The virtual water maze: the pool test for people

Opening

A student sits at a desk in a darkened room, one hand on a joystick. On the monitor, a round arena stretches out in front of her, its walls hung with pictures. There's a window on one side, a door on another. The floor is a flat, featureless expanse of blue.

She pushes forward. The view glides across the floor. She turns, and the pictures slide past. Somewhere out there is a hidden spot, invisible until she happens to pass over it. When she does, a platform rises, a tone sounds, and the trial ends.

A minute later, she starts again from a different side of the room. This time she heads, a little uncertainly, towards the window.

A maze without water

That's a virtual Morris water maze. There's no pool, no water and no rat. Yet the logic is the same. A hidden goal, cues around the walls, and a person who has to remember where things are. How did a test built for rodents end up on a computer screen, and lately inside a headset? And what's gained, and lost, on the way?

From pool to screen

The original water maze is a round pool of cloudy water with a platform hidden just beneath the surface. A rat or mouse learns to find it using landmarks around the room. It's been a standard test of spatial memory in animals for decades.

The virtual version keeps the structure and swaps the medium. The arena is drawn by a computer. The participant moves through it from a first-person view, using a keyboard, a joystick, a gamepad, or in newer systems, their own body. The hidden platform is a patch of floor or water that only reveals itself when crossed.

A session looks familiar to anyone who knows the animal task. There are training trials from varied start points, with the platform in a fixed place. There's usually a probe trial, where the platform is removed and the software records where the person searches. And there's often a visible platform trial, to check that the participant can see the goal and steer to it.

The measures carry over too. Time to reach the goal. Path length. The share of the probe spent in the right quadrant. Heading direction in the first few seconds, which shows whether the person set off the right way. Because everything happens inside the computer, position is logged perfectly, many times a second, with no camera needed.

The first virtual studies

The idea of moving animal tasks into virtual space took hold in the nineteen nineties, as desktop graphics became good enough. A landmark study came in 1998, from Robert Astur and colleagues. They built a computerized version of the Morris water task and asked a straightforward question. Would it work in people, the way the real pool works in rodents?

Across three experiments, it did. Participants learned to find the hidden platform using the cues around the virtual room. The study also reported a consistent sex difference. On average, men navigated to the hidden platform better than women, on a range of measures. The authors described those effect sizes as some of the largest ever reported, and robust across experiments. They presented the results as the first demonstration of the virtual water task's usefulness for humans.

It's worth pausing on why that mattered. For years, animal researchers had a gold standard for place learning, and human researchers had a patchwork of paper tests, questionnaires and real-world route tasks. Those were hard to compare with anything a rat did. A virtual pool offered common ground. The same hidden goal, the same distant cues, and very nearly the same measures.

In 2004, Astur's group went further. They tested sixty-one undergraduates on three tasks: a virtual water maze, a virtual eight-arm radial maze, and a classic paper test of mental rotation. Men again did better at finding the hidden platform. But on the virtual radial maze, there was no sex difference at all. And the tasks didn't correlate in the way you might expect. Only mental rotation and the water maze probe trial were significantly linked. The authors concluded that the two virtual mazes don't measure spatial memory in the same way, and shouldn't be treated as interchangeable.

Since then, the virtual water maze has spread widely. In 2021, Conor Thornberry, José Cimadevilla and Sean Commins reviewed more than two hundred virtual reality navigation papers. The most popular task, by some distance, was the virtual Morris water maze.

A bridge between species

What does the virtual task show? Most importantly, it builds a bridge. Researchers can run something close to the same experiment in a rat and in a person, using the same logic and similar measures. That makes it easier to ask whether a finding in animals has a human counterpart.

It also opens doors that are hard to open with animals. You can test patients with a condition directly, rather than relying on an animal model. You can combine the task with brain imaging, since a person lying in a scanner can still steer with a button box. And you can change the environment instantly, moving cues, shrinking the arena or removing landmarks, with no carpentry at all.

The control is exquisite. Every participant sees exactly the same room, under the same light, with the same cues in the same places. That's hard to promise in a real laboratory, where a moved chair or a new poster can change what an animal sees.

The visible platform trial earns its place here too. In the classic 1982 rat study, animals with hippocampal damage struggled to find a hidden platform, yet could swim perfectly well to one they could see. That contrast showed the problem was memory for place, not eyesight or swimming. A visible target in the virtual pool does the same job for people. If someone can steer straight to a goal they can see, but not to one they have to remember, the difficulty is unlikely to be the controls.

There's a practical side too. Every step is logged automatically, so there's no video to score and no tracking errors to clean up. And nobody has to be fished out of a pool.

The virtual room also makes classic cue experiments easy. Rotate all the pictures on the walls by ninety degrees between trials, and watch where the participant searches. If they follow the pictures, they were using those cues. If they don't, they may have been relying on something else, such as the shape of the arena or a count of steps. In a real lab, that kind of test means moving furniture. On a computer, it's a setting.

Real limits

For all that promise, the virtual task has real limits. The first is consistency. Thornberry and colleagues found that labs use very different set-ups. Arena sizes vary. The number and type of cues vary. So do the training schedules and the way trials end. Using different versions makes it very hard to replicate findings or compare results between research groups. They called for standard protocols, if the virtual water maze is to become a gold standard for human spatial research, as the real one is for animals.

The second limit is the body. On a desktop, the participant sits still. When they turn on screen, their head doesn't turn. When they move forward, their legs don't walk. In real navigation, those body signals feed into the brain's sense of position and direction. A desktop task leaves them out, which may change how people solve it.

That's one reason immersive systems have appeared. In a headset, turning your head turns the view, so the inner ear's balance organs come into play. Some set-ups add walking on a treadmill or a small treadplate, so the legs contribute too. HVS 6D VR is one example. It offers rodent-style protocols, including hidden and visible platforms, probe and reversal trials, inside a fully immersive headset, with walking options for body-based input.

Immersion brings its own questions. Some people feel queasy in a headset, so studies have to watch for that. And familiarity with video games may make steering easier for some participants than others, which is worth recording, since steering skill is not the same thing as spatial memory.

There's also the question of what counts as the same task. A rat in water is escaping something unpleasant. A person at a screen is, at most, mildly curious, or keen to do well. Motivation is different, and stress is different, so the two versions may not engage exactly the same systems.

Finally, humans aren't rats. People can count steps, use verbal rules, or say to themselves, it's halfway between the window and the door. A virtual pool can invite strategies no rodent could use, so any comparison across species needs care.

Back to the student

Back to that student at the desk, steering uncertainly towards the window. She's doing something a rat in a pool has done for decades, piecing together where she is from the room around her. The virtual version can't reproduce everything about the real thing. But it lets researchers ask the same question of a person that they ask of an animal, and that's a powerful thing to be able to do.

You'll find the papers we mentioned, and a transcript, at water-maze.com.