Word lists › Method
Frequency-based language learning: how Zipf works
Most courses teach vocabulary by topic: colours, food, the airport. Frequency-based learning teaches it in the order the language actually uses it, so the first thing you learn is the thing you will meet most. Zipf is built entirely around that order.
Why order matters
In every language a small set of words does most of the work. Zipf's law describes the shape: the most common word is used about twice as often as the second, three times as often as the third, and the curve keeps falling. So the first few hundred words you learn buy far more of the language than any later few hundred. Study them first and every hour is spent where it counts.
What the first 500 words buy you
| Language | Share of everyday text covered by the 500 most common words | |
|---|---|---|
| French | 67% | See the list → |
| Spanish | 67% | See the list → |
| Portuguese | 65% | See the list → |
| Japanese | 64% | See the list → |
| Italian | 63% | See the list → |
| German | 60% | See the list → |
| Greek | 58% | See the list → |
| Russian | 55% | See the list → |
Each word is counted together with its inflected forms (είναι and ήταν count toward είμαι). Source: the wordfreq corpus of web text, subtitles, news, books and Wikipedia.
Coverage is not comprehension
These figures measure the share of words you would recognise on a page. They do not measure how much you would understand, because the most frequent words are mostly the small ones: articles, prepositions, pronouns, the verb to be. Comfortable reading typically needs 95–98% coverage. So a first-500 figure of 60% means the scaffolding of the language is in place and the content words still have to come. Zipf shows both numbers honestly rather than promising fluency at 500 words.
How Zipf uses it
- Ten levels of 500 words, in frequency order. A level opens once you have met most of the one before it.
- Spaced repetition decides when a word comes back; mastery needs several correct answers across different inflected forms, not one lucky recall.
- Five kinds of practice card: meaning, recall, inflected form, typing and listening, chosen by you.
- Every inflection table with its article (των χρόνων, not a bare χρόνων), recorded letter sounds, and an AI tutor for the sentences in between.
Questions
What is frequency-based language learning?
Learning vocabulary in the order words actually occur in the language, most common first, rather than by textbook topic. Because a small number of words account for most of everyday speech and writing, the first few hundred words repay far more study time than any later few hundred.
How much of a language do the first 500 words cover?
It depends on the language. Across the eight languages Zipf teaches, the 500 most common words account for between 55% (Russian) and 67% (French) of the words in everyday text, counting each word together with its inflected forms.
Does 60% coverage mean I understand 60% of a conversation?
No. Coverage is the share of words you would recognise on a page, and the most frequent words are mostly function words (the, of, and, to be). Comfortable reading typically needs 95–98% coverage. A 60% figure means the scaffolding of the language is in place and the content words still have to come.
Why is it called Zipf?
After Zipf's law, the observation that in any language the most common word occurs roughly twice as often as the second, three times as often as the third, and so on. The consequence is that a short list of words covers most of what is said, which is the premise of frequency-based learning.
Where do the frequencies come from?
From the wordfreq corpus, which combines web text, film and TV subtitles, news, books and Wikipedia. Each word's share counts the word together with its inflected forms. The figures are estimates of everyday running text, not of any one genre.
Start with the 500 words that matter → Free for 7 days, then $4 / €4 a month.