How many words do you need to know to read a book in English?

The table is arithmetic about text tokens, not a comprehension test. A token is an occurrence in the passage: repeated occurrences of one unfamiliar word count repeatedly. Word families are a different unit used to describe vocabulary size. Confusing the units makes an apparently precise requirement misleading. Nation, 2006 Brysbaert et al., 2016

Where does the 8,000–9,000 estimate come from?

Nation used frequency-based word-family lists to examine coverage of written and spoken English. The conclusion is conditional: if about 98% coverage is needed for unassisted comprehension, the examined written materials require approximately that vocabulary range. Proper names and the treatment of related forms are part of the analysis, not details to omit from a universal learning target. Nation, 2006

Other vocabulary-counting research shows that lemmas, families and written forms can give very different totals. A result of 8,000 families is not directly comparable with a test reporting 8,000 lemmas, and it is not a count of all words in English. Brysbaert et al., 2016 Nagy et al., 1984

Is 98% a magic comprehension threshold?

Schmitt, Jiang and Grabe tested 661 participants from eight countries on two texts. They found a relatively linear relationship between vocabulary coverage and comprehension, without a sharp point at which understanding suddenly became complete. Their results support 98% as a useful target for academic-text coverage rather than an all-or-nothing gate. SCHMITT et al., 2011

Why books differ

A familiar story, specialist textbook and historical novel contain different vocabulary and assume different background knowledge. The coverage estimate does not remove grammar, inference or subject knowledge from comprehension. Reading-rate research likewise shows why moving through words and understanding them are separate outcomes. Nation, 2006 Brysbaert, 2019

Receptive knowledge is the relevant starting point for recognising text, but a learner can know a word partially. Studies distinguishing recognition, recall and productive knowledge show why ticking a word as familiar does not guarantee complete access to all its meanings in a passage. Laufer et al., 2004 Webb, 2008

Can you begin reading with fewer words?

Yes: the estimate concerns unassisted coverage of the examined materials, not permission to read. Easier material, support and different reading goals change the situation. A learner using a dictionary is doing a different task from the unassisted condition. The number should help choose material rather than discourage reading until a test reaches a particular total. Nation, 2006

Repeated encounters can develop different aspects of word knowledge, but repetition research does not yield a universal number of exposures that makes every word mastered. Track comprehension and the kind of help needed, rather than assuming that a fixed number of newly recognised words guarantees fluent reading. Webb, 2007

What about age, sex and other languages?

This is a text-coverage analysis in English, not an age- or sex-based ability norm. Other languages count forms and families differently, and different genres require different vocabulary. Applying the number across languages without checking the units would overstate the evidence. Nation, 2006 Brysbaert et al., 2016

The useful practical question is how much of the book you understand at your chosen level of assistance. A recognition score can give context, but a short comprehension check on the actual material is closer to that goal than a universal vocabulary cutoff. SCHMITT et al., 2011 Laufer et al., 2004