How many words do you need to speak a language?

The numbers, honestly

Word frequency in a language is brutally uneven. A few hundred words do most of the work in any conversation, and the tail of rare words is very long. Researchers who count these things, most famously Paul Nation, arrive at broadly similar figures across languages:

• The most frequent thousand or so word families cover roughly three quarters of the words in everyday speech.
• Two to three thousand cover most of a conversation and a simple text, enough to follow and to be understood.
• Around five thousand is where reading an ordinary newspaper or novel stops being a struggle.
• Educated native speakers know somewhere between fifteen and twenty thousand word families, and keep adding.

So the encouraging reading is true: a first real conversation is a matter of a few hundred words, well chosen. And the discouraging reading is also true: "learn a language in a thousand words" is enough to survive a trip and not enough to have a life in it.

Why the count lies

The figures above count word families, and a word family is a lemma with everything derived from it. In a language with cases and aspect, a single family is a lot of forms. The Serbian word for city is one entry on the frequency list, and fourteen forms in your mouth.

A learner who has memorised a thousand dictionary forms does not have seventy-five percent coverage. They can recognise seventy-five percent of a text, with effort, and produce very little of it, because the form they memorised is the one that appears in one grammatical role out of seven. The same thousand words, with their forms produced automatically, is a different achievement entirely, and there is no shortcut between the two except producing the forms.

The other thing the count hides is which words. A thousand words picked from a travel phrasebook and a thousand from the actual frequency list overlap less than you would think. The frequent words are mostly small and dull: pronouns, prepositions, the verbs to be, to have, to go, to want, the little particles that hold a sentence together. Nobody makes a flashcard for them, and every sentence needs them.

What a course has to do with this

Two things follow. The words have to be introduced in an order that respects frequency, so the ones every sentence needs come first, and so that at any point what you know is a usable language and not a collection of nouns. And every word has to be learned with its forms, in sentences, because the coverage figures only become true when the forms are automatic.

The course does both. Words arrive lesson by lesson, each lesson bringing the words its rule needs, and every word comes with a picture, a voice, the translation and every form. Every form is its own item in the repetition schedule, with its own level, so "knowing a word" means having its forms above a level, tested in sentences you wrote, not having seen it on a card.

The guarantee the count makes possible

Because the course knows exactly which words it has taught you, it can promise something a phrasebook cannot: you will never meet a word you have not learned. The texts, the dialogues and the sentences of a lesson are built only from the words the course has already given you. No guessing from context, no dictionary in the other hand, no sentence you can half-read. When you open a text, you can read it, and the only question is whether you can read it fast.

That is what a word count should mean, and it is the only version of the count worth tracking. Not how many words you have looked at, but how many you could use right now, in every form they take, without stopping to think. You can see that number, and the level behind every word in it, at any time.