Symbol Search Test

Every row asks the same thing. Two symbols sit on the left. Five sit on the right. Is either of the two over there?

There is nothing else to do and nothing to work out. You are not matching a pairing, spelling anything or holding a sequence. You read two shapes, sweep five, and press one of two buttons. That is the whole task, repeated.

Because the work is that small, the time it takes is worth measuring directly. The page reports the middle of your decision times rather than how many rows you got through. A middle time is a description of one typical decision, and a count is a description of a run.

The two symbols on the left change on every row. Hold them fixed and by the third row they would already be in your head. Reading the target group would then drop out of the task. What is left is a hunt for two shapes you were told about a minute ago. That is a memory test wearing this one's clothes.

The settings panel offers 20, 40 or 60 rows, and any of the three gives a number that means the same thing. That is a property of a median rather than a kindness.

A count cannot do this. Count how many rows you clear and the answer only means something next to the clock it was taken against. So the clock has to be the same for everyone, every time. That is exactly why the digit symbol test fixes its window at ninety seconds and offers no choice at all. The two pages sit in the same hub and take opposite decisions here, for the same reason.

A middle decision time is not built that way. Twenty decisions and sixty decisions have the same middle, give or take the wobble that comes with fewer measurements. So the length is a comfort setting. Twenty rows is a quick look, sixty is a steadier figure, and your history stays comparable across all of them.

Sixty is steadier because of how a median behaves. Any single slow row moves it hardly at all, and more rows means the middle is pinned by more of them. If you plan to compare two runs, take both at the same length and take them the same distance from a coffee.

Half the rows hold one of the two targets and half hold neither, exactly, in an order you cannot read. The result card times the two kinds separately, and they will not match.

The reason is worth a sentence. When a target is there, you can stop the moment you hit it, and on average that happens somewhere in the middle of the five. When nothing is there, no such moment arrives. The only way to be sure is to reach the end, so an empty row costs you the whole sweep every time.

That gap is one of the most reliable findings in the study of visual search. It is also why this test balances its rows rather than leaving the mix to chance. A run that happened to draw more empty rows would post a slower headline for no reason belonging to the person taking it.

A difference between two conditions being the real finding is a shape this site uses more than once. What the Stroop effect actually shows is the clearest write-up of it here, and the logic transfers directly.

• A large gap usually means you are checking the row properly rather than guessing early.
• A gap near zero with high accuracy is unusual and often means you are sweeping all five whether or not you have found the target.
• A gap near zero with poor accuracy means the empty rows are being answered before they have been read.

The comparison block on this page draws no bar. No percentile, no band, no better than most. That is deliberate and it is not what most sites offering this test will do.

Symbol search is a named subtest and it has been measured carefully for decades. All of that measurement is of a pencil. A person is handed a printed page and marks a yes or no box on every row. The published figure is how many rows were marked in two minutes, scaled for age. The number contains the speed of a hand moving down a sheet.

Nothing in that transfers to a decision timed in a browser. The hand is gone, the sheet is gone, and a per-decision time was never what was collected in the first place. There is no arithmetic that turns one into the other, only a guess dressed as a fact.

So this page shows your number, your best and your own history, and says the ranking does not exist. Where a figure does sit against published data we say so and show the curve. Average reaction time by age is the clearest example of one we can stand behind.

There is a second one, and it is relative rather than absolute. Processing speed by age gives the age curve as percentages of a young adult median. A relative curve survives the move from pencil to browser in a way that a raw count never does.

A decision here is around half a second, and only part of that is the search. Some of it is the eye moving, which it does in jumps rather than smoothly, and each jump costs a few tens of milliseconds. Some is the press itself. Some belongs to your screen, which paints when its refresh comes round rather than when it is asked.

That is why the honest comparison is you against you, on the same machine, rather than you against a stranger. Swap a 60 Hz laptop for a 144 Hz monitor and the number moves without anything about you changing. The methodology page explains what we can and cannot see from inside a browser.

What a quick time here says is that you found or ruled out a shape in a row of five quickly, on this screen, today. It is not a measure of how fast you think. It does not detect or rule out any condition, and it is not a diagnosis of anything.

Practise and the number will improve, and what improves is this task. Getting quicker at scanning rows of symbols makes you quicker at scanning rows of symbols. If something about your concentration or your vision genuinely worries you, that belongs with a doctor rather than with a browser tab.