AbilityBenchPerceptual speed
Processing speed test — three markers, not one
Three thirty-second blocks measure how fast you get through simple material correctly: substituting a digit for a shape, deciding whether two rows of shapes match, and finding one shape in a grid of thirty-two. Each block returns correct decisions per minute, an accuracy figure with an interval on it, and how even your rhythm was — free, no account, two minutes including the rests. There is no press-the-key-when-it-flashes block here on purpose: that measures the display and your finger, not the rate you work at.
- 100% free
- No signup
- 3 blocks
- 30 s each
- No reaction-time block
- Rates, not ranks
Three blocks of 30 seconds each, back to back with a rest you control between them: substitute a shape for a digit, decide whether two rows of shapes match, and say whether one shape is somewhere in a grid of 32. Each one is a clerical decision about material sitting in front of you, and the result is how many of them you got right per minute in each block.
Block 1 — substitution
Six shapes, each paired with a digit, in a key that stays on screen. One shape appears below it; type its digit. The pairing is drawn fresh for each run, so nothing here rewards having been before.
Block 2 — paired comparison
Two rows of shapes, three, six or nine of them, side by side. Say whether the rows are identical or whether exactly one position differs. Longer rows take longer and that is the point of using three lengths.
Block 3 — visual search
One shape at the top, then a grid of thirty-two. Say whether the shape is somewhere in the grid. It is there on about half the items, and when it is there it appears exactly once.
Why there is no press-when-you-see-it block
A single key pressed at a single light is mostly display latency, nerve conduction and a finger, and it moves with the age of your monitor as readily as with anything about you. Putting it in a speed composite would let the equipment vote. Every block here makes you decide something about what is on the screen before the hand is allowed to matter, which is the difference between how fast you process and how fast you can press.
Two minutes of work in total including the rests. You can stop a block early and the page will report the seconds you actually worked rather than pretending it was thirty.
How to run the three blocks
Ninety seconds of scored work, split into thirds you rest between.
Take the three untimed items at the top of each block
Each block opens with three items that have no clock on them and tell you the right answer afterwards. They exist because the three blocks want three different things from you and a visitor who is still working out which key means what is being measured on their reading of the instructions. Turn them off with the checkbox if you have run this before; nothing in them enters any rate.
Answer with F and J in blocks two and three, digits in block one
Block one wants the digit paired with the shape, typed on 1 through 6 wherever your keyboard keeps them. Blocks two and three are two-choice — F for same and J for different in the comparison block, F for present and J for absent in the search block — so both hands rest on the keys and neither answer costs more to give than the other. Six digit buttons and two wide buttons are on screen throughout for anyone on a phone, and a run answered by tapping is reported as such rather than mixed in silently.
Rest between blocks, then read the three rates side by side
Each block is timed from the first item that counts and runs thirty seconds, throwing away whatever was up when it expired. Take the rest between blocks: three blocks run back to back measure a progressively more tired person, and the second and third would carry that instead of what they are for. At the end you get correct decisions per minute for each block, the accuracy behind each rate, the median time per item, and a steadiness figure that says whether you kept an even pace or alternated between bursts and stalls.
Technical specifications
| Structure | 3 blocks of 30 seconds — 90 seconds of scored work in total, with two rests the visitor controls and no time limit on the rests |
|---|---|
| Block 1, substitution | 6 shapes paired with the digits 1-6 in a key that stays on screen, reshuffled from a fresh seed each run. One item in six (16.7%) comes out right from a guess |
| Block 2, paired comparison | Two rows of 3, 6 or 9 shapes drawn with equal probability. Half the items are identical; on the rest exactly one position differs, so nothing is decidable from the row length. Chance is 50% |
| Block 3, visual search | A target shape and a grid of 32 cells, 8 across by 4 down, filled from the other 5 shapes. The target is present on about half the items and, when present, appears exactly once. Chance is 50% |
| Not included | Simple reaction time. One key at one light is dominated by display latency, input latency and nerve conduction, and a composite containing it partly ranks equipment. Every item here needs a decision about what is on the screen first |
| What each block reports | Correct decisions per minute, accuracy with a 95% Wilson interval, median item time with its outlier rule, the exact one-sided binomial probability of that accuracy from guessing, and steadiness as the standard deviation of item times over their mean |
| Practice | 3 untimed items per block, 9 in total, each showing the correct answer for 750 ms. Excluded from every rate, and they run on the same key the scored block will use |
| What leaves the page | Nothing. The copy button writes the three rates, the accuracies and the run seed to your own clipboard as plain text; the item-by-item times never leave the tab and are not stored in it either |
Frequently asked questions
Why three short blocks instead of one longer task?
Because processing speed is not a task, it is what several tasks have in common, and one task on its own cannot tell you which part of your score was the ability and which was that task. The paper literature never scores a single clerical test and calls it perceptual speed: it administers two or three, converts each to a standard score and averages them, precisely because each one has quirks of its own — a shape set you find awkward, a row length that suits your eye, a grid too dense for your screen. Running three different kinds of decision means an unusual result in one of them is visible as an unusual result in one of them, instead of arriving as your speed.
Why is there no reaction-time block in a speed test?
Because a simple reaction time is mostly not about you. Press one key when one square appears and the number you get contains the display's own latency, the keyboard's scan interval, the retina's transduction delay and the conduction time down your arm, all of which happen before any thinking does. Two people with identical cognition and different monitors get different scores. The moment an item requires you to work out what to press — which digit this shape takes, whether these rows match — the decision becomes the largest term in the interval, and that is the term this page is trying to measure. Reaction time is a real measurement of a real thing; it is a measurement of the apparatus plus a reflex, and it does not belong in a composite about how fast material gets handled.
Can I compare my three rates with each other?
No, and the page deliberately does not do it for you. Nine shapes to check against nine other shapes is more work per item than substituting a single shape, so the comparison block will always return a lower rate than the substitution block for the same person, and that gap says nothing about either ability. Comparing them would need each rate expressed against a population that had done that exact block, which is the thing that does not exist here. What the three rates support is the pattern: a run where all three sit unusually low together is a different observation from one where a single block does.
What is the steadiness figure and why is it worth having?
It is the spread of your item times divided by their average, so a smaller percentage means a more even rhythm. It earns its place because two people can produce the same number of correct items a minute in completely different ways: one working at a constant pace, the other alternating short bursts with stalls of two or three seconds. The average hides that entirely, and the variability is often the more informative half — an inconsistent pace under an unremarkable rate is one of the more reproducible findings in the literature on attention, while a fast, even rate is hard to produce accidentally. It is reported as a ratio rather than in milliseconds so that a slow, even worker and a fast, even worker come out looking similar, which they should.
Does tapping the buttons instead of typing change the numbers?
It raises every item time and it lowers every rate, so yes. Moving a finger to a target and landing on it takes time that a finger already resting on F does not spend, and the cost is tens of milliseconds per item — small against a whole item but not small against ninety seconds of them. The three blocks also do not pay it equally: six small digit buttons need more aiming than two wide ones, so a tapped run compresses the gap between the substitution block and the other two. The page counts how many answers arrived each way and says so beside the result, which is the only honest thing to do with a mixed run.
I scored 60% on the comparison block. Is that good?
It is barely distinguishable from guessing, and the page tells you so in as many words. On a two-choice block, a run of pure coin flips lands near 50%, and with the number of items thirty seconds allows, 60% happens by luck often enough that it is not evidence of anything. That is why each block prints the exact one-sided binomial probability of its own accuracy arising from guessing rather than a bare percentage: on a two-alternative task a percentage without that probability beside it is the single easiest way for a test to flatter someone. If your accuracy lands there, the rate printed next to it is not measuring speed of processing — it is measuring speed of pressing.
Is this the same thing an employer's timed assessment measures?
It overlaps with part of one and is not a substitute for it. Commercial cognitive-ability assessments used in hiring are mostly reasoning under time pressure — verbal, numerical and abstract items with a global time limit — where the speed component is one contributor to a total score rather than the thing being reported. This page isolates the clerical-speed component and reports nothing else, which makes it useful for finding out whether working quickly through simple material is a strength or a weakness for you, and useless as practice for the content of any particular assessment. If a specific test is coming, practise its item types rather than its clock.
Processing speed as a construct, and what a single-task version gets wrong
The reason anyone cares about perceptual speed is that it turned out to sit underneath a surprising amount of everything else. Salthouse (2000), Aging and measures of processing speed, Biological Psychology, is the convenient summary of a body of work showing that when several clerical-speed markers are combined into one composite, that composite accounts for a large share of the age-related variance in reasoning, memory and problem solving — and that no individual marker does the job on its own, because each one carries task-specific noise the composite averages out. The markers themselves have been stable for decades: substitution, in which an arbitrary key has to be consulted and applied; paired comparison, in which two strings are judged same or different; and search, in which a target has to be located among distractors. The paired-comparison form used here follows the pattern-comparison and letter-comparison tests introduced by Salthouse and Babcock (1991), Decomposing adult age differences in working memory, with the row lengths of three, six and nine those tests use.
The search block is where a browser version usually goes wrong, and the reason is worth stating because it changes what the number means. Finding one shape among five other kinds of shape is a serial search: attention has to be moved from cell to cell, so the time grows with how many cells were checked before the target turned up. That is a different regime from the case where the target differs from its neighbors in a single visual feature, which pops out in roughly constant time however many distractors there are — the distinction Treisman and Gelade (1980), A feature-integration theory of attention, Cognitive Psychology, set out. A grid built carelessly, with the target the only round thing among angular distractors, produces a pop-out task and a rate several times higher, which looks like a faster visitor and is a different experiment. Here every cell is drawn from a set of six shapes that share no single distinguishing feature, which keeps the block serial and keeps its rate comparable to your own next run.
What none of this supports is a rank. A published composite is a z-score inside the sample that produced it, and there is no sample here — so the page reports rates, says what each block would return from guessing alone, and stops. Two runs of this page a week apart are a real comparison; this page against somebody else’s test is not. If you want the substitution marker at the length the clinical literature uses, the digit symbol substitution test gives it ninety seconds and nine pairs instead of thirty and six, and the brain age test is the page that explains why a speed figure cannot be turned into a number of years. For the neighbors that speed contributes to rather than defines, the Mensa-style practice test puts reasoning under a global clock, the numeracy test asks for quantitative work with no clerical component at all, the attention span test measures whether performance holds up rather than how high it starts, and the Tower of London test removes speed from the scoring entirely and counts planning instead.
Nothing here is gated. There is no registration step, no address to hand over and no partial result that turns into a full one once you do; all three rates, their accuracies, their medians and the steadiness figure appear together as soon as the third block ends. One button copies them out as plain text, with the seed, so a run can be quoted or set beside a later one instead of being lost with the tab.
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.
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 a slow-processing profile, an attention disorder, or the mental fog people arrive here trying to explain. Only a qualified professional, working with more than a browser, can make that judgment.
Where the three rates are worked out
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 three blocks keep their finished items in a JavaScript array in this tab, which is why the result can show all three at once after the last one ends. Nothing in that array is written to storage or sent anywhere, and it is discarded when the tab closes — a reload loses a finished run, so copy the summary if you want to compare it with next week’s.