Free cue validity at 100 and 700 ms

Posner cueing task at two cue-target delays

Keep your eyes on the center cross. One of two boxes thickens for 50 ms, and then a dot appears in one of them — press F for the left box, J for the right. The flash is a coin toss and tells you nothing, and it arrives either 100 ms or 700 ms before the dot, so the run reports what a meaningless cue was worth at each delay: usually a head start at 100 ms and a penalty at 700 ms. 64 scored trials, free, no account, and an optional 32-trial central-arrow block that shows why the penalty happens.

100% free
No signup
64 scored trials
Two delays in one block
Delay measured, not assumed

Keep your eyes on the cross in the middle. One of the two boxes will thicken for 50 ms, and then a dot appears in one box or the other — press F for the left box, J for the right. The flash is a coin toss and tells you nothing, which is the point: any head start it gives you was not a decision.

The left box as it looks while cued: border 6 px instead of 2 px and a gray fill. Nothing here depends on hue, so the cue is visible with any form of color vision. Box centers are 128 px either side of the cross and nothing on this screen is timed.

Everything the flash block does

Cue: one box thickens for 50 ms; cue side and dot side are drawn independently, so it is right half the time
Delay: 100 ms or 700 ms from cue onset to dot onset, mixed trial by trial
Your job: F when the dot is in the left box, J when it is in the right
Cells: four — rightly or wrongly cued, crossed with the two delays, 16 trials in each
Geometry: boxes 56 px square with centers 128 px either side of the cross — 33.9 mm, about 3.2° if you sit 60 cm away
Timing: a cross for 500-900 ms, the cue, the gap, then the dot until you answer or 1500 ms passes; 400 ms between trials
Void: anything answered inside 150 ms of the dot, anything answered before it, and any trial the browser stopped painting because this tab went to the background

Get both fingers onto their keys before you press start. The arrow block is the other experiment on this page: a center arrow that is right 75% of the time, at the longer delay only, where the flash has stopped helping. Run the flash block first — the arrow block is only interesting next to it.

How to run the cueing task

Two fingers, one cross to stare at, and a flash you are asked to ignore and cannot.

Look at the cross and keep looking at it
The cross in the middle is the only thing you are meant to be looking at. Both boxes sit within a few degrees of it, so you can see either one without turning your eyes, and the whole task is built on the difference between moving your attention and moving your eyes. Nothing here can tell whether you looked, so if you catch yourself glancing at the flash, that run is measuring your eye muscles instead — say so to yourself and start again.
Answer the dot, not the flash
Press F when the dot lands in the left box and J when it lands in the right. Both fingers belong on their keys before the first cross appears, because a hand that has to hunt for a key spends the hunt inside the interval being measured. The flash is right half the time by construction, which means there is no strategy available and no reason to obey it: whatever it does to your reaction time, it does without your permission. Eight practice trials count for nothing and report right or wrong after each one, so the choice between the two keys has stopped being a decision by the time anything is timed.
Read the two delays against each other
The result gives the wrongly-cued-minus-rightly-cued difference at each delay with a 95% interval, then the change between them. A positive number at 100 ms and a negative one at 700 ms is the classic pair. Underneath, the page prints the delay it actually delivered — measured from the frame that painted the cue to the frame that painted the dot — because a delay a page merely asked for is not a delay it achieved.

Technical specifications

Trials	64 scored in the flash block: 16 in each of four cells, rightly or wrongly cued crossed with the two delays. The optional arrow block adds 32. Eight untimed practice trials come first and appear in no figure
Cue and its validity	One 56-pixel box's border goes from 2 to 6 pixels and its fill from white to gray, for 50 ms. Cue side and dot side are drawn independently, so it is right on half the trials. The arrow block replaces it with a 200 ms glyph at the cross that is right on 24 of 32. Neither cue uses hue, so both are visible with any form of color vision
Cue-target delay	100 ms and 700 ms from cue onset to dot onset, shuffled trial by trial so the delay cannot be anticipated; the arrow block runs 700 ms only. A phase boundary can only fall on a frame, so the delivered delay is the requested one rounded up to whole frames — measured per trial and printed with the result rather than assumed
Geometry	Boxes 56 CSS pixels square with centers 128 pixels either side of the cross, which is 33.9 mm where a CSS inch is 96 pixels, and about 3.2 degrees of visual angle for a reader sitting 60 cm back. The degree figure is conditional on purpose: this page cannot ask how far away you are sitting and cannot measure how large a pixel is on your panel, so the pixel count is what it commits to
Response	F for the left box, J for the right, taken from physical key position rather than from the character the layout produces. On-screen buttons do the same job on a touchscreen and are counted separately, because aiming a thumb costs more than lifting a resting finger
Trial structure	A cross for 500-900 ms drawn uniformly, then the cue, then the gap, then the dot until a key lands or 1,500 ms passes, then 400 ms of nothing. One rest after trial 32 of the flash block; the arrow block is short enough not to need one
Discarded trials	Anything answered within 150 ms of the dot appearing, anything answered before it appeared — during the cross, the cue or the gap — and every trial that spent part of itself behind another window. A trial broken that way is not thrown away: it goes back into the queue, so the block still hands you 64 scored answers
Reference figure	20-50 ms — Posner (1980), Orienting of attention, Quarterly Journal of Experimental Psychology. It is a span collected across studies at cue-target intervals that mostly sat below 300 ms, so it belongs beside the short-delay figure and has nothing to say about the long one. On the reversal: Posner & Cohen (1984), Components of visual orienting, Attention and Performance X

Frequently asked questions

Why does this run two cue-target delays instead of one?

Because the delay decides the sign of the answer, so a single-delay version reports a number nobody can interpret. A flashed cue speeds responses to the place it flashed for roughly the first quarter of a second and slows them after about 300 ms, which means two honest pages running one delay each can publish opposite findings and both be right. Running 100 ms and 700 ms in the same block, shuffled so you cannot tell which is coming, puts the whole curve's two ends on one screen and makes the direction a result rather than a coincidence of setup.

What is inhibition of return, and does a negative number mean I have it?

It is the finding that attention which has been pulled somewhere and then left resists going back, and a negative number at the 700 ms delay is what it looks like on a keyboard. Posner and Cohen described it in 1984: the same peripheral flash that helps early hurts late, as though the location had been tagged as already-checked so that searching does not revisit it. Everyone shows it under the right conditions, so having one distinguishes nobody; not showing one on 16 trials a cell is equally uninformative, because the effect is smaller than the facilitation it follows and needs more trials to separate from zero.

Why is the flash a coin toss when Posner's cues predicted the target?

Because an uninformative cue is the only way to measure reflexive orienting, and an informative one is the only way to measure voluntary orienting — they are two different experiments that cannot share a block. Posner's central arrows were valid on 80% of trials precisely so that obeying them paid; if the arrow is right half the time there is nothing to obey and the effect largely disappears, which is the commonest fault in browser versions of this task. A peripheral flash is the opposite case: it is deliberately worthless, so any advantage it produces was not a decision. That is why the arrow block here runs at 75% valid and is a separate run rather than a third condition.

Nothing stops me from looking at the box. Does that ruin the measurement?

It changes what is being measured, and this page cannot detect it. The paradigm is about covert attention — the shift you make without moving your eyes — and a laboratory version watches fixation with an eye tracker and throws away every trial containing a saccade. A browser has no such instrument. If you look at the flash, a rightly cued trial finds your eyes already at the dot and a wrongly cued one has to bring them back, which produces a validity effect several times larger than the published span and is really a measurement of two saccades. A very large effect on this page is therefore a reason to suspect your eyes, not a reason to be pleased.

Posner's task was a single key for 'I saw it'. Why does this one ask which side?

Because a single key on a web page cannot be audited and this one can. Detection is the faster and more classical version, and it is also the version where a participant who presses on a rhythm produces a full set of plausible reaction times that no analysis can distinguish from real ones — laboratories handle that with target-absent catch trials and an experimenter in the room. Asking which box makes every trial carry a right answer, so a rhythm shows up immediately as an accuracy hole in the cell table. The trade is real: choosing between two keys adds a response-selection stage, so the absolute times here run above published detection latencies and only the differences between cells are comparable.

My effect is far bigger than 20-50 ms. Is the run wrong?

Probably not wrong, but probably not measuring only attention either. The three usual causes are eye movements, a slow block and a small sample. Eye movements inflate it as described above. A block answered near the 1,500 ms window rather than near the dot inflates everything, because a difference of tens of milliseconds sits on top of whatever else is happening in a distracted second. And 16 trials in a cell is few enough that the 95% interval printed beside each effect is often 40 or 50 ms wide, so a single number that lands well outside the published span may not be distinguishable from one inside it. The interval is the part of the result to read; the point estimate is the part that moves.

Does the arrow block have to be run second?

No, but it means much less on its own. Its whole point is the comparison: at the same 700 ms delay where an uninformative flash has usually stopped helping, an informative arrow is still worth following, which is the cleanest available demonstration that the reversal belongs to reflexive orienting rather than to the passage of time. Run alone it produces a single validity effect with nothing to sit beside. Either block can be re-run without touching the other, and re-running one replaces only its own figures.

What the cueing task measures, and the instrument a browser does not have

Posner’s 1980 paper set out to separate attention from the eyes. A participant fixates a point, something indicates where a target is likely to appear, and the target then appears there or somewhere else; the difference between the two response times is the evidence that something was moved in advance of the eyes moving. The published span for that difference is 20-50 ms, per Posner (1980), Orienting of attention, Quarterly Journal of Experimental Psychology, and for four years it looked like a single quantity. Then Posner & Cohen (1984), Components of visual orienting, Attention and Performance X showed that a flashed peripheral cue reverses: the advantage decays within about 200 ms and becomes a reliable cost by 300 ms or so, as though the visited location had been marked so that attention would not waste itself going back. That reversal is why the cue-target interval is not a detail of setup on this page but the manipulation itself, and why any cueing result quoted without its interval is uninterpretable.

The distinction that organizes everything here is between two ways of aiming attention. An exogenous cue — the peripheral flash — captures attention reflexively, works within a hundred milliseconds, cannot be resisted, and is the only kind that produces inhibition of return. An endogenous cue — a central arrow, a word, anything symbolic — has to be read and then acted on, takes a few hundred milliseconds to have its effect, can be deliberately ignored, and keeps helping for as long as it stays worth obeying. Two consequences follow, and browser versions of this task routinely miss both. The first is that a symbolic cue must actually predict something: a 50%-valid arrow is a decoration, and a page reporting a “cue validity effect” from one has measured almost nothing. The second is that the two cue types cannot be interleaved at their proper validity ratios, because the moment an arrow predicts, the flash beside it stops being uninformative. Hence the two blocks here, run separately, with the arrow set to 75% valid and the flash left at chance.

The departure a browser cannot avoid is the eye tracker. Every laboratory version of this task monitors fixation and discards trials containing a saccade, because the covert shift the paradigm exists to isolate is exactly the thing an eye movement destroys — and this page has no way to know whether you looked. It states that plainly rather than quietly reporting a number, and it is the reason an unusually large effect here should be read as a suspicion about your eyes rather than a finding about your attention. The other departures are stated where they occur: the geometry lives in CSS pixels because degrees of visual angle need a viewing distance no web API reports, and the response is a choice between two keys rather than a single detection press, which adds a selection stage to every absolute time. For interference that arrives beside the target instead of before it, run the flanker task or the Stroop test; for the same filtering question with no speeded key at all, the selective attention test. Attention held over minutes rather than shifted over milliseconds is the attention span test, speed as throughput over two minutes rather than as a difference of a few milliseconds is the digit symbol substitution test, and a decision under uncertainty with no clock in it anywhere is the balloon analogue risk task.

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 ADHD, hemispatial neglect or any disorder of attention. Only a qualified professional, working with more than a browser, can make that judgment.

Where 64 cue-and-key timestamps go

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.

Each trial leaves three numbers in this tab — when the cue painted, when the dot painted, when your key arrived — and they exist as an array that is read once to form four cell means and then dropped when the tab closes. Neither storage nor a network ever sees them, so a run that is finished is finished: reloading ends it, and the summary button is the one exit those numbers have. Re-running a block overwrites that block’s array in memory and leaves the other block’s alone.