AbilityBenchVocabulary depth
Synonym test that finds your level
Pick the word closest in meaning to the one on screen, four options at a time, and the ladder moves: two right answers and the next word is rarer, one wrong and it is commoner. Twenty-four items, free, no account, no clock on any of them. The run ends with the band it settled on, the words you missed with what each one means, and the option you should have taken.
- 100% free
- No signup
- 24 adaptive items
- Six frequency bands
- Missed words defined
24 items, four options each, and the words get harder or easier depending on how you are doing. A fixed list would spend most of itself on words you certainly know and words nobody knows; this one heads for the band where you are genuinely unsure and stays there, which is where the information is.
The six rungs, easiest first
- 1Everyday — spoken conversation, a hundred or more times in every million words of text (Zipf 5 and above)
- 2Common written — any newspaper page, tens of times per million words (Zipf about 4.5)
- 3Newspaper — the leader column rather than the sports report (Zipf about 4)
- 4Educated prose — a long-form magazine essay, a few times per million words (Zipf about 3.5)
- 5Literary — novels and criticism, around once per million words (Zipf about 2.5 to 3)
- 6Rare — you have probably met it in print once, if at all (Zipf 2 and below)
Two rungs apart is roughly a hundredfold difference in how often the word appears in print. Each of the 48 words was placed in a band by hand against that scale, not read out of a corpus, so the rung is this page’s judgment about the word and the page says so rather than dressing it as a lookup.
The warm-up words are easy ones from outside the bank, they tell you the answer afterwards, and none of them reaches the score. There is no clock on any item here — take as long over a word as you want.
How to take the synonym test
One word, four options, and a ladder that follows you rather than the other way round.
Take three warm-up words, or skip them
The warm-ups are easy words held outside the scored bank, so nothing is spent on them. Each one tells you afterwards which option was right and what the word means, which is the only place in the run where you are told anything mid-flight. Skip them if you have done this before — the scored items are drawn fresh either way.
Answer 24 items, pressing 1 to 4 or tapping
Every item starts at the band the ladder has walked to, beginning at band three of six. Two correct answers in a row move you up a band, a single wrong answer moves you down one, and the sequence quickly starts oscillating around the level where you are about seven-in-ten right. There is no time limit and no penalty for guessing: a wrong guess simply moves the ladder down, and it will climb back if the next two are right.
Read the band, then take the words back
The result leads with the band the walk settled on, worked out as the mean of the turning points rather than as a count of correct answers, and says how many turning points went into it. Under that sits the per-band record, then every word you missed with its definition and its answer, then the full list of the 24 words you saw. That list is the part worth keeping — it is a reading list of exactly the words sitting at the edge of what you already know.
Technical specifications
| Items | 24 scored, drawn one at a time from a bank of 48 arranged in six bands of eight, plus 3 optional warm-up words held outside the bank. No word is shown twice in a run |
|---|---|
| How the ladder moves | Two consecutive correct answers step one band toward the rare end; one wrong answer steps one band back. That is the 1-up/2-down rule, and a walk under it settles where the answerer is right 70.7% of the time — printed on the page rather than asserted, because the figure follows from the rule |
| The six bands | Everyday, common written, newspaper, educated prose, literary and rare. 1-7 Zipf, so band one sits near the top of that range and band six near the bottom — two bands apart is roughly a hundredfold difference in how often the word appears in print |
| Where the bands come from | van Heuven, Mandera, Keuleers & Brysbaert (2014), SUBTLEX-UK: A new and improved word frequency database for British English, Quarterly Journal of Experimental Psychology. The scale is published; the placement of these 48 particular words into its bands was done by hand for this page and is a judgment rather than a corpus lookup |
| Options and chance | Four per item, one correct, so blind guessing scores 25% and a full run answered at random averages six right. The three wrong options are related to the item word by topic or register but never by meaning — an item with two defensible answers is a scoring fault, not a hard item |
| The reported level | The mean of the turning points after the first two are discarded as the approach run, given to one decimal place. A run that reaches the top of the bank and stays there reports no level at all: the walk was pinned against the ceiling, and the number would then describe the bank rather than the reader |
| What voids an item | An answer arriving within 150 ms of the word being painted, an answer arriving before it was painted at all, and any item the tab was hidden or unfocused during. A void item moves the ladder in neither direction and is asked again with a different word |
| What leaves the page | Nothing at all unless you press copy, which puts a plain-text summary on your own clipboard. The 48-word bank ships with the page, so the whole run happens with no request to any server |
Frequently asked questions
Why did the words suddenly get so much harder?
Because you got two in a row right, which is the signal the ladder is built to act on. The jump feels abrupt because a band here is not a small step: each one is roughly a tenfold difference in how often the word turns up in print, so moving from the newspaper band to the literary band takes you from words you meet weekly to words you meet once a year. The run is trying to find the level at which you are wrong about three times in ten, and the fastest route to that level is to keep climbing until you are wrong.
Two of the four options both look right. What am I supposed to do?
Pick the one that matches in register as well as in sense, and if it still reads as a genuine tie, the item is faulty rather than hard. Near-synonyms almost always differ in something: how formal they are, what they usually attach to, whether they carry approval. The bank was written so that the three wrong options are related to the item word by topic, by sound or by the kind of thing it describes, and never by meaning, precisely so that a knowledgeable reader is not being asked to split hairs. An item where two answers are genuinely defensible measures nothing except which of the two the writer preferred.
Does a synonym test measure the same thing as knowing a word?
It measures recognition, which is the larger and easier half of knowing a word. Recognizing that austerity and severity are close is not the same as being able to produce austerity when you need it in a sentence, and the gap between the two is wide enough that the passive and active vocabularies are usually treated as different quantities. Four options make the recognition easier still, because the options themselves are evidence: seeing plentiful next to expensive tells you the item is about quantity before you have thought about the word at all.
Why 24 items? Other tests ask 40 or 100.
Because an adaptive run buys its precision from where the items are, not from how many there are. A fixed 100-item list spends most of its length on words you certainly know and words nobody knows, and neither kind tells anyone anything; the ladder puts almost every item within a band of your edge, which is where an answer is informative. The cost is that the estimate rests on turning points rather than on a total, so a run with only two or three of them reports no level — the results panel says so rather than averaging what it has.
Is the band number a percentile, or comparable to anyone else's?
It is neither. The band is a position on this page's own six-rung ladder, built from 48 words that were placed into bands by hand, so it says where you sit against this bank and nothing about where you sit against other people. There is no standardization sample behind it and no distribution to rank against, which is why the result never converts the band into a rank. Two people can compare bands only if both ran this same bank, and even then the comparison is between two adaptive walks rather than between two scores.
I answered everything right and got no level. Is that a bug?
No — it is the ladder telling you the truth about its own ceiling. A staircase that never gets an answer wrong has no turning points, so there is nothing to average, and it spends the run pushed against the rarest band the bank holds. Reporting a level in that case would describe the bank rather than the reader, so the panel says the ceiling did not hold you and stops. The honest fix is more items above band six, and eight rare words is what this page has.
Should I guess when I have no idea?
Yes, and the ladder is designed to absorb it. A wrong guess costs one step down, which two correct answers undo, and the reversals that the estimate is built from are exactly the moments where a guess went wrong — so a run with a few guesses in it converges just as well as one without. What does distort the estimate is refusing to answer, since an item with no response is scored as wrong without you having looked at the options. If you can narrow four to two, the guess is worth considerably more than the 25% floor.
Why frequency is what makes a word hard
The difficulty of a vocabulary item is mostly not a fact about the word’s meaning. It is a fact about how often you have met it, and that quantity is measurable in a way that difficulty is not: count occurrences in a large corpus, take the logarithm, and you have a scale on which words can be ordered without anybody’s opinion entering. van Heuven, Mandera, Keuleers & Brysbaert (2014), SUBTLEX-UK: A new and improved word frequency database for British English, Quarterly Journal of Experimental Psychology sets that scale out — 1-7 Zipf, where 3 is one occurrence per million words and 6 is a thousand. American frequency counts built the same way are in Brysbaert & New (2009), Moving beyond Kucera and Francis: A critical evaluation of current word frequency norms and the introduction of a new and improved word frequency measure for American English, Behavior Research Methods. The reason this page names the scale rather than quietly ordering its words is that the two halves have very different standing: the scale is published, and the placement of these particular 48 words into its bands was done by hand here. One is a citation and the other is an editorial judgment, and a page that blurs them is claiming a precision it did not buy.
Two things follow that most synonym tests get wrong. The first is the ceiling. A test whose hardest item is a band-six word cannot distinguish anybody above band six, so a perfect run is not a measurement of the reader — it is a report that the bank ran out. Fixed-list tests almost never say this, and the ones that convert a perfect score into a confident number are converting an absence of information. The second is what a recognition item can be turned into. Choosing the closest of four options is evidence about the word, but it is not evidence about how many words you know: the arithmetic from one to the other needs a sampling frame and a correction for guessing, which is a different instrument and lives on the vocabulary test. A synonym ladder that prints a word count has skipped the step where the count would have to be defended.
What the ladder is genuinely good at is finding your edge quickly, which is what makes the list at the end more useful than the band above it. Words sitting one band past where you settled are the ones you will meet often enough to be worth learning and rarely enough that you have not yet — a better reading list than any fixed vocabulary course, because it was assembled by your own answers. If the question behind your visit is a level rather than an edge, the English level test covers grammar and reading alongside vocabulary; if it is how fast you can produce words rather than recognize them, the verbal fluency test asks for output against a clock, which is the half a multiple-choice item cannot reach. And if you came here because reading feels slow rather than because words feel unfamiliar, the reading speed test measures the thing you actually noticed.
A reaction time here is the interval between the frame that painted the stimulus and the timestamp the browser attached to your key, both read from the same monotonic clock. What neither can see is the display pipeline behind it, so on a 60 Hz screen roughly 16 ms of every figure below is the machine rather than you. That is the timing floor: two numbers closer together than that are the same number, and this page reports no precision it cannot support.
Nothing on this page is scored against a clock, and the floor matters here for one rule only: an answer arriving within 150 ms of a word being painted is discarded, because a word has to be read and four options scanned before any choice is possible, and no run of that is finished in a sixth of a second. Time per item is reported in seconds, which is the precision the measurement supports.
This is a measurement exercise, not a clinical assessment. It reports what you did on this page against a stated reference and nothing more — it cannot establish dyslexia, a language disorder or a reading difficulty. Only a qualified professional, working with more than a browser, can make that judgment.
Where the 24 answers are checked
Every number on this page is worked out by JavaScript running in the tab you are reading it in. Your answers, your reaction times and your score are never uploaded, logged or kept — which is also why the test carries on working after you disconnect from the network, and why nothing here can be held back behind an email address.
The word bank and the ladder both live in the JavaScript this page loaded, so every comparison between your answer and the key happens in this tab and nothing is asked of any server during a run. That is also why a reload loses a finished result, and why the copy button carries the run seed: the seed is the only thing that can rebuild the exact sequence of words you were shown.