How many words do you need to know to read a newspaper?

The short answer from the research is 8,000 to 9,000 word families for reading a newspaper without help. That number is real, and it is also the reason a lot of people never open a newspaper in the language they're learning.

So here is the longer answer: where the figure comes from, what it actually feels like at each level, and why the threshold moves once you stop reading unassisted.

Where the number comes from

The standard reference is Laufer and Ravenhorst-Kalovski's 2010 study in Reading in a Foreign Language. They measured learners' vocabulary size, how much of a text that vocabulary covered, and how well they understood it. Two thresholds came out:

Threshold	Vocabulary	Text coverage
Optimal	8,000 word families	98% of the running words
Minimal	4,000 to 5,000 word families	95% of the running words

Their regression is blunter than the table. A 3,000-word vocabulary predicts a reading score of 56%. Four thousand gets you to 63%. Five thousand gets you to 70%. Each extra thousand buys about seven points, and the curve flattens after that.

The 98% figure itself traces back to Hu and Nation (2000), who did something clever: they took a text, replaced a controlled share of the words with invented non-words, and made four versions at 80%, 90%, 95% and 100% coverage. Almost nobody reached adequate comprehension below 95%. Some managed it at 95%, but they were the minority. At 98% most people were fine.

What 98% coverage actually feels like

Percentages are hard to picture. Convert them to gaps:

• 98% known is one unknown word in every fifty. On a 250-word news article, that's five words. You'll guess three of them from context and not notice the other two.
• 95% known is one in twenty. Two or three per paragraph. Readable, but you're working, and the study says most people's comprehension has already dropped below "adequate."
• 90% known is one in ten. Every second line stops you.
• 80% known is one in five. This is not reading, it's decoding.

The jump from 95% to 98% is only three percentage points and roughly 3,000 extra word families. That's the gap between "I finished the article" and "I gave up halfway," which is also the gap that shows up in reading speed.

Word families are bigger than they sound

A word family is a base word plus its inflections and common derivations. Govern, governs, governed, government, governor, governance is one family. So 8,000 families is considerably more than 8,000 entries you'd count in a wordlist, and considerably fewer than the raw "50,000 words" figures you see quoted for educated native speakers.

It also means your family count grows faster than your word count once you know the common prefixes and suffixes. Meeting unmistakably when you already know mistake costs almost nothing.

Languages that don't count words the same way

For Japanese, the question people actually ask is how many kanji. Japan's jōyō list, the cabinet-designated set of everyday characters last revised in November 2010, runs to 2,136. Newspapers largely stay inside it and add furigana for what falls outside.

But 2,136 characters is not a vocabulary. You still need the words those characters build, and a fair amount of newspaper text is kana-only grammar and katakana loanwords that no kanji list covers. The character count is an entry fee, not the ticket. Same story for Chinese: knowing 2,500 characters and knowing 8,000 words are different achievements.

The catch: the number is for reading unassisted

Every figure above measures comprehension with no dictionary, no glosses, nothing. That's the right way to run the study and the wrong way to run your week.

If a word's meaning arrives the moment you want it, the 98% wall stops behaving like a wall. The unknown words are still there, but they cost a second instead of a stop, and the article gets finished. Finishing is what puts those words in front of you a second and third time, which is the only thing that grows the number.

This is the whole design of Mossglen (ours, free): any web page becomes a reading view where tapping a word gives you the meaning and saves it with the sentence it came from, so a 93% article reads more like a 98% one while your actual vocabulary catches up.

What to read at your current size

• Under 3,000 families. Newspapers are the wrong target. Graded readers, easy-news editions, and articles about things you already know. Start here.
• 3,000 to 5,000. Real articles, with lookups. Pick topics you follow in your own language, because background knowledge does some of the work vocabulary would otherwise have to.
• 5,000 to 8,000. Newspapers, unassisted, in familiar sections. Sports and business before opinion columns, which lean on idiom.
• Above 8,000. You're not asking this question anymore.

The short version

Question	Answer
Newspaper, no help	8,000 to 9,000 word families
Bare minimum	4,000 to 5,000 families, 95% coverage
What 98% feels like	One unknown word every fifty
Japanese	2,136 jōyō kanji is the entry fee, not the vocabulary
Fastest route	Read below the threshold, a lot, with lookups that don't stop you

The threshold is real but it isn't a gate you wait at. You get to 8,000 by reading, and reading is how you get there, not what you do once you arrive.