How many Dutch words do you need to read a book?

How many words you need to read: the research says 95 and 98 percent of the page

Researchers do not ask how many words you need. They ask what share of the running words in a text you must know to understand it, and then work out how many words that takes. The question has been studied for nearly forty years, almost always on learners of English.

Batia Laufer (1989) looked for the vocabulary a learner needs to pass a reading test at the University of Haifa, where the lowest pass mark was 55 percent. Her learners underlined the words they did not know, and the share of known words at which they reached that mark was 95 percent: one unknown word in twenty. That figure, the lexical threshold, has been the starting point of the field since.

Marcella Hu and Paul Nation (2000) made the test sharper. They gave learners a short story in four versions, with 80, 90, 95 or 100 percent of the words known (the unknown words were replaced by invented ones), and tested what they had understood. Comprehension rose with every step: at 80 percent nobody understood the story well enough, and at 95 a few readers did and most did not. The authors estimated that reading fiction without help, for pleasure, takes about 98 percent: one unknown word in fifty. A replication published in 2023 could not fully confirm the result, but 98 percent is the figure the paper is cited for.

Laufer and Ravenhorst-Kalovski (2010) measured vocabulary size, text coverage and comprehension on the same learners and found two thresholds: a minimal one at 95 percent of the words known, which took 4,000 to 5,000 word families, and an optimal one at 98 percent, about 8,000 families. They also found that small gains in vocabulary improved comprehension even when they hardly changed the coverage figure.

Paul Nation (2006) asked the question the other way round: how many word families does it take to cover 98 percent of an ordinary text? Counting with fourteen lists of 1,000 word families each, drawn from the British National Corpus, he arrived at 8,000 to 9,000 families for written text and 6,000 to 7,000 for speech.

Schmitt, Jiang and Grabe (2011) tested 661 learners from eight countries on the words of two texts and then on the texts themselves, and found no threshold at all: comprehension rose in a more or less straight line with the share of words known, with no point where it suddenly took off. Every extra percent of coverage helps.

Put together: at 95 percent known you can read with help, a glossary or a dictionary; at 98 percent you read for the story; below 90 percent you are translating, not reading. The rule of thumb on this site, that a text with more than one unknown word in ten is too hard, is the same line.

Word families, base words and word forms

The research counts word families: a base word with its inflected forms and the common words made from it, so that in English read, reads, reading, reader and readable are one family. Eight thousand families is far more than 8,000 words as a dictionary lists them. A frequency list, on the other hand, counts word forms: loopt, liep and gelopen are three entries. The lists on this site sit in between. The forms are grouped under their base word, lopen, and the base words are ranked. A base word is a smaller unit than a family (lopen and de loper count as two), so 1,000 base words is somewhat less vocabulary than 1,000 families, and the 4,000 to 5,000 families of the 95 percent threshold are more than the 4,000 base words of the B1 band below. The count on this page is honest about that: it says what the frequency list covers, not what a given reader knows.

All of it was measured on English

Every study above is about learners of English, with English texts and English word lists. No study of this kind for Dutch as a second language could be found for this page. The nearest is Hazenberg and Hulstijn (1996), who asked how many base words a non-native student needs for university study in Dutch and arrived at about 10,000, more than the 3,000 or 5,000 of earlier Dutch estimates. That is university reading, not a story. For texts written for learners of Dutch there was no figure. The count below is one.

What a count of every text and book on this site shows

Every Dutch text on this site goes through a word check before it is published, and the same check can answer the question for the texts themselves: of the running words on a page, how many are among the 1,000, 2,000 or 4,000 most common Dutch words? The table gives the answer for every free text and every chapter of the books, counted on 3 November 2026, and for one novel of 1860 as a contrast.

How the count works. Every word of a text is read as its base word: the lemma the UD Dutch Alpino treebank gives the form where the treebank has seen it, and otherwise the regular Dutch inflection undone (liep and gelopen under lopen, huizen under huis). A compound that is not itself common counts at the rank of its rarer half, because a reader who knows basis and school reads basisschool. The base words are ranked by wordfreq, a free Dutch frequency list by Robyn Speer (CC BY-SA 4.0) that counts Wikipedia, subtitles, news, websites and social media, with the forms grouped under their base word. A list of 258 everyday beginner words (numbers, food, the house) counts as common at every rank, because a frequency list ranks kaas and twaalf too low for a learner. Names are left out of the count. So “1,000 words” in the table is the share of running words, names excluded, whose base word is among the 1,000 most common or on the everyday list; “2,000 words” and “4,000 words” the same for the larger lists; “Beyond 4,000” the rest.

One thing to know before reading the table: the texts are written to these lists. The house rule is that at least 90 percent of the words of a text fall within its level's band, the 1,000 most common words at A1, 2,000 at A2, 4,000 at B1, and that every word outside the band is explained on the page. So the figures describe texts made for learners, which is the point. They show what a learner with 1,000 words can read, and what the same 1,000 words are worth in a novel.

Scroll sideways to see the whole table

The pattern is the same in every row written for learners. With the 1,000 most common words you know about nine words in ten; with 2,000 between 93 and 95 percent, around the minimum in the research; with 4,000 about 97 percent, close to the comfortable 98. The stories come out a little easier than the texts about the Netherlands (the KNM texts and the news-style B1 texts), whose topics, the government and the history among them, pull in rarer words. Of the books, Saïdjah and Adinda has the most words beyond the 4,000, because its story needs words for a village on Java that no list of modern Dutch ranks high; the glossary carries them.

The last row is Multatuli's Max Havelaar, the novel of 1860 that contains the story of Saïdjah and Adinda, whole and in its original Dutch (Project Gutenberg ebook 11024). The same 1,000 words cover 72.7 percent of it, and even 4,000 leave more than one word in six unknown. Part of that distance is the 19th century: Multatuli's spelling (hy, zyn and myn for hij, zijn and mijn, zoo, menschen) and words no longer in use fall outside a list of modern Dutch, so the row says less about a learner's chances with a modern novel than about the distance between 1860 and today. A count of modern Dutch novels is not possible here, because their texts are not in the public domain. The row still answers the question. A learner with 1,000 words reads the retold Saïdjah and Adinda with its glossary, and does not read Multatuli.

For Dutch at large there is one more figure. On the frequency list itself, which measures all of Dutch across Wikipedia, subtitles, news and the web, the 1,000 most common words cover about three in four running words (the figure is on the page of the 1,000 words). A newspaper is written from the whole language, so at 1,000 words it is nearer that three in four than the nine in ten of a text written for your level. That is why a graded text is readable at 1,000 words and a newspaper is not yet.

What the numbers mean for you

• Learn the first 1,000 first. The 1,000 most common Dutch words are on one page, with their meanings and de or het, and as a PDF and an Anki deck. With them and the everyday words you know about nine words in ten of every A1 and A2 text on this site. In the A1 texts and Fairy Tales the glossary explains the tenth; in the A2 texts and books about half of it, the words between 1,000 and 2,000, is yours to learn.
• Then the next 1,500. The Dutch B1 vocabulary list takes you from word 1,001 to 2,500, by theme. At 2,000 words the A2 texts and books are yours with their glossary.
• Read texts written for your level. At 1,000 words a free A1 text or a book for A1 is readable and a newspaper is not yet; at 2,000 the A2 texts and the A2 books; at 4,000 the B1 texts and easy news. How to learn Dutch by reading has the method.
• The glossary covers the rest. The words outside your band, up to one in ten, are underlined in every text, with their meaning a tap away, so the 95 percent of the research becomes 100 on the page. That is what makes 1,000 words enough for a book written for A1, and 2,000 for one written for A2.
• The honest limit. Knowing a word on a list is not knowing it in a sentence. The count says that koe is among the 4,000 most common words; it does not say that you will recognize koeien in a story. The research measured words that learners knew, by test; a frequency list measures words that are common. The gap between the two closes by reading: every word you meet in a text is a word you half knew from the list and now know in a sentence.

See it in a text

Three sentences from the free texts, with the words outside the 1,000 most common named.

• From Melk op zondag (A1): Achter de huizen staan koeien in de wei. Behind the houses, cows stand in the meadow. Eight words; two are outside the 1,000: koeien (cows; koe is among the 4,000) and wei (meadow; beyond the 6,000). Both are underlined in the text with their meaning.
• From Fietsles (A1): Maar Tesfay kan niet fietsen. Als kind had hij geen fiets. But Tesfay cannot cycle. As a child he had no bike. Eleven words, one of them a name, which does not count. One word is outside the 1,000: the verb fietsen, ranked between 1,001 and 2,000, while the noun fiets is on the everyday list.
• From Trakteren (B1): Op de wc van het kantoor hing een kalender met namen en datums, de verjaardagen van alle collega's. On the office toilet hung a calendar with names and dates, the birthdays of all the colleagues. Eighteen words; four are outside the 1,000: verjaardagen (birthdays; within the 2,000), wc (within the 4,000), kalender (calendar; between 4,001 and 6,000) and datums (dates; datum is within the 2,000, though the count does not join this plural to it). A B1 reader with 4,000 words meets one new word here; a reader with 1,000 meets four, and still follows the sentence, because kantoor, namen and collega's are in the 1,000.

The research on this page

• Laufer, B. (1989). What percentage of text-lexis is essential for comprehension? In C. Laurén and M. Nordman (eds), Special Language: From Humans Thinking to Thinking Machines. Multilingual Matters.
• Hu, M. and Nation, P. (2000). Unknown vocabulary density and reading comprehension. Reading in a Foreign Language 13(1).
• Nation, I. S. P. (2006). How large a vocabulary is needed for reading and listening? The Canadian Modern Language Review 63(1), 59 to 82.
• Laufer, B. and Ravenhorst-Kalovski, G. C. (2010). Lexical threshold revisited: lexical text coverage, learners' vocabulary size and reading comprehension. Reading in a Foreign Language 22(1), 15 to 30.
• Schmitt, N., Jiang, X. and Grabe, W. (2011). The percentage of words known in a text and reading comprehension. The Modern Language Journal 95(1), 26 to 43.
• Hazenberg, S. and Hulstijn, J. H. (1996). Defining a minimal receptive second-language vocabulary for non-native university students: an empirical investigation. Applied Linguistics 17(2), 145 to 163.
• The count: the word check of this site, with the Dutch list of wordfreq (Robyn Speer, CC BY-SA 4.0) and the lemmas of the UD Dutch Alpino treebank; the texts and chapters as published on 3 November 2026. You may quote the table with credit to Dutch Readers.

Read next

Pick a text at your level and count for yourself: read it once and note how many words you had to tap.

• A1: Melk op zondag, a Sunday in a village in Friesland, a visitor at ten and no milk. The book at this level is Fairy Tales, eight tales retold at A1, out now in paperback.
• A2: De fiets van 40 euro, a blue bike bought at the station for 40 euro, and why it was so cheap. The book at this level is Dracula, out now in paperback and for Kindle.
• B1: Trakteren, a first birthday at a Dutch office. The B1 books of the series are coming soon; the whole collection is on one page.

Not sure of your level? The Dutch levels explained says what each level can read, and the free level test takes about ten minutes.