How many words do you need to understand a language?

"How many words do I need?" is one of the most asked questions in language learning, and one of the hardest to answer honestly. Numbers float around online, often with no source. This article sticks to what a published research review actually says, and is careful about what it cannot say.

What is lexical coverage?

Researchers who study vocabulary often work in terms of coverage: the proportion of words in a text or speech that a learner already knows. If you know 98 of every 100 words in a passage, your coverage is 98 per cent and you will meet two unknown words in every hundred. The question then becomes: how large must a vocabulary be to reach a coverage level that lets a learner understand without much strain?

What numbers have researchers proposed?

A 2017 paper in Language Teaching by Norbert Schmitt and colleagues reviewed the key studies and called for replications (Schmitt, Cobb, Horst and Schmitt). In their summary of earlier work, all about English:

• Reading. About 98 per cent coverage is the most commonly cited target for written text, with 95 per cent as a minimum that may need extra help such as a dictionary. An earlier analysis by Nation (2006) estimated roughly 8,000 to 9,000 word families (plus proper nouns) for 98 per cent coverage of written English.
• Listening. Van Zeeland and Schmitt (2012) suggested 95 per cent coverage may be adequate for listening, which the paper places at roughly 2,000 to 3,000 word families. Nation's estimate for 98 per cent coverage of spoken English was about 6,000 to 7,000 word families.

Studies in this area usually count word families rather than every separate word form, which is why their totals look smaller than a count of every different word a learner might meet. We have not defined the term more exactly than the studies do, so treat the figures as rough.

Why should you treat these numbers carefully?

The authors list a long set of caveats, which is the most useful part for a parent:

• The figures come from counting words in texts, not from testing what learners actually understood.
• The original text collections were small, and results were combined in ways the authors question.
• One listening study used short, well-organised stories, allowed repeated listening and did not allow questions, so its result may be a best case.
• How you count word families, and whether you account for set phrases, changes the totals.
• Specialised areas such as science or law may need different vocabularies.
• Coverage figures do not say whether a particular unknown word is critical to meaning.
• Learners who can use resources, such as a dictionary or context, may get by with less.

The paper is about English. We did not find a readable source giving equivalent word counts for Japanese, Mandarin, Spanish or any other language, and these figures should not be assumed to transfer.

What does this mean for a child or adult learner?

A few grounded takeaways:

1. The number depends on the goal. Following a slow, simple story needs far fewer words than reading a newspaper.
2. The most common words matter most. The idea of coverage is that frequent words appear again and again, so learning them first pays off. This is how coverage works in general, though the exact list depends on the language.
3. Thousands of words take time. There is no quick route to several thousand words, so regular practice and revisiting matters. A review of spacing studies found that the most useful gap between sessions widens as the target retention period widens (Cepeda and colleagues). See our piece on spaced repetition.
4. Early levels are described in tasks. In the global scale as Cambridge reproduces it, A1 and A2 are framed as everyday jobs such as introducing yourself or handling simple shopping talk, not as word counts (Cambridge Assessment English). The Council of Europe's own pages were not reachable when we checked, so we lean on that Cambridge reproduction. Our CEFR guide lays out the steps.

What about bilingual children and vocabulary size?

Parents sometimes compare word counts. A review for parents says a bilingual child typically has fewer words in each separate language than a monolingual peer, but pooling the words from both languages, and counting a word and its translation as one, gives a total of about the same size (Byers-Heinlein and Lew-Williams). So a raw word count for one language can understate what a child knows.

How should a family use word counts?

Treat them as a rough map, not a target. Count what your child can use, not only what they can recognise. Notice whether a short text in the language has many words they do not know: if it has a lot, it is probably too hard right now, and if almost none, it may be too easy. Aim for texts where most words are familiar and a few are new.

A calm next step

XLingo builds vocabulary through word cards, short stories, exercises and spaced review, in eleven languages at six levels. It makes no promise of results. The free trial and price are on the pricing page.

The closing thought from the research is a modest one: vocabulary grows by steady exposure to words you meet again and again, in contexts that make sense to you, and no published count replaces that. Set a goal you can describe in plain terms, check it against real texts and keep going.