How Many Words Do You Need to Know to Read in a Foreign Language?
Here is a number that tends to stop language learners cold: in a typical Spanish novel, a learner with 3,000 word families under their belt already knows roughly 95% of the running words on the page. One unknown word in every twenty. That sounds almost good enough — until you sit down to actually read and discover that one stranger per sentence is enough to derail meaning. The research on exactly how many words you need to read a foreign language comfortably is four decades old, surprisingly consistent, and almost completely unknown to the average self-taught learner. This article lays it out plainly, connects it to the text types you actually want to read, and explains what the numbers mean for you right now.
The Two Thresholds You Need to Know
Vocabulary researchers talk about lexical coverage — the percentage of running words in a text that a reader already knows. Two coverage percentages appear again and again in the literature, and understanding what each one actually delivers is the foundation of everything else.
95% coverage: the floor, not the ceiling. Laufer's (1989) study indicated that L2 learners may have adequate reading comprehension of a general academic text at 95% coverage. That study was the first of its kind. At 95%, you encounter one unknown word in every twenty — roughly one per sentence in normal prose. Successful reading generally requires 95–98% word coverage, with 95% allowing adequate comprehension with support (e.g., dictionaries, glosses) and 98% supporting unassisted understanding. The "support" caveat is important. At 95%, most learners are not reading independently; they are reading with a dictionary close at hand.
98% coverage: the target for independent reading. Hu and Nation (2000) found that 98% coverage was needed for adequate unassisted reading comprehension of a relatively easy L2 fiction text. They reported that 98% coverage may be necessary if comprehension test scores of 60% or higher are needed. At 98%, only one word in every fifty is unknown — about one unfamiliar word per page, not per sentence. That is the difference between reading and decoding.
This supports Laufer and Ravenhorst-Kalovski's (2010) suggestion that two lexical coverage thresholds based on the degree of comprehension are used: 95% for minimal and 98% for optimal comprehension. The two-threshold model has since become the field's working consensus. In a large-scale study, 661 participants from 8 countries completed a vocabulary measure, read texts, and then completed a reading comprehension test. The results revealed a relatively linear relationship between the percentage of vocabulary known and the degree of reading comprehension. There was no indication of a vocabulary "threshold" where comprehension increased dramatically, but results suggest that the 98% estimate is a more reasonable coverage target for readers of academic texts.
The important takeaway from Schmitt et al. (2011) is that the relationship is linear, not cliff-shaped. Every extra percentage point of coverage adds comprehension. There is no magic level where suddenly everything clicks — but 98% is where most learners can finally sustain reading without constant interruption.
What These Percentages Feel Like on the Page
Abstract percentages become concrete fast. Here is a short authentic-register Spanish passage, profiled at the 3,000-word-family band. Beeblio measures what fraction of its tokens a learner at that band already knows:
Hoy es un buen día. Voy a la ciudad con mi amigo y después comemos en casa. Por la noche leo un libro.
The passage above comes back at 100% known for a learner at the 3,000-word band — every single token falls inside the top-3,000 Spanish word families. That is exactly the sweet spot Beeblio targets: the text is fully within reach, and any new words introduced in the next session will land just beyond the known boundary, not in the void. It also illustrates why coverage percentage matters more than raw difficulty: a learner at 2,000 words reading the same passage would hit several unknowns, which is fine for push practice but exhausting as a sustained diet.
Word Families to Coverage: The Translation Table
Knowing that you need 98% coverage is useful. Knowing how many word families translate to that coverage in the text type you actually want to read is actionable. An analysis of various kinds of texts using 1,000-word family lists shows that between 3,000 to 4,000 word families are needed to get 95% text coverage, and between 6,000 and 9,000 word families are needed to gain 98% coverage. For novels and newspapers, 4,000 word families reach 95% coverage, while 9,000 word families plus proper nouns provide 98% coverage.
| Text type | 95% coverage (word families) | 98% coverage (word families) |
|---|---|---|
| Conversation / everyday speech | ~3,000 | ~6,000–7,000 |
| TV, songs, podcasts | ~2,000–3,000 | ~5,000–7,000 |
| Graded readers (simplified fiction) | ~2,000 | ~3,000–4,000 |
| Newspapers | ~4,000 | ~8,000–9,000 |
| Novels (unsimplified) | ~4,000 | ~9,000 |
| Academic spoken English | ~4,000 | ~8,000 |
Sources: Nation (2006); Coxhead (2013); Webb & Nation (2021); Dang et al. (2022).
Several things are worth noting in this table. First, spoken language is cheaper than written language. Nation (2006) found that including proper nouns, 3,000 word families accounted for 95% coverage of unscripted spoken English. Van-Zeeland and Schmitt (2012) suggest that to reach 95% lexical coverage of spoken text, learners would need from 2,000 to 3,000 word families. This is why a learner who "understands TV fine" can still be lost in a newspaper. Second, the jump from 95% to 98% is enormous. Studies on the lexical demands of various text genres pointed out that learners who have vocabulary knowledge at the 95% threshold would have to double or even triple their lexical resources if they wished to gain 98% coverage of the same text genres. That doubling-or-tripling is why intermediate plateaus feel so permanent — learners are trying to close a gap that is much larger than it looks.
The Gap Nobody Talks About: Graded Readers to Real Novels
The table above contains a buried crisis. The highest graded readers typically cap out around 3,000 word families. A real, unsimplified novel needs 9,000 for 98% coverage. A gap exists between the end of graded readers at around 3,000 word families and the vocabulary size needed to read unsimplified texts at around 8,000 word families. This is a 5,000-to-6,000-family chasm. It is the single most common reason intermediate learners give up on reading in their target language: they finish the graded reader series feeling confident, crack open a real novel, and are immediately overwhelmed.
The research-backed solution is not to white-knuckle through unsimplified texts with a dictionary. It is to stay in highly comprehensible texts — text where you know 95% or more of the tokens — long enough to build vocabulary through incidental acquisition, which is learning words from context as a by-product of reading for meaning. Nation (2006) reports that 98% coverage is most appropriate for incidental vocabulary learning. Put differently: to learn new words from reading, you already need to know almost all the surrounding words. The scaffolding must be in place before the new word can stick.
Krashen, Nation, and the i+1 Alignment
Stephen Krashen's comprehensible input hypothesis, first published in the 1970s and 80s, proposed that language is acquired when learners are exposed to input that is slightly beyond their current level. Krashen's "comprehensible input" hypothesis suggests that learners acquire language by taking in and understanding language that is "just beyond" their current level of competence. He defines comprehensible input as 'i+1', where 'i' is the current level of proficiency and '+1' is the level of proficiency just beyond the learner's current level. Krashen's framework was deliberately qualitative — he never specified a precise coverage percentage. Nation's corpus work did the quantification. The concept has been quantified in vocabulary acquisition research; Nation reviews various studies which indicate that about 98% of the words in running text should be previously known in order for extensive reading to be effective.
The two frameworks converge neatly: Krashen's "+1" is exactly the handful of new words that appear in a text where you know 98% of everything else. The new words are contextualised, retrievable, and memorable precisely because the surrounding 98% is holding the meaning together. Below 95%, the context itself starts to collapse, and the "+1" becomes "+20."
Extensive Reading: Why Volume at the Right Level Beats Difficulty
The practical implication of all the above is that extensive reading — reading large quantities of level-appropriate text for pleasure and general comprehension — is one of the most efficient vocabulary-building strategies available. Long-term language development requires exposure to understandable input, autonomy, and enjoyment, all of which are fostered by extensive reading. A plethora of studies has revealed that extensive reading enables students to reap numerous linguistic benefits, including improved reading fluency, vocabulary acquisition, grammar, and writing ability.
The key word is volume. A book per week is recommended to achieve the benefits of extensive reading, for vocabulary acquisition and for establishing a reading habit. But volume only works when the material is genuinely comprehensible. Difficult texts you slog through at 85% coverage build frustration faster than vocabulary. The most efficient path is: stay well inside your coverage band, read a lot, and let the high-frequency words just above your known boundary accumulate through repeated encounters in context.
A Note on "Word Families" vs. Raw Word Counts
Throughout this article the unit has been word families, not individual word forms. A word family groups a headword together with its inflected and derived forms: read, reads, reading, reader, readable, unreadable all belong to one family. This matters because learners often feel they know far fewer words than they actually do, because they count forms rather than families. Reliable studies suggest that educated native speakers of English know around 20,000 word families. The coverage targets above — 3,000 for 95%, 8,000–9,000 for 98% — are stated in families. If you have been tracking individual forms or lemmas, your effective vocabulary size is likely higher than you think.
For Spanish specifically, the morphological richness of the language means one family covers more surface forms than in English (a single Spanish verb can have over 50 conjugations). This is both a challenge and an advantage: once you know a family, you recognise a large number of tokens automatically. The coverage figures from English-language frequency research generalize in principle to Spanish, though Spanish's richer morphology shifts exact numbers.
FAQ
Is 95% coverage really enough to enjoy reading?
Barely, and only with support. Minimal text comprehension requires at least 95% known-word coverage, whereas independent comprehension requires approximately 98%. At 95% you meet an unknown word every twenty words — roughly every sentence. For reading with a dictionary or gloss available, 95% is a reasonable practice level. For pleasurable, unassisted reading, you want to be at or above 98%.
Do these numbers apply to Spanish, French, Japanese — or only English?
The landmark studies (Laufer 1989, Hu & Nation 2000, Schmitt et al. 2011) were conducted primarily in English-language contexts, but the coverage-comprehension relationship is a property of reading, not of any one language. The thresholds of 95% and 98% are widely accepted across applied linguistics as language-general targets. The number of word families needed to hit those thresholds will vary by language depending on morphological complexity and corpus composition, but the coverage percentages themselves hold.
What about children's books or social media posts — are they easier to reach 98% on?
Significantly. Songs, soap opera, sitcom and podcast were relatively less demanding as they only required approximately 2,000–3,000 word families for 95% coverage, and 5,000–7,000 for 98% coverage. Children's literature and simplified graded readers sit in a similar range. This is exactly why beginner and intermediate learners benefit from level-appropriate materials long before they are ready for authentic novels.
How long does it take to go from 3,000 to 9,000 word families?
Nation's advice for course design is that language courses should aim for at least 1,000 word families per year to ensure meaningful progress. At that rate, the 6,000-family gap between a comfortable graded reader level and independent novel-reading is approximately six years of dedicated study. Extensive reading, spaced repetition, and deliberate vocabulary study in combination can compress that timeline significantly — but there is no honest shortcut past the volume of encounters required for a word to become automatic.
Does knowing 98% of words guarantee comprehension?
No, it is a necessary but not sufficient condition. Background knowledge, syntax, discourse structure, and topic familiarity all play roles. Results from large-scale research revealed a relatively linear relationship between the percentage of vocabulary known and the degree of reading comprehension, but there was no indication of a vocabulary "threshold" where comprehension increased dramatically at a particular percentage. Comprehension is a gradient, not a switch. What 98% coverage gives you is a platform stable enough to deploy all those other skills effectively.
Sources
- Hu, M. & Nation, I.S.P. (2000). "Unknown Vocabulary Density and Reading Comprehension." Reading in a Foreign Language, 13(1), 403–430.
- Nation, I.S.P. (2006). "How Large a Vocabulary Is Needed For Reading and Listening?" Canadian Modern Language Review, 63(1), 59–82.
- Schmitt, N., Jiang, X., & Grabe, W. (2011). "The Percentage of Words Known in a Text and Reading Comprehension." The Modern Language Journal, 95(1), 26–43.
- Laufer, B. & Ravenhorst-Kalovski, G.C. (2010). "Lexical Threshold Revisited." Reading in a Foreign Language, 22(1), 15–30.
- Krashen, S.D. (1982). Principles and Practice in Second Language Acquisition. Pergamon Press.
- Day, R. & Bamford, J. (1998). Extensive Reading in the Second Language Classroom. Cambridge University Press.
- Webb, S., Pellicer-Sánchez, A., & Wang, C. (2024). "How Does Lexical Coverage Affect the Processing of L2 Texts?" Applied Linguistics, 45(6).
- Webb, S. & Nation, I.S.P. (2021). "Research Investigating Lexical Coverage and Reading Comprehension." Reading in a Foreign Language, 33(2).
- Dang, T.N.Y. et al. (2022). "Lexical Coverage and Text Type Demands." Frontiers in Psychology, 13.
Every text you read in Beeblio is profiled against your current known-words band, keeping coverage at or above 95% while introducing a controlled number of words just beyond it — the precise zone where, according to Nation, Hu, and Schmitt, incidental acquisition is most likely to occur. As your known-words count climbs, Beeblio's passages shift with it, so you are always reading at the coverage level that research says matters most.