How many things can you count at a glance?
Fewer than it feels like. Once a pile is too big to count in the time you have, you are estimating, and estimates are rough. In a 2019 study at Vanderbilt University, 71 students who saw arrays of dots for half a second and typed how many were off by 46% on average; even the most accurate student was off by 26%.
This test asks the same question with marbles. A lid lifts off a dish for about 0.7 seconds, too short to count, and you type your guess on a number keypad. Then your guess lights up on a board of 200 pads and the marbles are poured onto it, one to a pad, so everyone sees the real count and how far your guess was from it. There are five dishes, each fuller than the last. Your result gives your typical miss, whether you lean high or low, and your error in the formula Yeo, Wilkey and Price used, |guess − count| ÷ count, next to their students’ average and range.
It is not a maths test, an IQ test or a screen for dyscalculia. Five guesses are a small sample, so one lucky or unlucky dish moves the average, and the comparison with the study is a reference point rather than a ranking: their students saw up to 350 dots and were never told the count, while here you see the answer after every dish.
What the test measures
One thing: how close your eye gets to the true number of things it only glimpsed.
The task
- Five dishes of glass marbles under a domed lid, each pile bigger than the last
- The lid opens for a blink, about 0.7 seconds fully open: enough to see the pile, not to count it
- Type your guess on a round number keypad (on a computer, the number keys work too)
- After each guess it lights up on a counting board of 200 pads and the marbles are poured out onto it, one to a pad, so you see the true count and exactly how far off you were
Your result
- A verdict, from Built-in counter to Vibes-based counting
- Your typical miss as a percentage, and whether you lean low, high or neither
- Whether your misses change as the piles get bigger
- Your absolute error rate, scored like Yeo, Wilkey & Price (2019), beside their 71 students’ mean of 46% and range of 26–75%
- A strip of all five dishes: how many there were and what you said
Before you start
- About a minute; free, no sign-up
- Works on a phone or a computer; the 3D dish needs WebGL
- Sound effects are optional and can be muted
Counting at a glance FAQ
Is this test free? Do I need to sign up to see my result?
Yes, it’s free, and there’s nothing to sign up for: no account, no email, no payment. Your result appears on this page the moment you finish, and you can share it or play again straight away.
How many dots can you count at a glance?
Beyond a small handful, nobody really counts in under a second; people estimate. In Yeo, Wilkey and Price’s 2019 study, 71 university students saw dot arrays for 500 milliseconds and typed how many: their answers were off by 46% on average, and by 26% for the most accurate student.
How is the counting test scored?
The counting test scores each guess as a ratio of the true count, which gives a typical miss and a lean towards guessing low or high. It also applies the formula from Yeo, Wilkey and Price (2019), the size of each miss divided by the true count, averaged over the five dishes, so you can set your result beside their students’.
What is a normal error when estimating a number of dots?
In Yeo, Wilkey and Price’s 2019 sample of 71 students, the average absolute error was 46%, with a range of 26% to 75%. Their task was harder than this one: up to 350 dots, half a second to look, 270 guesses and never the true count. Seeing the answer helps a lot; in Izard and Dehaene’s 2008 study, one display that adults were told held 30 dots was enough to line up their estimates with it.
Why is there no percentile?
No published sample did this exact task. The closest study used bigger piles, a shorter look and no feedback, so a percentile against it would flatter you. The counting test shows that study’s average and range as a reference point instead, and the site does not publish its own players’ data.
Why did I get a different result the second time?
Each run draws new piles, and five guesses are a small sample: one lucky or unlucky dish moves the average. Seeing the true count after every dish also teaches your eye as you go, so a second run is often closer than the first.
Is this a maths test or a test for dyscalculia?
No. The counting test measures one thing, how close a quick look gets to the true number, and five guesses can’t tell you anything about your maths ability, your intelligence or whether you have dyscalculia. Those need proper assessment by a professional.
Where the numbers come from
Your absolute error rate uses the formula of Yeo, Wilkey & Price (2019) and is shown beside their 71 students’ mean and range, with no percentile: this task is easier than theirs, with smaller piles, a longer look, the true count after every dish and five guesses instead of 270. The site does not publish its own players’ data.
How the score works
Typical miss
Each miss is measured as a ratio, so guessing double and guessing half count the same, and a miss on a small pile weighs as much as one on a big pile. The average is shown as a percentage.
Lean
The same ratios averaged with their signs, so misses in opposite directions cancel. A clear average to one side means you tend to guess low or high.
Absolute error rate
|guess − count| ÷ count, averaged over your five dishes: the measure Yeo, Wilkey and Price (2019) used for each of their students.