📈 Word Frequency & Comprehension Reference
A quick-reference guide showing commonly cited relationships between vocabulary size (most frequent words known) and everyday text comprehension — useful for prioritizing which words to learn first.
What "85% coverage" actually feels like
Coverage is the share of running words on the page you recognise, not the share of the meaning you get. Turn the table’s percentages into unknown words on a 250-word page:
1,000 words known → ~85% → ~38 unknown words per page
2,000 → ~90% → ~25
5,000 → ~95% → ~13
10,000 → ~98% → ~5
That is why the curve flattens: the first 1,000 words buy 85 percentage points, the next 9,000 buy 13. Reading research generally puts comfortable unassisted reading near 98% coverage – about one unknown word in fifty – and treats 95% as the level at which a dictionary is still needed every few lines.
Reading the table honestly
- These are corpus averages, not your text. Conversation and children’s books reach a given coverage on far fewer words than a newspaper editorial or a chemistry paper.
- "Word" usually means word family. Counts of this kind group run, runs, running, ran as one item, so a 5,000-family vocabulary is many more individual forms.
- The high end is language-dependent. A morphologically rich language spreads the same text across more distinct forms, which flattens the early gains English learners enjoy.
Frequently asked questions
How many words do I need to read a novel in a foreign language?
For comfortable reading without constant lookups, figures in the region of 8,000 to 10,000 word families are usually quoted, which is where the table puts 97–98% coverage. Graded readers work far below that by restricting themselves to the top few thousand.
Which words should I learn first?
The most frequent ones, in order. The top 100 words alone carry about half of everyday text, mostly function words such as pronouns, prepositions and the common verbs, and no amount of specialist vocabulary substitutes for them.
Is 90% comprehension good enough to follow a conversation?
Less often than it sounds. One unknown word in ten is roughly two or three per spoken sentence, and unlike a page, speech does not wait while you work them out.