"How many words do I need to know to read a novel?" It is one of the most common questions language learners ask, and it has a surprisingly well-researched answer. The answer is also a little sobering: it is more than most people expect. But it comes with good news, which is that you do not need to wait until you reach that number to start reading real stories.
This article explains what the research says, what the numbers actually mean, and what to do with them.
The key idea: lexical coverage
The useful question is not "how many words do I know?" but "what percentage of the words in this text do I know?" That percentage is called lexical coverage.
Coverage counts running words, not unique words. If a page has 300 words and you know 285 of them, your coverage is 95 percent, even if the 15 unknown words are all different.
Coverage matters because comprehension does not rise smoothly. When coverage drops, the unknown words start to collide: two in one sentence, three in one paragraph, and suddenly you cannot use context to guess any of them.
What the research says
The most cited study on this is by Marcella Hu and Paul Nation (2000). They tested how well learners understood a fiction text at different levels of coverage. Their conclusion, simplified, was that for most learners to achieve adequate comprehension of a text without help, they needed to know around 98 percent of the running words. At lower coverage, some learners still understood a fair amount, but comprehension became much less reliable.
Earlier work, notably by Batia Laufer, had pointed to 95 percent as a threshold for reasonable reading comprehension. Later research by Laufer and Ravenhorst-Kalovski suggested treating these as two thresholds: 95 percent as a minimal level, and 98 percent as an optimal one.
So the two numbers you will see quoted are:
- 95 percent coverage: about one unknown word in every twenty. Reading is possible but effortful, and comprehension is often partial.
- 98 percent coverage: about one unknown word in every fifty. Reading becomes comfortable for most learners.
To make this concrete, a typical paperback page might hold somewhere around 250 to 350 words. At 95 percent, that is perhaps a dozen or more unknown words per page. At 98 percent, it is perhaps five or six. Most people who have tried to read with a dozen unknown words per page will recognise the feeling: it is a puzzle, not a story.
Translating coverage into vocabulary size
How big a vocabulary gives you 98 percent coverage of a novel? Paul Nation's research on English (Nation 2006) estimated that reaching 98 percent coverage of a wide range of written texts, including novels, requires knowing something like 8,000 to 9,000 word families, plus proper nouns. For spoken English, the estimate was lower, around 6,000 to 7,000 word families.
Two caveats are important:
- These are estimates for English. Other languages will differ. Languages with rich inflection or compounding, and languages where the notion of a "word" works differently (such as Chinese, where the relationship between characters and words matters), are not directly comparable. Treat the English figures as an order of magnitude, not a universal law.
- A word family is not a single word. A word family includes a base word and its common inflected and derived forms. In English, "read", "reads", "reading", "reader" and "readable" might count as one family. So 8,000 families represents many more individual forms.
Fewer words give you 95 percent coverage, and the research generally suggests that the first few thousand most frequent words do a disproportionate amount of the work. That is the good news: the most common words are common precisely because they show up constantly, so they are also the easiest to learn through reading.
Why the first few thousand words matter so much
Word frequency is very uneven. In almost any language, a small number of words, such as articles, pronouns, common verbs and prepositions, account for a large share of everything written. Beyond that core, each additional word you learn covers a smaller and smaller slice of text.
This has two practical consequences:
- Early vocabulary has a huge payoff. Learning the most frequent words quickly opens up a large share of most texts.
- Later vocabulary is a long tail. Getting from "most texts are readable with effort" to "novels are comfortable" takes many thousands of less frequent words. These are hard to learn from word lists because each one appears rarely, so they are best picked up through lots of reading.
That second point is the paradox at the heart of reading in a foreign language: you need a big vocabulary to read comfortably, and reading is the best way to build a big vocabulary.
Breaking the paradox: graded reading
The way out of the paradox is to read texts where the coverage is high for you, even though your vocabulary is still small. That is what graded readers are for. By controlling vocabulary at each level, they let a learner with a few hundred or a couple of thousand words read at 95 to 98 percent coverage. If you have not used them before, see what are graded readers.
Graded reading also solves the long-tail problem gradually. Each level introduces new vocabulary at a controlled pace, recycling it so that you meet each new word several times. Over many books, your vocabulary grows towards the size you need for native texts, and you get there by reading rather than by memorising lists.
This is the core argument for extensive reading: a large volume of easy, high-coverage reading.
What about lookup tools?
The 98 percent figure is about reading without help. Tools change the situation somewhat. If you can tap a word and see its meaning in a second, a text at, say, 93 or 95 percent coverage becomes much more manageable than it would be with a paper dictionary.
That does not make coverage irrelevant. Even with instant lookups, a text with many unknown words per sentence is tiring, you will not remember most of the words you look up, and you lose the thread of the story. Lookup tools are best used to stretch slightly beyond your comfortable level, not to read texts far above it. We discuss this balance in tap-to-translate without getting dependent.
How to estimate your own position
You do not need a lab to apply this research. Three practical ways to estimate where you are:
1. The page test. Take a page of the book you want to read. Count words you do not know and cannot guess. Divide by the total number of words on the page. If the unknown share is above about 5 percent, the book is going to be hard work.
2. A vocabulary size test. A test can estimate how many of the most frequent words you know, which tells you roughly which texts are within reach. You can try our vocabulary size test (also available for other languages).
3. A text difficulty check. Paste a passage into a text difficulty tool to see how much of it falls into common versus rarer vocabulary bands.
None of these is exact, and all of them have limitations (for example, recognising a word in a test is easier than understanding it in a sentence). But together they give you a reasonable picture.
A realistic reading ladder
Putting it all together, here is how many learners progress:
| Stage | Typical reading material | What it feels like |
|---|---|---|
| Absolute beginner | Very simple graded stories, short sentences, heavy repetition | Slow but possible, lots of support needed |
| Beginner to lower intermediate | Graded readers at A1–A2 / HSK1–3 | Whole stories become readable |
| Intermediate | Graded readers at B1–B2, learner news, simple non-fiction | Reading becomes fluent at the right level |
| Upper intermediate | Easier native books, books you already know in translation | Possible with some lookups |
| Advanced | Most native fiction and non-fiction | Comfortable, with occasional unknown words |
The move from graded material to native books is not a cliff if you approach it well. Many learners find a mix of graded and native reading works best for a long time. See graded readers vs native books for how to manage that transition.
How ReadNative tracks vocabulary
ReadNative is built around this ladder. Each book retells a real story starting at absolute beginner level and gradually increasing in difficulty, so the coverage stays high while your vocabulary grows. You can tap a word for a gloss or a sentence for a translation when you need it.
The app also counts a word as "known" once you have met it three or more times without looking it up. It is a deliberately simple, transparent rule: not a precise measure of vocabulary size, but an honest signal of the words that are becoming automatic through reading. Looked-up words go into spaced-repetition review, so the ones that are not sticking get extra attention.
If you want to try it, the first book in each language is free. For example, Russian learners can start with The Captain's Daughter, and Italian learners with Pinocchio. Language overview pages such as learn Russian and learn Italian list what is available.
The takeaway
Reading native novels comfortably requires a large vocabulary, probably in the thousands of word families, and the research suggests you need to know around 98 percent of the words in a text for comfortable unassisted reading. That sounds discouraging until you realise you do not have to wait for it. By reading texts that are matched to your current vocabulary, you can read real stories from the start, and let that reading build the vocabulary you need for everything else.
FAQ
How many words do I need to know to read a novel?
For English, research by Paul Nation estimated roughly 8,000 to 9,000 word families for 98 percent coverage of a wide range of written texts, including novels. Other languages differ, so treat this as an order of magnitude rather than a target for every language.
What does 98 percent coverage mean?
It means you know 98 out of every 100 running words in a text, or about one unknown word in fifty. Research by Hu and Nation (2000) suggested that this level allows most learners to understand a fiction text adequately without help.
Can I read with 90 percent coverage?
You can, but it is hard work. At 90 percent, roughly one word in ten is unknown, which usually means several unknown words per sentence and frequent loss of meaning. It is better suited to intensive study with a dictionary than to fluent reading.
Is it better to learn words from lists or from reading?
Both have a role. Lists and flashcards can help you learn the most frequent words quickly. Beyond that core, most vocabulary is learned more effectively through repeated encounters in context, which is why large amounts of reading matter.
Do these numbers apply to Chinese or Japanese?
Not directly. The best-known figures come from research on English, and languages that use characters, or that build words very differently, are hard to compare word for word. The general principle, that you need very high coverage for comfortable reading, still applies.