Vocabulary size

How accurate are vocabulary size tests? testyourvocab, VST and LexTALE compared

Accurate enough to place you in the right range, not to give an exact count. testyourvocab states a margin of error of about ±10%, assuming you tick only words you really know; Nation's Vocabulary Size Test reports word families, but guessing on its multiple-choice items can push scores up; LexTALE gives a percentage that tracks English proficiency, not a number of words. Scores from different tests cannot be compared directly, because each counts words differently.

By the Readlang Mobile team · Updated 29 October 2026

How accurate are vocabulary size tests?

Good enough to tell a learner with 3,000 words from one with 9,000; not good enough to tell 9,000 from 10,000, which lie within one another's ±10% margin. Every vocabulary size test works the same way: it checks a small sample of words drawn from frequency bands and scales the result up to the whole language. Three things make the final number less exact than it looks:

• Sampling. A test can check only a few hundred words at most. testyourvocab calculates its own sampling error at about ±10% (Preply); for the VST, Gyllstad, Vilkaitė and Schmitt (2015) found that its 10 items per 1,000-word band do not represent the band well enough.
• Guessing and overclaiming. On multiple-choice tests, learners can pick the right meaning without knowing the word, which inflates scores (Stoeckel, McLean & Nation, 2021). On self-report tests, they can tick words they only half know; that is why yes/no tests such as LexTALE mix in invented words (Lemhöfer & Broersma, 2012).
• The unit. Tests count word families, dictionary headwords or lemmas, and the same person gets very different totals in each (below).

A score is most useful as a rough level, and for comparing yourself with your own earlier score on the same test.

What do testyourvocab, the VST and LexTALE measure?

The three free tests that come up most often answer different questions. testyourvocab and the VST estimate how many words you know; LexTALE estimates how good your English is.

Sources: Preply test page and methodology; Nation & Beglar, 2007, and Nation's test page; Lemhöfer & Broersma, 2012.

How accurate is testyourvocab?

By its own calculation, to about ±10%: an estimate of 20,000 means somewhere between 18,000 and 22,000 (Preply, How it works). The address testyourvocab.com now redirects to a page on Preply, the online tutoring company, which hosts the test and its methodology; we checked both on 28 October 2026.

How it works. Step one shows about 40 words spread from the easiest to the hardest in English; step two shows about 120 words around the level the first step found. You tick each word you know at least one definition of, and the instructions say to leave a word blank if you are not sure of its meaning. The words are ranked by frequency in the British National Corpus, from a list of about 45,000 dictionary headwords. Regular inflections and derived forms are not counted separately, and words whose meaning is easy to work out, words easily confused with others, and cognates and false friends of Portuguese were removed (Preply).

Where the ±10% comes from. The methodology page derives it from sampling: with an average of 22.5 sampled words in the calculation and a spread of about 0.25 times the vocabulary size, the standard error is 0.0527, and 1.96 times that gives ±10.33%. Halving the margin to 5% would take about 380 more words; a 1% margin would take about 12,000.

What the ±10% does not cover:

• Overclaiming. The figure assumes you tick only words you know. The methodology page does not describe invented words or any other check for ticking too many, which yes/no tests such as LexTALE and Brysbaert's test use for exactly this reason (below).
• The comparison group. The test's FAQ says that "what really matters is not your absolute number, but rather your score relative to others who take the same test", and that test-takers report an average verbal SAT score of around 700, against around 500 for the US population. Comparisons with other test-takers are comparisons with a self-selected, highly verbal group (Preply).
• Retakes. Looking up words you met in the test inflates the next result; the FAQ suggests taking it at most once a year.

For learners, the FAQ gives rough benchmarks for people taking private English courses of about 3 hours a week: 1,500–3,000 words after a couple of years, 4,000–6,000 for intermediate English after 4–6 years, and 8,000–10,000 for a particularly good student at an advanced level after about 8 years. We found no independent published study of testyourvocab's accuracy.

How accurate is the Vocabulary Size Test?

It ranks learners well, but its raw score is likely to overstate how many words a learner can understand when reading. The Vocabulary Size Test (VST) by Paul Nation and David Beglar (2007) is widely used in research on learners of English. The original version has 140 multiple-choice items, ten from each 1,000-family band up to 14,000, and the score times 100 estimates your size in word families. Two 20,000-family versions, A and B, have 100 items each and are multiplied by 200. The tests are free to download from Paul Nation's resources page, with bilingual versions in Arabic, Gujarati, Korean, Japanese, Mandarin, Tamil, Thai, Vietnamese and Russian.

What the validation studies found:

• The items work as a scale. Beglar (2010) analysed the 140-item test with the Rasch model, and his results supported using it to measure learners' vocabulary.
• Guessing inflates scores. Because each correct answer is multiplied by 100 or 200, lucky guesses add up quickly, and Stewart (2014) argued that the Rasch analysis could not detect their overall effect. Applying a model that allows for guessing to 3,373 Japanese university students' answers, McLean, Kramer and Stewart (2015) concluded that a substantial share of what lower-level learners got right in the least frequent of the eight bands tested could be put down to guessing, and that such learners should not sit every band.
• Ten items per band are too few. Gyllstad, Vilkaitė and Schmitt (2015) compared VST answers with a criterion measure of the same learners' knowledge and found a gap between the two; they concluded that a higher sampling rate would be needed to represent each band.
• Good for ranking, not for an exact size. Stoeckel and colleagues (2019) gave 200 Japanese university students one of four VST variants (monolingual or bilingual, each with or without an "I don't know" option) and a test asking them to recall the meaning of the same words. Three variants correlated .88 to .91 with meaning recall, the bilingual one with "I don't know" .77. The authors concluded that the three variants other than the bilingual one with "I don't know" can rank or group learners, but that to estimate the vocabulary a learner can use for reading, raw scores need adjusting or a meaning-recall test should be used instead.

Who it is for. Nation's test page says the 20,000 versions can be used with native speakers and advanced non-native speakers, prefers bilingual versions "especially for lower proficiency learners", and says the VST is not suitable for learners who know fewer than 5,000 words; for them it recommends the Vocabulary Levels Test (Nation, checked 2026). The online version at VocabularySize.com showed a maintenance notice when we checked on 28 October 2026; Nation's page also lists the Compleat Lexical Tutor as an online host.

Is LexTALE a vocabulary size test?

No. LexTALE uses words to estimate how proficient you are in English, and reports a percentage, not a number of words. Kristin Lemhöfer and Mirjam Broersma (2012) built it as a quick check for psycholinguistic experiments with medium to highly proficient learners. You see 60 letter strings, 40 real English words and 20 invented ones, and decide for each whether it is a word. The score averages the percentage correct on words and on non-words, so saying yes to everything is penalised as much as saying no to everything. In the original study it took 3.5 minutes on average.

So LexTALE worked much better for the Dutch learners than for the Korean ones. For the Dutch group only, the authors proposed rough bands: 80–100% for C1 and C2, 60–80% for B2, and below 59% for B1 and lower, and they warned that the bands rest on limited data and that predictions for the Korean group were not accurate enough to use. A partial replication with learners of English whose first language was Spanish or Chinese found only low to moderate correlations with a standardised proficiency test, and its authors advise caution in using LexTALE as a measure of overall proficiency (Puig-Mayenco et al., 2023).

Versions exist for other languages, which makes LexTALE the easiest free check outside English: the original English, Dutch and German versions at lextale.com (Lemhöfer & Broersma, 2012; Brysbaert, 2013), French with 56 words and 28 non-words (Brysbaert, 2013), Spanish (Izura, Cuetos & Brysbaert, 2014), and a browser version with English, German, Dutch and Mandarin Chinese (LexTALE online). They too report a percentage, not a word count.

Why does each test give a different number?

Mostly because they count different things. One person can know 11,000 words and 42,000 words at the same time, depending on the unit:

A testyourvocab result in the tens of thousands and a VST result of 9,000 can therefore describe the same reader. Compare a score only with figures in the same unit, and only after checking which unit a test reports. Our page on how many words the average person knows explains the units in more detail. The Ghent test that produced the lemma figures, which mixed 67 real words with 33 non-words, could not be reached when we checked on 28 October 2026.

Which vocabulary test should you take?

• A learner of English who knows fewer than about 5,000 words: the Updated Vocabulary Levels Test or the New Vocabulary Levels Test, which report how well you know each 1,000-word level rather than one total (Nation's test page).
• An advanced learner of English, or one who already knows more than about 5,000 words, who wants a size in word families: a VST 20,000 version, bilingual if one exists for your first language. Its result can be set against research thresholds for reading and listening (next section).
• A quick check of general English level, for example for a study: LexTALE, about five minutes, with the limits above.
• A native speaker or very advanced learner who is curious: testyourvocab or the VST 20,000; read the testyourvocab number as headwords and as a comparison with a highly verbal group of test-takers.
• A learner of Dutch, French, German, Spanish or Mandarin: the LexTALE version for that language. We know of no free size test in word families for these languages that is as widely used as the VST is for English.

What does a VST score mean for reading?

Because the VST counts word families, its result can be compared with research on how many families it takes to know most of the words in a text. The figures below are from our guide to how many words you need; they count names as known.

A VST result of 5,000 families suggests you know about 95% of the words in an unsimplified novel: readable with effort. At 98%, one unknown word in fifty, reading becomes comfortable. Given the guessing problem above, treat a result near a threshold as possibly a little lower. The quickest cross-check needs no test at all: count the unknown words on one page of a book you want to read, as in the two-minute check in our first-week plan.

How can you get a more reliable result?

• Do not guess and do not tick half-known words. testyourvocab tells you to leave a word blank if you are not sure of its meaning; on the VST, a wild guess that happens to be right adds 100 or 200 families.
• Use a bilingual VST if there is one for your first language. Nation's page recommends them ahead of the English-only versions, especially for lower-proficiency learners.
• Skip what is clearly beyond you. For lower-level learners, a substantial share of correct answers in the least frequent bands comes from guessing (McLean et al., 2015); a levels test fits better.
• Take two forms rather than one. The VST 20,000 has versions A and B with different words; two results give a better idea of your range than one.
• Retest rarely and on the same test. Progress shows in months, not days, and a change from one test to another is mostly a change of unit.

References

1. Beglar, D. (2010). A Rasch-based validation of the Vocabulary Size Test. Language Testing, 27(1), 101–118. doi:10.1177/0265532209340194
2. Brysbaert, M. (2013). Lextale_FR: A fast, free, and efficient test to measure language proficiency in French. Psychologica Belgica, 53(1), 23–37. doi:10.5334/pb-53-1-23
3. Brysbaert, M., Stevens, M., Mandera, P., & Keuleers, E. (2016). How many words do we know? Practical estimates of vocabulary size dependent on word definition, the degree of language input and the participant's age. Frontiers in Psychology, 7, 1116. doi:10.3389/fpsyg.2016.01116
4. Gyllstad, H., Vilkaitė, L., & Schmitt, N. (2015). Assessing vocabulary size through multiple-choice formats: Issues with guessing and sampling rates. ITL – International Journal of Applied Linguistics, 166(2), 278–306. doi:10.1075/itl.166.2.04gyl
5. Izura, C., Cuetos, F., & Brysbaert, M. (2014). Lextale-Esp: A test to rapidly and efficiently assess the Spanish vocabulary size. Psicológica, 35(1), 49–66. hdl.handle.net/10651/26190
6. 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)
7. Lemhöfer, K., & Broersma, M. (2012). Introducing LexTALE: A quick and valid Lexical Test for Advanced Learners of English. Behavior Research Methods, 44(2), 325–343. doi:10.3758/s13428-011-0146-0; open-access text (PMC). Test: lextale.com
8. McLean, S., Kramer, B., & Stewart, J. (2015). An empirical examination of the effect of guessing on vocabulary size test scores. Vocabulary Learning and Instruction, 4(1), 26–35. doi:10.7820/vli.v04.1.mclean.et.al
9. Nation, P. Vocabulary tests. Victoria University of Wellington, School of Linguistics and Applied Language Studies. wgtn.ac.nz, checked 28 October 2026.
10. 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
11. Nation, I. S. P., & Beglar, D. (2007). A vocabulary size test. The Language Teacher, 31(7), 9–13. Open-access PDF
12. LexTALE online (English, German, Dutch, Mandarin Chinese). gasparl.github.io/lextale, checked 28 October 2026.
13. Preply. Test your vocab (formerly testyourvocab.com), test page and FAQ. preply.com, checked 28 October 2026.
14. Preply. Test your vocab: How it works. preply.com, checked 28 October 2026.
15. Puig-Mayenco, E., Chaouch-Orozco, A., Liu, H., & Martín-Villena, F. (2023). The LexTALE as a measure of L2 global proficiency: A cautionary tale based on a partial replication of Lemhöfer and Broersma (2012). Linguistic Approaches to Bilingualism, 13(3), 299–314. doi:10.1075/lab.22048.pui
16. Stewart, J. (2014). Do multiple-choice options inflate estimates of vocabulary size on the VST? Language Assessment Quarterly, 11(3), 271–282. doi:10.1080/15434303.2014.922977
17. Stoeckel, T., Stewart, J., McLean, S., Ishii, T., Kramer, B., & Matsumoto, Y. (2019). The relationship of four variants of the Vocabulary Size Test to a criterion measure of meaning recall vocabulary knowledge. System, 87, 102161. doi:10.1016/j.system.2019.102161
18. Stoeckel, T., McLean, S., & Nation, P. (2021). Limitations of size and levels tests of written receptive vocabulary knowledge. Studies in Second Language Acquisition, 43(1), 181–203. doi:10.1017/S027226312000025X
19. Webb, S., & Rodgers, M. P. H. (2009). The lexical coverage of movies. Applied Linguistics, 30(3), 407–427. doi:10.1093/applin/amp010