How many words do you need to read a novel in another language?

The short answer: to read an unsimplified novel written for adults with about 98% of the words known, the level at which most readers can follow a story without help, research on English suggests you need a vocabulary of roughly 8,000 to 9,000 word families.

That number can sound like a wall. It isn’t, for two reasons. The words you need come in a very particular order, so progress is faster at the start than it looks. And you don’t have to reach the number before you start reading novels. This article explains where the figure comes from, what it means, and how learners get there.

First, what counts as a word?

Researchers count vocabulary in a few different ways, and the differences matter.

• Word forms are every spelling you see: walk, walks, walked, walking.
• Lemmas group a word with its inflected forms: walk, walks, walked and walking are one lemma.
• Word families go further, adding closely related derived words: teach, teacher and teaching belong together, as do happy, happily, happiness and unhappy.

The figures in this article are mostly in word families, because knowing one member of a family usually helps you recognize the others.

Languages differ here. A single Spanish verb such as hablar has dozens of forms: hablo, hablas, hablamos, hablé, hablaba, hablaré, hablaría, hable, hablara and more. A Finnish or Turkish noun can take many endings. Most of the research below was done on English, so treat its numbers as a guide to the shape of the problem rather than an exact target for every language.

Where the 8,000–9,000 figure comes from

The best-known estimate is from Paul Nation (2006, “How large a vocabulary is needed for reading and listening?”). Nation used frequency lists built from a large corpus of English and checked how much of various kinds of text each slice of vocabulary covered. For five adult novels taken together, his figures looked like this:

Vocabulary	Share of the words in the novels
The 2,000 most frequent word families	about 88%
4,000 families, plus names	about 95%
9,000 families, plus names	about 98%

Names made up about one and a half percent of the text, and Nation counted them as known, since readers can usually recognize a name as a name. Newspapers needed about the same vocabulary as novels. Spoken English needed less, about 6,000 to 7,000 families for 98% coverage.

Why 98%? Because research on reading comprehension suggests that most learners need to know about that share of the words to understand a text adequately without help (Hu & Nation 2000, “Unknown vocabulary density and reading comprehension”). At 95% many readers can follow with effort and support; below that, most struggle. The 98% rule explains those studies.

Why the last few thousand words are the hardest

Look at the table again. Two thousand word families already cover almost nine words in ten. The next seven thousand add only another ten percent.

That’s the shape of vocabulary in every language. A small number of words, such as the, and, be, have, go and say, appear constantly. Thousands of others, like lantern, cutlass or reluctant, appear rarely. In fiction, the most frequent thousand word families alone account for about four in every five running words (Hu & Nation 2000).

The rare words matter far more than their share suggests, because they often carry the meaning. Here’s the first sentence of Kafka’s The Metamorphosis (1915):

Als Gregor Samsa eines Morgens aus unruhigen Träumen erwachte, fand er sich in seinem Bett zu einem ungeheueren Ungeziefer verwandelt.

“When Gregor Samsa woke one morning from troubled dreams, he found himself transformed in his bed into a monstrous vermin.” Most of the German words are very common: als, eines Morgens, aus, fand er sich, in seinem Bett. A learner with a modest vocabulary knows them. But the words that make the sentence one of the most famous openings in literature, unruhig (troubled), erwachen (to awake), ungeheuer (monstrous), Ungeziefer (vermin) and verwandelt (transformed), are exactly the ones that learner is likely to miss.

Here’s a version we wrote at about A2:

Eines Morgens wacht Gregor Samsa auf. Er liegt in seinem Bett, aber er ist kein Mensch mehr. Er ist ein riesiges Insekt.

(“One morning Gregor Samsa wakes up. He’s lying in his bed, but he’s not a person any more. He’s a giant insect.”) The meaning survives; the rare words are gone. That’s what adapted editions do.

Children’s books aren’t the shortcut they seem

Many learners try children’s books first, on the reasonable assumption that they must be easy. For language learners, they often aren’t.

Webb and Macalister (2013, “Is text written for children useful for L2 extensive reading?”) compared three collections of text of the same size: a school magazine written for native-speaking children aged 7 to 13, a collection of news and fiction for older readers, and a set of graded readers written for learners. To reach 98% coverage, the children’s texts needed about 10,000 word families, as many as the texts for adults. The texts for the youngest children needed even more. The graded readers needed about 3,000.

Children’s writers may use short sentences and simple plots, but they draw on a wide range of concrete, playful and unusual words, because native-speaking children already know a great many. An earlier study of three short novels for young readers, Alice in Wonderland among them, estimated that a reader would need about 5,000 words for comfortable reading (Hirsh & Nation 1992, “What vocabulary size is needed to read unsimplified texts for pleasure?”).

The lesson: “simple for native speakers” and “simple for learners” are different things. Books written or adapted for learners control exactly what matters, the range of words.

How long does it take?

There’s no single answer. It depends on the language, how close it is to languages you know, how much time you have and how you spend it. But research can show the scale of the task.

Paul Nation (2014, “How much input do you need to learn the most frequent 9,000 words?”) calculated how much a learner would need to read to meet the words in each thousand-word band about twelve times on average, which he treated as a reasonable number of encounters for learning. Using a collection of novels, he found:

To meet these words about 12 times each	Read about	Roughly
The 2nd thousand most frequent families	170,000 words	2 novels
The 3rd thousand	300,000 words	3 novels
The 5th thousand	1 million words	9 novels
The 9th thousand	3 million words	25 novels

He assumed that learners would be reading text in which no more than 2% of the words were unknown, which means graded or adapted texts until a learner gets close to the target. At 150 words a minute, five days a week for forty weeks of the year, reading enough to cover the ninth thousand in a single year would take well over an hour and a half a day.

So reading alone is a slow route to the rarest words. Tom Cobb (2007, “Computing the vocabulary demands of L2 reading”) went further, arguing that at the amounts most learners actually read, words beyond the 2,000 most frequent simply don’t come around often enough to be learned from reading alone. That’s why many researchers, Nation among them, recommend combining extensive reading with some deliberate vocabulary study: reading gives words depth and context, and study speeds up the first meeting.

The good news: you can read novels long before 9,000

Adapted editions change the arithmetic. A good adaptation keeps the story and replaces rare words with common ones, so a learner with a few thousand words can read a whole book at 98% coverage, as the graded readers in Webb and Macalister’s study show.

Help lowers the bar. With a quick way to check the occasional unknown word, text at slightly lower coverage becomes readable, especially for shorter sessions.

Stories teach their own words. A novel repeats its key words. By the third chapter of Treasure Island, you’ll probably know ship, map and treasure, whatever your vocabulary was on page one.

Related languages give a head start. An English speaker learning Spanish, French, Italian or Portuguese recognizes many words on sight, like información or nation, though false friends need watching.

Each book makes the next one easier. Reading at your level builds the vocabulary for the level above.

A realistic path, level by level

Here’s how vocabulary and reading tend to line up, using rough vocabulary figures for learners of English from Milton (2010, “The development of vocabulary breadth across the CEFR levels”). His test counted how many of the 5,000 most frequent words learners knew, so it can’t show anything above that ceiling.

Level	Known of the 5,000 most frequent words	Comfortable reading
A2	1,500–2,500	Graded readers and novels adapted to A2
B1	2,750–3,250	Intermediate adaptations; simple articles
B2	3,250–3,750	Upper-intermediate adaptations; easier unsimplified texts with help
C1	3,750–4,500	Lightly condensed novels; many unsimplified texts with some help
C2	4,500–5,000	Unabridged novels

For more on the levels themselves, see the CEFR levels explained.

How Lexo helps

Lexo is an iPhone and iPad app that lets you read whole novels long before you know 9,000 word families. Each of its roughly 100 classic novels is available at five CEFR levels: rewritten for learners at A2, B1 and B2, close to the original and lightly condensed at C1, and complete at C2. You choose the level for each book, so you can read The Metamorphosis in German or A Christmas Carol in English at A2 now and come back to it at a higher level later; another level of a book you own is $1.99 at planned launch pricing.

For the words you don’t know, tap: Lexo shows the meaning in that sentence, the base form, the phrase or idiom the word belongs to and a translation of the whole sentence, and plays the word aloud. You can save words to your Words list as you go. Lexo is in beta and launching on the App Store soon. Get notified at launch.