The radial arm maze: eight arms and a running list

Opening

From above, it looks like a wheel with the rim taken off. A small round hub sits in the middle of a table, and eight narrow runways stick out from it like spokes, each one over a metre long. At the far end of every arm there's a little food cup, set down so the contents can't be seen from the centre.

A hungry rat is placed on the hub. It picks an arm, trots to the end, and eats the pellet. Back to the centre. It picks another arm, a different one, and eats again. Back to the centre. Seven choices later, it has emptied all eight cups, and it has barely set foot in an empty arm.

That's a neat trick for an animal with no map, no notebook and no counting system. How does it know where it's already been?

A forager's question

This is the radial arm maze, and it asks a question every forager has to answer. Of all the places I could visit, which have I already cleaned out? Before the water maze came along, this wheel of runways was one of the leading tests of spatial memory. It still has things to teach us.

How the maze works

The classic version has eight arms of equal length, spaced evenly around a central platform. Each arm ends in a food site. In the simplest form of the task, every arm is baited at the start of a trial. The best strategy is to visit each arm once, and never go back. Researchers call this win-shift. Once you've won food in one place, shift to somewhere new.

Every visit forces a return to the hub before the next choice. So after each meal, the rat faces the full set of eight options again. To do well, it has to hold in mind which arms it has already visited during this trial. That kind of short-lived, trial-by-trial memory is called working memory. A return to an arm that's already empty counts as a working memory error.

There's a popular variation. Bait only some of the arms, and keep the same ones baited every day. Now the rat has to learn which arms are ever worth visiting, which is a longer-term reference memory, as well as remembering which of those it has already emptied today. Entering an arm that's never baited counts as a reference memory error.

That ability to separate two kinds of memory in one simple session is a big part of the maze's appeal.

Picture a single trial from start to finish. The experimenter baits the cups, then places the rat on the hub, sometimes inside a small clear cylinder that's lifted to start the clock. The rat chooses freely. Someone, or these days a camera and software, records each arm entered and the order. The trial ends when all the food is gone, or when a set number of choices or minutes has passed. Then the maze is cleaned, and the next animal takes its turn.

Remembrance of places passed, 1976

The radial arm maze was introduced by David Olton and Robert Samuelson in 1976. Their paper had a memorable title: remembrance of places passed, spatial memory in rats. It appeared in the Journal of Experimental Psychology, in its section on animal behaviour processes.

Olton was a psychologist at Johns Hopkins University in Baltimore, and he would go on to use the maze for years in studies of memory and the hippocampus. But the first paper started with a plain question about behaviour.

The question was simple to state and surprisingly hard to answer. Could a rat remember a list of places it had just visited? In the wild, that skill would matter. An animal that keeps returning to a patch it has already stripped wastes time and energy.

The results were striking. In their first eight choices, the rats made about seven visits to new arms on average. That's roughly eighty-eight percent correct. If the rats were choosing at random, you'd expect only about five new arms in eight choices.

Olton and Samuelson worked hard to rule out cheating. Could the rats smell the food in the remaining arms? Could they follow their own scent trails? The researchers built in controls for both. In one clever test, partway through a trial, they swapped arms around. Arms that had already been run along were moved into positions the rat hadn't yet visited. The rats tended to go to the places they hadn't visited, even when that meant running down an arm they'd already used. They seemed to be remembering locations in the room, not the arms themselves.

That finding put the room at the centre of the story, just as the water maze would a few years later. Rats were using cues outside the maze to know where they'd been.

The maze caught on fast. Labs built versions with as few as three arms and as many as forty-eight, and used it to study the effects of drugs, brain lesions and ageing. For a while, it was one of the default ways to ask whether an animal's spatial memory was intact. When the water maze arrived in the early nineteen eighties, it offered a way to avoid food restriction and scent trails altogether. A widely used protocol from 2006 notes that the water maze went on to become more widely used than its predecessors, the radial arm maze among them. But the wheel never disappeared. It's still used today, often alongside the water maze, because the two tasks ask subtly different things. One asks an animal to remember a single hidden place. The other asks it to keep track of a list of places, and to update that list with every choice. As we'll hear, those turn out not to be the same skill.

What the maze shows

What does the maze show? At its core, it shows that rats have a capable memory for places, held over the course of a trial. It also lets researchers pull apart working memory and reference memory, since each produces its own kind of error. A drug, a lesion, or ageing might affect one more than the other.

The numbers are easy to picture. Imagine a rat with all eight arms baited. Perfect performance is eight choices for eight pellets. A rat that makes eleven choices to empty the maze has made three working memory errors. Track that over days, and you have a learning curve. You can also look at the very first eight choices, as Olton and Samuelson did, and count how many were new.

The maze has also travelled into human research. In 2004, Robert Astur and colleagues gave sixty-one undergraduates three tasks: a virtual eight-arm radial maze, a virtual water maze, and a paper test of mental rotation. On the water maze, men found the hidden platform better than women. On the virtual radial maze, there was no sex difference in working memory errors, reference memory errors, or distance travelled. And performance on the two mazes didn't correlate. Of all the measures, only the water maze probe trial and mental rotation were significantly linked.

That last point is important. It suggests the two tasks don't tap spatial memory in quite the same way, even when motivation, stress and movement are matched. The authors cautioned against treating them as interchangeable.

Weak spots

For all its elegance, the maze has well-known weak spots. The first is motivation. The classic version uses food, so animals are usually kept on a restricted diet to make them hungry enough to work. That takes time and care, and hunger itself can affect behaviour. Drugs that change appetite can look like drugs that change memory.

The second is smell. Even with controls, food and scent trails are a constant worry in a dry maze. Many labs rotate the maze, clean the arms between trials, or bait every arm and block some, to make odour useless as a clue.

The third is strategy. A rat doesn't have to remember anything if it simply goes round the maze in order, taking each next arm clockwise. That would produce a perfect score with no spatial memory at all. Researchers look out for that kind of sequence, sometimes called chaining, and some designs add short delays or closed doors at the hub to break it up. In the water maze, a related pattern appears when animals circle at a fixed distance from the wall.

And then there's the question of what the maze is really measuring. Rats are excellent at it, which makes it less sensitive to small effects. And as the human work suggests, a good score on one spatial task doesn't promise a good score on another.

Back to the wheel

Back to the wheel of runways, and the rat that empties eight cups in eight visits. That small piece of foraging hides a real feat of memory, a running list of places held in mind and updated with every choice. It's a reminder that spatial memory isn't one thing. A pool measures some of it. A wheel of arms measures some more.

hvsimage.com explains chaining and circling, and how to spot them in a swim path. You'll find the papers we mentioned, and a transcript, at water-maze.com.