Tilt the Glass. Where’s the Water?

weirdly accurate.

    Tilt the Glass. Where’s the Water?

    What is the water-level task?

    The water-level task is a classic test of spatial thinking: you see a tilted glass and show where the surface of the water goes. This version lets you try it on six glasses and find out how close you come, and which way you tend to drift. The task comes from Jean Piaget and Bärbel Inhelder’s work on how children come to understand space, and in a 2026 registered replication Tenney and colleagues gave the same glass to 370 adults in Germany.

    This test puts the same task on screen. Each of six glasses hangs tilted against a wall, with a mark where the water meets its side; you drag the surface to turn it about that mark, or use the adjust bar under the glass, and commit once. Glass 1 is the replication’s own item, a tumbler tilted 50° clockwise, answered first so that it can be set beside the 370 people who drew it on paper. Glasses 2 to 6 get steeper, from a 22° coupe to a 68° flask. Each answer is measured in degrees from where the water really settles, and within 5° counts as right: the criterion Hecht and Proffitt used in 1995 and the replication kept. You get your average error, how many glasses you got right and which way you tend to lean.

    It is not an IQ test and makes no claim about intelligence. Glass 1 is compared with people who used pencil and paper, while you turn a line with a finger or a mouse, so the percentage is a reference point, not a calibrated rank. The other five glasses have no population data at all; for them you see only your own numbers. A run is six answers, so one careless drag moves your average noticeably.

    What the test measures

    One thing: how close you set the surface of the water in a tilted glass to where it really goes.

    The task

    • Piaget and Inhelder’s water-level task, in the form used by Hecht & Proffitt (1995) and Tenney et al.’s registered replication: set the water’s surface through a point marked on the glass
    • Six glasses tilted both ways: the replication’s tumbler at 50° first, then a coupe at 22°, a carafe at 35°, a beaker at 44°, a bottle at 62° and a flask at 68°
    • Drag the surface to turn it about the mark, or use the adjust bar under the glass (arrow keys on a keyboard); one answer per glass, no retries
    • After each answer the water settles where it really goes, and an arc shows how many degrees off you were

    Your result

    • Your average error over the six glasses, in degrees, and how many of the six you got right (within 5°)
    • Which way you tend to lean: along with the glass, against it, or neither
    • Glass 1 against the 370 adults who drew the same glass on paper: the share of them you came closer than
    • A shareable card and a link that challenges a friend to get more glasses right

    Before you start

    • About two minutes; free, no sign-up
    • Works with a finger on a phone or a mouse on a computer
    • Needs a browser with 3D graphics (WebGL) switched on; sound is optional

    Water-level task 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.

    What is the water-level task?

    The water-level task is a test of spatial thinking from Jean Piaget and Bärbel Inhelder’s research on how children understand space. You see a tilted container and draw the surface of the water inside it. Many adults find it harder than it looks, which is why psychologists still study it. The test shows the right answer after each glass.

    How many people get the water-level task right?

    The result page tells you after the sixth glass, next to your own answer on glass 1. The figures come from Tenney et al.’s 2026 registered replication, in which 370 adults in Germany drew the water in a glass tilted 50°, the same glass you answer first. The page keeps the share who got it right, and which way the others went wrong, until you have committed to your own answer.

    Why is the water-level task harder than it looks?

    The result page explains it after the sixth glass, alongside which way your own answers leaned. Knowing the usual mistake first would change how you set the water, so the page keeps it back until you finish. The explanation comes from Heiko Hecht and Dennis Proffitt’s 1995 study, set beside the open data of Tenney and colleagues’ 2026 replication with 370 adults.

    Are bartenders and waiters better at the water-level task?

    No clear difference has held up. Hecht and Proffitt (1995) reported that experienced servers and bartenders did worse, and Vasta and colleagues (1997) found the opposite, both with about 20 people per group. Tenney et al.’s 2026 registered replication tested 370 people in Germany and found no significant difference between people who had served drinks and those who had not. The result page gives the figures for both groups after you finish.

    Do men do better than women on the water-level task?

    Older studies reported that men do better, but in Tenney et al.’s 2026 replication, with 370 adults, the gap between men and women was small and not statistically significant. In that sample the only background factor reliably linked to a smaller error was years of physics education, and that link was weak.

    Is the water-level task an IQ test?

    No. The water-level task measures one narrow piece of spatial reasoning, and this version gives no IQ score and makes no claim about intelligence. Glass 1 is compared with adults who drew with pencil and paper, so the percentage on the result page is a reference point rather than a rank, and the other five glasses have no population data.

    Can I do the water-level task on my phone?

    Yes. On a phone you drag the water’s surface with a finger to turn it about the mark, or use the adjust bar under the glass; on a computer the arrow keys work too. The test takes about two minutes, is free with no sign-up, and needs a browser with 3D graphics (WebGL) switched on.

    Where the numbers come from

    Glass 1 is compared with the 370 errors in the open data of Tenney et al. (CC BY 4.0), after their main-analysis exclusions: the percentage is the share of those people whose error was larger than yours, ties counting half. They drew with pencil and paper, so it is a reference only. The other five glasses have no population data. The share of people who leaned with the glass, shown on the result page, was counted from the same open data; the paper does not state it. The site does not publish its own players’ data.

    How it is scored

    Error

    The angle between your surface and where the water really settles, in degrees. Positive means the line leans along with the glass, negative means it leans the other way.

    Right

    An error of 5° or less, the cut-off Hecht and Proffitt (1995) used and Tenney et al. kept in their replication.

    Lean

    The average of your signed errors over the six glasses. Less than 2.5° either way counts as no habit.

    Glass 1 comparison

    Your glass 1 error, rounded to a whole degree, against the 370 errors in the replication’s open data. People with the same error count half.