Vocabulary size
How many words do you need for academic reading in English?
About as many as for any unsimplified English text: roughly 4,000–5,000 word families to know 95% of the words and 8,000–9,000 to know 98%, by the research estimates for written English. Academic vocabulary is a small layer inside that total: the 570 word families of the Academic Word List make up about 10% of the words in academic texts, against 1.4% in fiction (Coxhead, 2000).
Key points
- The Academic Word List (AWL) has 570 word families, 3,110 word forms in all, chosen from 3.5 million words of academic text in 414 texts and 28 subject areas. It covers 10.0% of the words in that corpus and 1.4% of a fiction corpus of about the same size (Coxhead, 2000).
- West's General Service List (GSL) and the AWL together cover about 86% of the same academic corpus, so about one word in seven is outside both lists, short of the 95% at which reading becomes possible with effort (Coxhead, 2000; Laufer & Ravenhorst-Kalovski, 2010).
- Outside the corpus it was built from, coverage varies: 8.5% of a second academic corpus, 10.6% of a 3.3-million-word corpus in a later study, and only 7.2% and 6.9% of the much larger academic sections of COCA and the BNC (Coxhead, 2000; Hyland & Tse, 2007; Gardner & Davies, 2014).
- Learn a general list first. The AWL was built to sit on top of the GSL, which alone covers about 76% of academic text. If you start from the New General Service List (2,809 words), its academic companion is the New Academic Word List (957 words) (NGSL; NAWL).
- Reuse terms differ: the NGSL and NAWL are under CC BY-SA 4.0; the AWL is copyright Victoria University of Wellington, with non-commercial reproduction allowed and commercial use needing written permission (checked 31 October 2026).
How many words do you need for academic reading?
Roughly the same number as for other unsimplified written English. The research thresholds are about 4,000–5,000 word families to know 95% of the running words of a written text, enough to read with effort, and about 8,000–9,000 to know 98%, the level at which reading becomes comfortable (Laufer & Ravenhorst-Kalovski, 2010; Nation, 2006). Nation's estimate was made on novels and newspapers, not on academic texts.
Inside that total, academic texts lean on a set of words that are uncommon in stories and conversation but frequent across subjects: research, process, data, analyse, theory, significant. The Academic Word List collects 570 such word families. They are about one word in ten of an academic text, so they are worth learning early, but they are a layer on top of a general vocabulary, not a replacement for it.
- 570
- word families in the Academic Word List (Coxhead, 2000)
- 10.0%
- of the words in the 3.5-million-word Academic Corpus are AWL words; 1.4% of a fiction corpus of about the same size
- 86%
- of the Academic Corpus is covered by the GSL and the AWL together
- 95% / 98%
- of words known: the thresholds for reading with effort and for comfortable reading (Laufer & Ravenhorst-Kalovski, 2010)
Our guide to how many words you need explains where the 95% and 98% thresholds come from and gives the figures for films and conversation.
What is the Academic Word List?
The Academic Word List is a list of 570 word families that occur frequently and evenly across academic texts in many subjects, leaving out the most frequent 2,000 words of general English. Averil Coxhead compiled it for her MA thesis at Victoria University of Wellington in New Zealand and published it in TESOL Quarterly in 2000 (Coxhead, 2000; Victoria University of Wellington). It replaced the older University Word List, which had 836 word families and covered 9.8% of the same corpus: the AWL reaches slightly more coverage with 266 fewer families.
The corpus. Coxhead's Academic Corpus has about 3,500,000 running words from 414 texts: journal articles, book chapters, textbooks, course workbooks, laboratory manuals and course notes, most of them written between 1993 and 1996. It is split into four faculties of about 875,000 words each, Arts, Commerce, Law and Science, and these into 28 subject areas from education and history to chemistry and constitutional law (Academic Corpus; Coxhead, 2000).
How words were chosen. A word family went into the list only if it was outside West's General Service List and if it met three criteria (AWL selection):
- Range: it occurred in all four faculties and in at least 15 of the 28 subject areas;
- Frequency: it occurred over 100 times in the whole corpus;
- Uniformity: it occurred at least 10 times in each faculty.
The criteria are what make the list general: a word frequent only in law or only in chemistry did not get in.
The sublists. The 570 families are ranked by frequency and split into ten sublists, nine of 60 families and a last one of 30. Coverage falls steeply from the first sublist to the last, so the first sublists repay study most (The sublists):
| Sublist | Word families | Coverage of the corpus |
|---|---|---|
| 1 | 60 | 3.6% |
| 2 | 60 | 1.8% |
| 3 | 60 | 1.2% |
| 4 | 60 | 0.9% |
| 5 | 60 | 0.8% |
| 6 | 60 | 0.6% |
| 7 | 60 | 0.5% |
| 8 | 60 | 0.3% |
| 9 | 60 | 0.2% |
| 10 | 30 | 0.1% |
The first 60 families give more than a third of the list's coverage. All of the example words above, research, process, data, analyse, theory and significant, are Sublist 1 headwords.
How much of a textbook does the Academic Word List cover?
About one word in ten in the kind of text it was built from, which includes textbook chapters, journal articles and course notes, and between about 7% and 11% in other collections of academic writing. Coverage depends on the subject and on the corpus used to measure it:
| Text collection | Size in words | AWL coverage | Source |
|---|---|---|---|
| Coxhead's Academic Corpus, all four faculties | 3.5 million | 10.0% | Coxhead, 2000 |
| Commerce section | about 875,000 | 12.0% | Coxhead, 2000 |
| Law section | about 875,000 | 9.4% | Coxhead, 2000 |
| Arts section | about 875,000 | 9.3% | Coxhead, 2000 |
| Science section | about 875,000 | 9.1% | Coxhead, 2000 |
| Second academic corpus, collected to test the list | about 678,000 | 8.5% | Coxhead, 2000 |
| Academic writing across disciplines and genres | 3.3 million | 10.6% | Hyland & Tse, 2007 |
| Corpus of Contemporary American English (COCA), academic section | about 120 million | 7.2% | Gardner & Davies, 2014 |
| British National Corpus (BNC), academic section | about 33 million | 6.9% | Gardner & Davies, 2014 |
| Fiction, for comparison | about 3.8 million | 1.4% | Coxhead, 2000 |
Two conclusions hold across the studies. First, AWL words are much more common in academic writing than in fiction: roughly seven times more in Coxhead's own comparison. Second, 10% is not a fixed property of the list: in other corpora it ranges from 6.9% to 10.6%, and it is lowest in the two largest corpora, COCA and the BNC, where it is closer to 7%.
There is also a limit that coverage figures do not show. Ken Hyland and Polly Tse (2007) found that, although the AWL covered 10.6% of their corpus, many of its words occur and behave differently across disciplines in range, frequency, collocation and meaning. They questioned whether students need a single core academic vocabulary and recommended that teachers help students build a vocabulary for their own discipline. In practice: knowing an AWL word in general does not guarantee knowing what it means in your field.
What do the GSL and the AWL leave out?
About 14% of the words in academic text. West's General Service List (1953), the roughly 2,000 most useful words of general English on which the AWL was built, covers about 76% of Coxhead's Academic Corpus; the AWL adds 10%, for a total of about 86% (Coxhead, 2000). The remainder is one word in seven: less frequent general words, the technical terms of each subject, and names.
That is well short of the thresholds. At 86% coverage a reader meets an unknown word about every seven words; at 95% it is one in twenty, and at 98% one in fifty (Laufer & Ravenhorst-Kalovski, 2010). So the two lists together are a foundation for academic reading, not the end point: the distance from 86% to 95% has to be closed with a larger general vocabulary and with the specialised words of your subject.
AWL or NGSL: which should you learn first?
A general list first, then an academic one. The AWL was designed on the assumption that the learner already knows the words of the General Service List; it leaves them out, and they cover about 76% of academic text against the AWL's 10% (Coxhead, 2000). Learning the AWL before the general high-frequency words means learning the smaller part of the text first.
The New General Service List (NGSL) is a modern replacement for West's list, published by Charles Browne, Brent Culligan and Joseph Phillips in 2013. Version 1.2, released in April 2023, has 2,809 words, and its authors state that it covers 92% or more of general English (NGSL). The two lists are therefore not a pair: the AWL was built against the old GSL, and some of its most frequent families are already in the NGSL. The Sublist 1 examples above, research, process, data, theory and significant, and analysis from the analyse family, all appear in the NGSL (NGSL word list).
So pick a matching pair:
- GSL, then AWL: the original pair, and the one used in the research cited on this page. Together they cover about 86% of Coxhead's Academic Corpus.
- NGSL, then NAWL: the New Academic Word List was built to follow the NGSL. Version 1.2 has 957 words; its authors report that the NGSL and NAWL together cover about 92% of a 288-million-word academic corpus, 5 percentage points more than the GSL and AWL; their site also cites a study by Hendry and Sheepy (2018) that found 90% coverage of a 5-million-word corpus of over 18,000 journal abstracts (NAWL). We did not check that study.
- NGSL, then AWL: workable, but expect overlap; skip the AWL words you already learned from the NGSL.
If you do not know where you stand, a vocabulary size test is the quickest way to find out whether the general high-frequency words are already secure; our comparison of vocabulary tests suggests which one to take.
How do the AWL, NGSL, NAWL and AVL compare?
Five lists come up in almost every discussion of academic vocabulary. They differ in what they count, what they were built from, and what they assume you already know.
| List | Authors, year | Size | Counted as | Built from | Assumes you know | Stated coverage |
|---|---|---|---|---|---|---|
| General Service List (GSL) | West, 1953 | about 2,000 | Headwords | Not checked by us | Nothing | About 76% of Coxhead's Academic Corpus |
| Academic Word List (AWL) | Coxhead, 2000 | 570 | Word families | 3.5 million words, 414 academic texts, 28 subject areas | The GSL | 10.0% of the Academic Corpus |
| New General Service List (NGSL) 1.2 | Browne, Culligan & Phillips, 2013; version 1.2, 2023 | 2,809 | Words | Not stated on the list's page | Nothing | 92% or more of general English |
| New Academic Word List (NAWL) 1.2 | Browne, Culligan & Phillips, 2013 | 957 | Words | 288-million-word academic corpus | The NGSL | With the NGSL, about 92% of that corpus |
| Academic Vocabulary List (AVL) | Gardner & Davies, 2014 | 3,000 lemmas; 2,000 families | Lemmas, with a word-family version | 120-million-word academic section of COCA, 9 disciplines | No base list | About 14% of COCA and BNC academic text |
Sources: West, 1953; Coxhead, 2000; NGSL; NAWL; Gardner & Davies, 2014; academicwords.info.
Why the Academic Vocabulary List is different. Dee Gardner and Mark Davies (2014) built the AVL without a general list underneath. They argued that the GSL is outdated and that word families lump together forms with different meanings and parts of speech, so they counted lemmas instead. A lemma qualified if it was at least 50% more frequent in academic text than expected, reached at least 20% of its expected frequency in 7 of 9 disciplines, was evenly spread (dispersion of at least 0.80), and was not more than three times as frequent as expected in any one discipline. Comparing like with like, the top 570 AVL word families covered 13.8% of COCA's academic section and 13.7% of the BNC's, against 7.2% and 6.9% for the 570 AWL families. The AVL was selected from COCA, so the BNC figure is the fairer test. And because it has no base list, the AVL can include frequent words that the AWL leaves out as GSL words, which is one likely reason it covers more.
Where does the Academic Word List fit in your vocabulary size?
Mostly inside the first 4,000 words, not beyond the 9,000. Gardner and Davies (2014) checked the AWL against modern frequency counts: 451 of its 570 word families (79%) are represented among the 4,000 most frequent lemmas of COCA, and 236 (41%) among the top 2,000. The AWL was defined as lying beyond the first 2,000 words by the standards of a 1953 list; by current frequency, many of its words are ordinary high-frequency English.
That changes what the list is for. A learner working towards the 4,000–5,000 word families needed for 95% coverage of written text (Laufer & Ravenhorst-Kalovski, 2010) will meet most AWL words along the way. The AWL's value is as a shortcut: it tells someone heading for university which words in that band to put first. It does not add a separate block of 570 families on top of the 8,000–9,000 needed for comfortable reading (Nation, 2006).
The units matter when you compare numbers. The AWL and Nation's estimates count word families; the AVL and the NGSL count lemmas or words; frequency lists such as our 500 most common English words in books count word forms. Our page on how many words the average person knows shows how far apart the same vocabulary looks in each unit, and the hub guide explains what counts as a word family.
How should you use academic word lists?
- Secure the general high-frequency words first. In Coxhead's corpus, the GSL covers about 76% of academic text and the AWL 10%. Gaps in the first 2,000–3,000 words cost more than gaps in an academic list.
- Use one matching pair. GSL with AWL, or NGSL with NAWL. Mixing the NGSL with the AWL repeats words.
- Start with Sublist 1. Its 60 families give more than a third of the AWL's coverage; the last sublist gives 0.1% (The sublists).
- Learn each word in your own field's texts. AWL words can mean different things and keep different company in different disciplines (Hyland & Tse, 2007), so a flashcard definition is a starting point, not the whole job.
- Close the rest of the gap by reading. From 86% to 95% is mostly less frequent general words and the terms of your subject. Reading widely in your discipline is how most of them are met often enough; our guide to what the research means for learners explains why lists stop paying off after the first few thousand words.
Exam preparation, such as for IELTS or TOEFL, uses the same academic vocabulary but is outside the scope of this page.
Can you download and reuse these lists?
We do not reprint the lists here. Each is available from its rights holder, on different terms; we checked them on 31 October 2026:
- Academic Word List. The headwords and sublists are free to view on Victoria University of Wellington's AWL pages, which carry a copyright notice of the university and no separate licence (AWL pages). The university's copyright page allows reproduction for non-commercial purposes if the material is unaltered, not used in a misleading way and credited; commercial reproduction needs the university's prior written permission (Copyright).
- New General Service List and New Academic Word List. Free downloads in several formats, released under the Creative Commons Attribution-ShareAlike 4.0 International licence, credited to Browne, Culligan and Phillips. The site states that the lists may be used commercially (NGSL; NAWL; word lists page).
- Academic Vocabulary List. The authors' site describes the lists as completely free and offers them as Excel files; it states no licence or terms of reuse (academicwords.info).
- General Service List. Published as a book by Michael West in 1953 (West, 1953); we did not check its reuse terms.
References
- Coxhead, A. (2000). A new academic word list. TESOL Quarterly, 34(2), 213–238. doi:10.2307/3587951; PDF at Victoria University of Wellington
- Gardner, D., & Davies, M. (2014). A new academic vocabulary list. Applied Linguistics, 35(3), 305–327. doi:10.1093/applin/amt015; author's PDF
- Hyland, K., & Tse, P. (2007). Is there an "academic vocabulary"? TESOL Quarterly, 41(2), 235–253. doi:10.1002/j.1545-7249.2007.tb00058.x
- Laufer, B., & Ravenhorst-Kalovski, G. C. (2010). Lexical threshold revisited: Lexical text coverage, learners' vocabulary size and reading comprehension. Reading in a Foreign Language, 22(1), 15–30. Full text (ERIC)
- Nation, I. S. P. (2006). How large a vocabulary is needed for reading and listening? Canadian Modern Language Review, 63(1), 59–82. doi:10.3138/cmlr.63.1.59
- West, M. (1953). A General Service List of English Words. London: Longmans, Green.
- Victoria University of Wellington, School of Linguistics and Applied Language Studies. The Academic Word List and AWL information. wgtn.ac.nz, checked 31 October 2026.
- Victoria University of Wellington. The Academic Corpus. wgtn.ac.nz, checked 31 October 2026.
- Victoria University of Wellington. Academic Word List selection. wgtn.ac.nz, checked 31 October 2026.
- Victoria University of Wellington. The sublists. wgtn.ac.nz, checked 31 October 2026.
- Victoria University of Wellington. Copyright. wgtn.ac.nz, checked 31 October 2026.
- Browne, C., Culligan, B., & Phillips, J. The New General Service List (NGSL 1.2). newgeneralservicelist.com, checked 31 October 2026.
- Browne, C., Culligan, B., & Phillips, J. The New Academic Word List (NAWL 1.2). newgeneralservicelist.com, checked 31 October 2026.
- The word lists. newgeneralservicelist.com, checked 31 October 2026.
- EAP Foundation. New General Service List (NGSL, Browne, Culligan & Phillips). eapfoundation.com, checked 31 October 2026.
- Davies, M., & Gardner, D. Academic Vocabulary List. academicwords.info, checked 31 October 2026.