A Vocabulary Quiz Claiming to Measure 170,000 English Words Draws Scrutiny

A web app hosted on Google Cloud Run asks a simple question — how many of the roughly 170,000 English words do you know? — and answers it with a 100-question multiple-choice test that ends in a single number. The submission to Hacker News drew 501 points and 555 comments, and much of the discussion focused less on the number than on how it is produced.

How the test says it works

The app's "How is this calculated" page, quoted by commenters, states that the Oxford English Dictionary (Second Edition) lists approximately 171,476 words in current use. Rather than testing random words, the tool says it uses stratified sampling, dividing the language into five difficulty bands based on frequency of use: Core Basics (~3,000 words), Intermediate (~7,000), Advanced (~10,000), Expert (~25,000) and The Obscure (40,000+). The estimate is the sum of accuracy in each band multiplied by that band's size.

That arithmetic drew immediate scrutiny. The bands add up to about 85,000 words, not 170,000, so a user who answers every question correctly still ends up with a "scientific estimate" of roughly half the language. Several commenters reported exactly that: 85,000 words after a perfect run.

Where the words and definitions were questioned

Users disputed the placement of individual words. One commenter said "breviary" was labelled intermediate, which they considered far too easy a slot for a word many readers have never seen; others defended it as familiar to Catholics and to readers of medieval manuscripts. "Metamorphosis" and "kinetic" were labelled expert, which some found implausible. "Hippopotomonstrosesquippedaliophobia" was called a stunt word — recognisable only because its parts are, as one reader put it, less vocabulary than trivia.

Some definitions were criticised as loose, including "lethargy" given as "the state of being lethargic". The multiple-choice format also made guessing cheap: commenters noted that the longest option was often the correct one, and that distractors frequently followed a pattern of correct answer, opposite meaning and two unrelated things.

Pacing, bias and what users asked for

The most common complaint was mechanical: three clicks per word where one would do, and 100 words to get through. Several people stopped early. One commenter argued the test is barely adaptive, if adaptive at all, while others suspected everyone was shown the same word list, though some reported seeing different words. The suggested fixes included testing a much larger word set and sampling questions per user to reduce selection bias, updating the estimate after every answer, adding an "I don't know" button and showing a summary of the answers a user got wrong.

Reported results in the thread ranged from about 61,000 to 85,000 words. One commenter, who checked the app's number against a random sample drawn from a 1970 edition of the Shorter Oxford English Dictionary, concluded the figure was inflated; another argued the word list is heavily weighted towards Latin and Greek roots, which hands an advantage to speakers of those languages.