Thigmotaxis and search strategies: the shape of a search

Opening

Two mice, two swims, the same pool. Both reach the hidden platform in about twenty seconds. On a results table, they look identical.

But look at the swim paths, drawn from above, and they couldn't be more different. The first mouse sets off from the wall and heads in a clean, nearly straight line to the platform. The second swims a wide circle, keeping a steady distance from the wall, round and round like a runner on a track, until it bumps into the platform on the way past.

Same time. Same score. Completely different ways of solving the problem.

The shape of a search

This episode is about the shape of a search. Not how fast an animal finds its goal, but how it goes about it. From the panicky wall-hugging of a first swim to the precise, direct route of an expert, the path tells a story that a stopwatch can't. And reading that story has become one of the most important developments in how the water maze is used.

Starting at the wall

Let's start at the wall. When a rodent is first placed in the water maze, it often swims close to the edge of the pool, sometimes fast, sometimes repeatedly, as if looking for a way out. That's thigmotaxis, the same wall-hugging we met in the open field. The word comes from Greek, thigma meaning touch, and taxis meaning arrangement.

In a new space, staying close to a wall makes sense for a small animal. It's safer from predators. In the water maze, though, it's a dead end, because the platform is never at the wall. So the first thing an animal has to learn is to let go of the edge.

A widely cited study from Dresden, published in 2009 by Garthe, Behr and Kempermann, described what happens next. They watched mice learn the water maze and saw a sequence of search patterns. They described thigmotaxis as a first and highly emotional response to a new and stressful situation, usually quickly overcome.

From random search to a direct swim

Next comes random search, where the animal covers the whole surface of the pool with no particular aim. Then scanning, where the search becomes concentrated in the central area of the pool, the place where the distant visual cues are easiest to see. The animal seems to be looking around for landmarks. In neither of these, the authors noted, is there any spatial or even directional preference.

Then there's chaining, the circling we heard in the cold open. The animal has learned roughly how far the platform is from the wall, and swims a ring at that distance. Notice what that strategy doesn't need. It doesn't need to know where in the room the platform is. Swim a circle at the right radius, and you'll hit it eventually from any starting point. It's a clever shortcut, but it's not spatial memory in the full sense.

The next step is directed search. Now the animal heads roughly in the right direction from the start, within a sort of widening triangle pointing from the start towards the goal. That's a sign that distant landmarks are being used, and in the authors' view the behaviour has become dependent on the hippocampus.

Finally, the two most precise strategies. Focal search, where the animal spends most of its time close to the goal. And direct swimming, where it heads straight to the platform, whatever its starting point.

There's one more pattern, and it appears when the platform moves. Perseverance is continuing to search where the platform used to be. That isn't necessarily a failure. It's evidence of what the animal learned before, and how quickly it lets go tells you something about flexibility.

Why the stopwatch misleads

Why does all this matter? Because the stopwatch can mislead. The Dresden team made the point directly. Simply practising non-spatial strategies, such as scanning or chaining, can produce comparable escape times, without any contribution from hippocampus-specific behaviour.

Their own experiment showed just how much that matters. It started from a computer model, which predicted that new nerve cells might be especially important when something novel turns up in a familiar situation. Moving the platform in a pool the animal already knows is exactly that kind of situation. They used a drug called temozolomide to suppress the birth of new nerve cells in the adult hippocampus, a process known as adult neurogenesis. Then they trained the mice in the water maze, six trials a day for five days, moving the platform to a new location on day four.

Both groups learned to find the platform. On a simple latency measure, the difference between them could look minimal. On the third day, for example, the average difference in escape time between the groups was at its smallest. But the strategies told a different story. The treated mice relied on scanning and directed search, and rarely advanced to focal search and direct swimming, especially after the platform moved. In the authors' words, they did not advance to spatially precise search strategies.

The study also showed how slowly that precision builds, even in healthy animals. In the control group, focal search and direct swimming were used by more than half of the animals only after more than two days of training.

It also showed how strategy analysis can rule things out. The treated mice didn't float passively, and they showed no difference in thigmotaxis. That argued against a general problem with motivation or panic, and pointed instead to a specific gap in spatial precision.

To classify thousands of swims consistently, they wrote a computer program that applied numerical rules to each path, for example how much time was spent in certain zones, and how tightly the path clustered around its centre or the goal. They reported that, tested on more than four thousand example trials, it was reliable more than ninety percent of the time. Paths it couldn't recognise were classified by hand.

The authors linked their result to an idea about what the hippocampus builds. Getting from a rough sense of direction to a precise spot needs what they called a metric map, one that holds the exact angles and distances between landmarks, rather than just a rough sense of how things are arranged. Their suggestion was that new nerve cells help to fold fresh cue arrangements into that map.

Thigmotaxis beyond the pool

Thigmotaxis has its own story too, beyond the water maze. We met the 1994 study from Rouen in our open field episode, showing that drugs which increase anxiety push wall-hugging up, and drugs which reduce it push it down. In the water maze, persistent wall-hugging is a warning sign. The Dresden authors noted that a wide range of drugs, brain lesions and gene knockouts lead to much higher levels of thigmotaxis, often with passive floating. An animal stuck at the wall can't show what it knows about the room. So if one group hugs the wall more, a longer escape time may reflect anxiety or motivation, not memory.

The same thinking applies to chaining. Specialist analysis guides note that, in a probe trial after full training, chaining usually indicates that spatial learning is impaired or absent. It could also mean the room lacks strong spatial cues, in which case the set-up needs adjusting. Or, early in training, it may simply mean that learning is still at an early stage. Tracking it across trials shows whether an animal makes the move from non-spatial to spatial searching, or never does.

The opening seconds of a swim deserve special attention. An animal may head accurately towards the right place, narrowly miss it, and then change its behaviour. In a probe trial, behaviour often changes once the animal reaches the remembered spot and finds nothing there. Whole-trial averages can hide that first moment of spatial intent. Looking at the initial path separately, for example a direct route after an initial turn, or after leaving the wall, can reveal knowledge that a single summary number misses.

The limits of strategy analysis

Strategy analysis has its own limits. The categories are human inventions laid on top of continuous behaviour. A swim might be half scanning and half directed search, and different labs draw the boundaries in different places. Change the thresholds and you change the counts.

Automated classification brings consistency, but it doesn't make the definitions any less of a choice. Results need to say which method was used, and ideally be checked against a sample of paths scored by eye.

And the strategies tell us how an animal searched, not why. Chaining might be a sign of hippocampal damage, a lack of cues, or simple inexperience. The path is evidence, but it still needs interpreting alongside the rest of the experiment, including cued trials, swim speed, and probe results.

Back to the two mice

Back to our two mice. One swims a straight line, the other a patient circle. A stopwatch scores them the same. Their paths tell us one has learned where the platform is in the room, and the other has learned how far it is from the wall. That difference, invisible in a single number, is exactly what researchers studying memory need to see.

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