it's 88 weeks for the same target.

Sources

  1. Principles and Practice in Second Language Acquisition. Krashen, 1982.
  2. How much Chinese and English vocabulary is needed to read and understand any text?. Applied Linguistics, 2000.
  3. How large a vocabulary needed for reading and listening?. English Language Research Quarterly, 2006.
  4. The effects of incidental and intentional learning of words and their meanings on L2 reading comprehension. System, 2003.
  5. Spacing effects in learning: A temporal ridgeline of optimal retention. Psychological Science, 2008.

Markdown