Throughput under a clock

Concentration test scored on speed, errors and the swing between them

Eighteen rows of fourteen glyphs arrive one row at a time with ten seconds on each, and your job is to mark every glyph whose bowl faces left and which carries exactly two marks on its stem. Everything else — the same bowl with one, three or four marks, or the mirrored bowl with any number — is a decoy. The page reports four things separately: how far you got, how many targets you passed over, how many decoys you marked, and how much your rate moved from your best row to your worst. It is free, the result is on screen at the end, and nothing is stored. There is no norm here and no cut score, because the published sheets in this family sell their tables along with their paper and neither transfers to glyphs drawn for a browser.

100% free
No signup
252 glyphs
10 s per row
Rate swing reported

A row of 14 glyphs appears. Mark every one with its bowl on the left and exactly two marks attached to the stem — counting the marks above and below together. Everything else is a decoy: the same bowl with one, three or four marks, or the mirrored bowl with any number.

mark it

leave it

The sheet, spelled out

Target: bowl on the left of the stem and exactly two marks on it, above and below counted together
Decoys: the same bowl with one, three or four marks, and the mirrored bowl with any of the four counts — 13 glyphs in all against 3 that qualify
Row: 14 glyphs, 4 of them targets — 29% — with no more than 3 of one kind together
Rows: 18, run back to back with no break
Clock: 10 seconds a row, or less if you call it swept
Reached: the last glyph you marked or stepped onto, so a target beyond it is not counted against you
Marking: tap a glyph, or walk the row with the arrow keys and mark with Space
Void: called swept inside 150 ms, called swept before the row painted, or open while this window was at the back

The bowl is a mirror rather than a different shape on purpose. A left-facing and a right-facing form of the same figure are the hardest pair in this family to tell apart at speed, and a decoy that differed by an obvious feature would turn the row into a search instead of a count.

Four rows is enough to see whether the glyphs are legible on your screen and not enough for the figure this page is built around: a swing between rows needs enough rows for the word swing to mean anything.

How to work the concentration test

One rule with two halves, eighteen rows, and a clock that does not wait.

Fix the rule before the first row: bowl on the left, two marks in total
Both halves have to be true at once. A left-facing bowl with three marks is a decoy, and so is a right-facing bowl with exactly two — which is the pair that catches people, because a mirror image of a familiar shape reads as the shape itself at speed. Marks above and below the stem are counted together, so one above and one below qualifies just as two above does. Five worked examples sit above the start buttons; look at them until the mirrored pair stops looking the same.
Work left to right and do not skip ahead
Tap a glyph to mark it, or walk the row with the arrow keys and mark with the spacebar. The row is scored on how far you reached — the last glyph you touched — so targets beyond that point are not counted against you, and jumping to whatever catches your eye produces a reach that is a lie about what you actually inspected. Press Enter or the button below the row when you have swept it and the next row starts immediately; otherwise the clock closes it at ten seconds.
Read the four figures as a set, never as a score
The results give you glyphs processed, targets found with a 95% interval, decoys wrongly marked, and a per-row table showing rate, misses and false marks for every row you finished. Speed and accuracy trade against each other, so a large count with a long list of misses and a small count with none are the same performance taken at two different settings. The figure that is not a trade is the swing between your rows: two people can process the same number of glyphs in three minutes and be completely different workers, and that difference lives entirely in the gap between their fastest and slowest rows.

Technical specifications

Sheet	18 rows of 14 glyphs, 252 in all, generated fresh from the run seed on every attempt so no two runs are the same sheet. A 4-row option exists to check legibility and is far too short for the swing figure
Target	Bowl left of the stem plus exactly two marks. 4 per row — 29% of the glyphs — with no more than 3 targets or 3 decoys in an unbroken stretch
Glyph set	16 forms in total: two mirrored bowls crossed with eight arrangements of one to four marks, drawn as vector shapes rather than letters so they render identically at any size and depend on no installed font. 3 of the 16 qualify
Clock	10 seconds per row, or less when you call the row swept. Practice rows run 14 seconds and mark themselves up afterwards, showing missed targets in green and wrongly marked decoys in red
Reached	The index of the last glyph you marked or stepped onto, following the paper convention of scoring to the last canceled character. It undercounts by a glyph or two when a row ends on a decoy, and does so identically for everyone
Speed figure	Glyphs reached divided by the seconds that row actually took, not a raw count. The paper procedure gives a fixed interval because the examiner cannot see when a row was finished; a browser can, so a row ended early is honestly reported as a shorter row rather than a slower one
Swing	Fastest row rate minus slowest row rate, in glyphs per second, printed with the standard deviation across all rows beside it. This is the figure that separates a steady worker from an erratic one at equal totals
Norms	None. The published cancellation instruments are commercial products whose tables are calibrated to their own character sheets, and a table applied to a different stimulus set is a number with nothing behind it

Frequently asked questions

Why do you use invented glyphs instead of the letters the real test uses?

Because the character sheet of a published instrument is its copyrighted content, and reproducing it would also be pointless: a score from a reproduced sheet still could not be read against that instrument's norms, since those were standardized on a supervised paper administration with a pencil. What is not proprietary is the structure of the discrimination — a target that differs from its neighbors only in a detail you have to count, embedded in a dense row under time pressure. That structure is what makes the task work, and it is what the glyphs here implement. Drawing them as vector shapes has a second benefit: they look the same on every device and cannot be broken by a missing font.

Ten seconds a row feels rushed. Is the pressure part of the test?

It is the whole reason the task measures anything. Given unlimited time almost everyone marks almost everything correctly, and the result becomes a measure of patience with no ceiling in sight. The clock forces a rate to be chosen, and choosing a rate is what exposes the trade this test is about: push it and you start passing targets over, hold back and you reach glyph nine instead of glyph fourteen. Neither is an error in the ordinary sense. The pressure is also why the page reports the per-row table — where in the eighteen rows you moved your rate is more revealing than the rate itself.

What does the swing between rows actually tell me?

How consistently the effort was applied, which is independent of how much of it there was. Consider two runs that both process 200 glyphs in three minutes: one at a steady 1.1 glyphs per second in every row, the other alternating between 1.6 and 0.6. The totals are identical and the working is not. A large swing usually means the rate is being set by something other than the task — a burst of effort, a lapse, a moment spent recovering from a mistake — and the paper versions of this test report it as a figure in its own right for that reason. It needs enough rows to mean anything, which is why the four-row option says it cannot form it.

The mirrored bowl keeps catching me out. Is that a real effect or bad design?

It is a real and well-known one, and it is why the decoy set is built that way. A shape and its mirror image are treated as the same object by a visual system that spent its whole evolution never needing to tell a left-facing thing from a right-facing one, which is also why children reliably write letters backwards for a while and why the mirror pair among the alphabet's letters is where reading errors cluster. Under speed pressure that tendency comes back. If the decoys differed by an obvious feature the row would become a search, which is a different task with a different measurement, and this site has that one on its own page.

Does the page count targets I never reached as misses?

No, and this is the single most important detail in the scoring. Only targets before the last glyph you touched are counted, so a row you got a third of the way through contributes a third of a row's targets to your denominator, not a whole row's. Without that rule the score would punish a careful worker twice — once by giving them a lower count and again by charging them for everything they never saw — and the two measures would stop being independent. It is the convention the paper versions use, and it is the reason the results panel prints 'targets in the stretch you reached' rather than 'targets'.

Is 'found minus wrongly marked' the score I should quote?

It is the traditional single figure for this family and it is printed here, but it should never travel on its own. Subtracting false marks from correct ones stops a run that marks everything from scoring well, which is what it was designed to do — but it also gives the same number to a slow careful worker and a fast careless one, who reach it by opposite routes. That is why it sits underneath its two parts on this page rather than above them, and why the copy button writes out every row instead of a headline.

Can I take this on a phone?

Yes, and it changes the run in a way worth knowing about. Fourteen glyphs across a phone screen puts each one at around twenty-four pixels, which is legible for the bowl and tight for counting marks, so a phone run tends to come back slower and more accurate. Tapping and arrow-key walking are also genuinely different tasks: the tapping hand chooses where to go and can skip, while the arrow keys step through every glyph in order, which usually produces a higher reach and a lower rate. The results panel records how each row was ended so you can see which you used.

What a cancellation task measures, and the four numbers it takes to say it

Cancellation tests are the oldest working format in this subject and among the least glamorous: a dense field of near-identical symbols, a rule that picks out a minority of them, and a clock. What makes them durable is that they produce several genuinely independent quantities from one procedure. Speed is how much material was worked through. Accuracy splits in two — targets passed over, which is a sweep that outran the looking, and decoys marked, which is the rule applied and applied wrongly. And consistency is the movement of the rate from row to row, which is uncorrelated with all three. Collapse those into one score and you get a number that a fast careless worker and a slow careful one can both produce, which is exactly the number every other page in this corner of the web reports.

The design of the decoys is not decorative. The rule here has two independent halves — which way the bowl faces, and how many marks are on the stem — and the decoy set is built so that every glyph satisfies exactly one of them or neither. That forces both to be checked on every item, and it is why the errors are informative: marking a right-facing glyph with two marks means the counting worked and the mirror caught you, while marking a left-facing glyph with three means the opposite. A cancellation task whose decoys differ from the target by an obvious feature turns into a pop-out search, where the whole row is processed at once and the clock stops mattering — that version is worth taking too, and it lives on the selective attention test, where the number of distractors is manipulated on purpose.

The one real departure from the paper procedure is that a row here can be ended early. On paper the interval is fixed because the examiner has no way of knowing when the participant reached the last character, so everybody gets the same seconds and the score is a raw count. A browser knows, so the row closes either at the cap or when you say you have swept it, and the speed figure becomes a rate. That is a better measurement and a worse comparison: it means a count from this page is not the same quantity as a count from a paper sheet, which would matter if there were norms to compare against — and there are not, for the reason above. Two neighboring pages take the time dimension instead of the density one: the attention span test holds the task trivially easy and watches five minutes pass, and the continuous performance test makes the target rare and the run ten minutes long. For interference generated inside a single item rather than by its neighbors, the Stroop test is the canonical case, and the language processing test asks the same speed-against-accuracy question of words rather than shapes.

Nothing on this page is measured in milliseconds, which is unusual here and worth stating: the display and input latency that every timed page on this site has to declare amounts to a hundredth of one of these ten-second rows and reaches none of the figures. What that buys is a task that measures the same thing on a slow machine as on a fast one. What it costs is that this page cannot say anything about reaction time, and the three siblings above are where that question belongs.

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 an attention disorder, a concentration deficit or any condition at all. Only a qualified professional, working with more than a browser, can make that judgment.

Where your marks and your row times stay

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.

Every sheet is generated in this tab from a seed shown with your result, so the 252 glyphs you worked through are reproducible from that number and are never stored anywhere. Which glyphs you marked exists as a set of indices in memory until the row closes, at which point it becomes four counts and the indices are discarded.