Key takeaways
- Word frequency is extremely uneven: a small set of very common words accounts for a large share of everyday speech, and the returns fall away sharply after that.
- Which words you learn matters far more than how many — a thousand high-frequency words outperform three thousand chosen at random.
- Passive recognition and active production are different sizes; you understand far more words than you can produce, in every language including your first.
- A word is not learned when you can translate it, but when you can retrieve it under time pressure in a real sentence.
- Word-count targets are useful as a rough milestone and misleading as a measure of ability.
The question has no single answer, and the reason is more useful than a number would be: word frequency is extremely uneven. A small set of very common words accounts for a large share of ordinary speech, and each additional thousand words after that buys progressively less.
That shape — steep at first, flattening quickly — is what makes word-count targets simultaneously encouraging and misleading.
The coverage idea
Vocabulary research often frames this as text coverage: what proportion of the words in a typical text or conversation you would recognise, given a vocabulary of a certain size.
The pattern that emerges consistently across studies of many languages is:
- The most frequent few hundred words cover a large proportion of everyday spoken language.
- The next few thousand add meaningfully but far less per word.
- Beyond that, each additional thousand words covers a progressively smaller share.
We are deliberately not quoting specific percentages here. The exact figures vary by language, by corpus and by how a "word" is counted, and picking one number from one study and presenting it as a general fact is the kind of precision that misleads. The shape of the curve is robust; the exact numbers are not.
Why which words matters more than how many
Because of that curve, a thousand high-frequency words outperform three thousand chosen at random, comfortably.
This has a direct practical consequence: study a frequency-ordered list, or study words grouped by the situations you will actually be in — which approximates frequency for your purposes. Do not study words because they appeared in a text you happened to read.
It also explains why vocabulary from a themed lesson feels more useful than vocabulary from a novel. A lesson on ordering food teaches words you will use; a novel teaches words the author needed.
Active and passive vocabulary are different sizes
You understand far more words than you can produce. This is true in your first language too — you recognise plenty of words you would never use.
| Passive (recognition) | Active (production) | |
|---|---|---|
| How it grows | Reading and listening | Speaking and writing |
| Typical size | Larger | Smaller |
| What it lets you do | Understand | Be understood |
The gap matters because conversation requires active vocabulary, and most study methods build passive vocabulary. Flashcards, reading and listening all grow recognition. Only production practice grows production.
The fix is not more input. It is retrieval: producing the word from memory rather than recognising it. Retrieval practice is one of the better-supported learning techniques in the Dunlosky review of effective learning techniques, and it is what makes a word available when you need it.
What "knowing a word" actually means
There are at least four stages, and conflating them is where word counts become fiction:
- Recognise it when you see it written.
- Recognise it when you hear it in speech — much harder, because speech is fast and unsegmented.
- Produce it when you have time to think.
- Produce it under time pressure, in a sentence you are building.
Only stage four is useful in conversation. A learner counting stage-one words and reporting a vocabulary size is not lying, but the number does not describe what they can do.
How to build vocabulary that survives
Learn in context, not in isolation. A word learned inside a phrase comes with its grammar, its collocations and a retrieval cue attached.
Learn the word with everything attached to it. Gender and article in Italian; tone mark in Vietnamese or Thai. Anything that attaches to every word should be learned with the word, or it has to be learned twice.
Space the reviews. Distributed practice substantially outperforms massed practice for verbal recall — the effect synthesised across a large body of studies by Cepeda and colleagues. Spaced repetition for language learning covers what that means in practice, including what our apps do and do not implement.
Practise retrieval, not recognition. Cover the answer. Produce it before checking.
Accept forgetting. Words drop out. Reviewing something you have forgotten is not evidence of failure, it is the mechanism working.
A better target than a word count
CEFR levels describe what you can do rather than what you know, which sidesteps the whole problem — CEFR levels explained covers them. So do concrete tasks: order a meal and handle the follow-up, understand a phone call about an appointment, read a news article without a dictionary.
Those are testable. A vocabulary count is a proxy, and a loose one.
Where our apps fit
Our courses group vocabulary by topic and situation rather than presenting alphabetical or random lists, which is the practical approximation of frequency ordering described above. Learn Italian for Beginners, Learn Vietnamese Language Fast and Regal - Learn Thai Language all take that approach, with review modes built around bringing back what you got wrong rather than what you have already mastered.
What they do not do is force production in conversation, which is the part that converts recognition into use. That still requires a person.
A limit on this article: it deliberately avoids quoting specific coverage percentages, because those vary by language and corpus and we are not in a position to verify a general figure. The rest of our courses are under language learning apps.
Good to know
Frequently asked questions
How many words do you need for a basic conversation?
Does the choice of words matter more than the number?
What is the difference between active and passive vocabulary?
When is a word actually learned?
Is a word-count target useful at all?
Sources
- Common European Framework of Reference for Language skills (opens in a new tab)Europass, European Commission — accessed
- Distributed practice in verbal recall tasks: A review and quantitative synthesis (Cepeda et al., Psychol Bull 2006;132(3):354–80) (opens in a new tab)PubMed, U.S. National Library of Medicine — accessed
- Improving Students' Learning With Effective Learning Techniques (Dunlosky et al., Psychol Sci Public Interest 2013;14(1):4–58) (opens in a new tab)PubMed, U.S. National Library of Medicine — accessed
