The T-maze: left, right and a change of mind

Opening

It could hardly be simpler. A short, narrow corridor with a crossbar at the end. Turn left, or turn right. That's all there is. From above, it's a capital letter T, built from wood or grey plastic, sitting on a bench in a quiet room.

A mouse is placed at the bottom of the stem. It wanders up the corridor, pauses at the junction, and peers both ways. Then it turns left and explores to the end of that arm. A few moments later it's lifted out, and put back at the start.

Up the stem again. Pause at the junction. And this time, more often than not, it turns right.

Why the mouse switches

That small change of mind is the whole point of the T-maze. It's one of the simplest pieces of apparatus in behavioural science, and one of the most revealing. Why would an animal choose differently the second time? And what can a single left or right tell us about memory and the brain?

How the T-maze works

The T-maze is a stem and two goal arms, joined at a junction. It can be raised off the bench on legs, or enclosed with walls. There's usually a small start area at the base of the stem, and often doors that can be slid in to block an arm or hold the animal in place.

The most common way to use it relies on a natural habit called spontaneous alternation. Give a rodent two trials in quick succession, and on the second trial it tends to choose the arm it didn't visit the first time. Nobody trains it to do this. It seems to be a built-in tendency to explore somewhere new, rather than go back to somewhere already checked.

That tendency only works if the animal remembers which arm it chose a moment ago. So alternation is, at heart, a test of short-term memory for place.

There's also a rewarded version. The animal is kept a little hungry, and food is placed in the arm it hasn't yet visited. Now alternation is not just a habit, but a learned rule that pays off. Typically, the first run of each pair is a forced run, with one arm blocked off, so the animal can only go one way. The second run is the free choice. Did it go to the new arm, or back to the old one?

Each trial is quick. A widely used protocol says each one should be completed in under two minutes, though the total number of trials depends on the experiment and the statistics.

A century of alternation

Researchers have noticed rodents alternating in simple mazes for about a century. T-shaped and Y-shaped junctions were among the earliest tools of animal psychology, partly because a single choice point is so easy to build and score. You don't need a camera or a computer to record left or right. A notebook will do.

The T also grew into bigger structures. Join many T-junctions together and you get a multiple T-maze, a branching puzzle where every junction has a right and a wrong answer. Classic experiments let rats wander such mazes with no reward at all. Later, when food was placed at the end, the rats that had explored beforehand found their way quickly. That kind of result was taken as evidence that rats build a mental map of their surroundings, even when nothing is at stake. It's an idea that runs straight through to the water maze.

There's a close cousin too, the Y-maze. Its arms meet at a gentler angle and are spaced evenly, so there's no sharp turn at the junction. Many labs use the Y-maze for spontaneous alternation, letting an animal explore all three arms freely and counting how often it visits them in a fresh order.

The T-maze became much more than a curiosity once researchers linked it to the brain. Alternation turned out to be sensitive to damage to the hippocampus, the same structure later shown to matter so much in the water maze. In 1979, Carol Barnes used T-maze alternation in her study of ageing rats. Among other findings, she reported that the after-effects of strong electrical stimulation of a hippocampal pathway selectively impaired the old rats' spontaneous alternation.

In 2006, Robert Deacon and Nicholas Rawlins at the University of Oxford set the method down as a standard protocol, published in Nature Protocols. Rawlins, as it happens, was also one of the authors of the 1982 water maze study on hippocampal damage. Their protocol describes both spontaneous and rewarded alternation, and makes two points worth hearing. Both versions are very sensitive to dysfunction of the hippocampus. But other brain structures are involved too.

That second point is the honest small print. A poor alternation score suggests something is wrong, but it doesn't tell you exactly where.

What it shows

So what does the T-maze show? Its great strength is speed. A clear result can come from a modest number of trials, run in a single session, with no long training. That makes it useful for screening, for example when comparing strains of mice, or checking a new genetic model, before committing to longer tests.

Because spontaneous alternation needs no reward and no deprivation, it's also gentle. There's no swimming, and in the spontaneous version no hunger. That matters for animal welfare, and it removes some of the confounds that complicate other tests.

The numbers are easy to picture. Chance is fifty percent, since there are only two arms. Imagine a healthy mouse that alternates on seventy percent of its pairs. Now imagine an animal with hippocampal damage drifting back towards fifty, choosing as if it had forgotten the first run. The gap between those numbers is the effect researchers are looking for.

Researchers can make the task harder by adding a delay between the forced run and the free choice. A few seconds is easy. A minute or more puts real weight on memory. Watching performance fall as the delay grows gives a picture of how long the memory lasts.

The T-maze also lends itself to other questions. Reward one arm consistently, and you can test simple discrimination learning, or how quickly the animal adapts if the reward switches sides. Put different things at the ends of the arms, such as a new smell or another animal, and the choice reveals a preference. The same simple shape can ask very different questions.

The limits of simplicity

The T-maze has limits that come straight from its simplicity. The biggest is that fifty percent chance level. With only two options, a lucky guess is right half the time. That means you need many trials, or many animals, to be confident that a difference is real.

Side bias is another trap. Some animals simply prefer to turn one way, perhaps because of a draught, a sound, or a light on one side of the room. A strong side preference can swamp the alternation tendency. Good protocols vary which arm is forced first and check for bias.

Then there's the question of how the animal is solving the task. It might remember the place it visited. It might remember the turn it made, left or right, using its body rather than the room. Or it might follow a smell left on the floor from the first run. Each of those is a different kind of memory, and the T-maze alone may not separate them. Rotating the maze, cleaning it between runs, and changing the room cues are ways of finding out.

Motivation matters in the rewarded version. A hungry animal works harder, but hunger has to be managed carefully, and some drugs change appetite. That can make an effect on eating look like an effect on memory.

And a single binary choice tells you less about how the animal is thinking than a full swim path or a long exploration. You know where it went, but not much about how it got there, or what it considered along the way. Video tracking can help, by recording hesitation at the junction, speed, and head turns. Researchers have long noticed that animals sometimes pause at a choice point and look one way, then the other, before committing. Measuring that hesitation turns a yes or no answer into something richer.

And the room still matters, just as in every other maze. The orientation of the maze, the cues on the walls, and whether anything gets moved between sessions can all change how animals solve the task. A T-maze in one lab isn't automatically the same test as a T-maze in another.

Back to the T

Back to the capital T on the bench, and the mouse that goes left, then right. It's a tiny decision, made in a second or two. But behind it sits a memory of a moment ago, and a preference for the new over the known. The T-maze strips spatial memory down to a single choice, and that simplicity is both its strength and its limit.

hvsimage.com has more on tracking and analysis for T-mazes and other multi-arm mazes. You'll find the papers we mentioned, and a transcript, at water-maze.com.