You Are Here (9 page)

As far as we know, no other animal can path integrate as well as the desert ant. Given the lifestyle of these hardy critters, this is not too surprising. These ants live in barren conditions with sparse local cues to help them navigate, and climatic conditions where mistakes could be
quickly fatal. To see how our own abilities to path integrate compare with those of other animals, it would be more sensible for us to look somewhat closer to home in the great tree of life. Fortunately, there is no shortage of research on path integration in mammals like us.

One of the first path-integration experiments with mammals was conducted by a husband-and-wife team of psychologists, Horst and Marie-Luise Mittelstaedt. The Mittelstaedts used female Mongolian gerbils for their experiments, and they took advantage of the well-honed maternal instincts of nursing mother gerbils. When young pups stray from the nest, their mothers are diligent about seeking out their errant children. They pick them up gently by the scruff of the neck and return them to their nest. Even in complete darkness, mothers can retrieve their pups by localizing the tiny, high-pitched squeaks the pups produce when separated from them. The Mittelstaedts designed a circular arena with a small container on the outside edge that could hold a nursing mother and her pups. While in complete darkness, one of the pups was removed from the container and placed in the middle of the arena. The mother would instinctively begin hunting for the lost pup. Like a desert ant, the gerbil mother would search in a somewhat meandering path, but once the pup was found the mother made a beeline for the nest, just like the ants collecting food in the desert.

To test whether the gerbils were using path integration, the Mittelstaedts added a small platform to the center of the arena, which could be rotated at different speeds. Again, a pup was removed from the nest and placed on this platform. When the mother stood on the platform with the retrieved pup in her mouth, the experimenters rotated the platform. If they rotated the platform very slowly so that the mother could not sense the movement, she set off for the nest in the wrong direction. The mother’s homeward course could be predicted simply by the magnitude of the platform rotation. This
proves that, like Wehner’s ants, the Mittelstaedts’ gerbils were using path integration to track their spatial location relative to the nest.
⁷

Similar experiments with other animals have suggested that the ability to path integrate is common in nature.
⁸
Hamsters led across a large space in the darkness, following a choice morsel of food like the legendary dangled carrot, will turn and run to a hiding place once they have received their treat. Dogs shown a biscuit and then led away on a winding course while wearing a blindfold and headphones can, when released, turn and run to the location of the food with considerable accuracy. Though the path-integration abilities of dogs, gerbils, and hamsters are impressive, there have been no tests of the ability of an animal to path integrate over the same spatial scales as routinely tested in ants. In a way, this would not be a fair comparison. Because of their sensitivity to light polarization, ants have a built-in compass that can always be used to assess direction. Most other animals don’t have such an accurate compass and must rely entirely on a record of their own movements obtained from their vestibular system. To understand why this is a disadvantage, we will need to turn our eyes upward.

In another desert, far from the home turf of the African desert ant, Robert Goddard toiled in the heat of Roswell, New Mexico, following a boyhood dream to send rockets far into space, a prelude to a mission to Mars. Goddard’s quest began at the dawn of the twentieth century when, as a seventeen-year-old boy, he sat in the bough of a cherry tree, looking down at the ground and imagining the view from Mars. He dreamed of a rocket that not only could escape from the earth’s atmosphere but could be guided to a target using some kind of navigational system. Later in life, as a rocket scientist in the New Mexican desert, Goddard designed a navigational system
based on the gyroscope, a device invented 300 years earlier by the French scientist Léon Foucault.
⁹

The wedding of Goddard’s solid rocket boosters and Foucault’s gyroscope produced one of the great shapers of twentieth-century world politics: the ballistic missile. The navigational problems of ballistic missiles are not very different from those of nursing gerbils finding their way home in the dark. In both cases, knowing where you are means understanding where you’ve been. In rocket mechanics, such problems are solved using a clever combination of accelerometers and gyroscopes.

A basic accelerometer can be thought of as nothing more than a mass, a spring, and a ruler. As the mass is accelerated, it exerts a force on the spring that causes the spring to stretch. The ruler measures the extent of the stretch, and this measurement yields the size of the acceleration.

As shown in Figure 3, gyroscopes are commonly constructed using a series of rotating rings called gimbals. As the object that carries the gyroscope rotates through space, the gimbals rotate. Measuring the size of the rotation can generate information about changes in heading, or direction.

Figure 3
: The rotating wings of a gyroscope provide directional information

Both gyroscopes and accelerometers rely on some basic physical laws describing how things that contain mass move. Anything with mass contains inertia, which can be thought of as resistance to movement. Gyroscopes and accelerometers, because they rely on inertia, are said to be the instruments of “inertial navigation.” Together, these machines provide all the information necessary to calculate position, provided that the arithmetic can be worked out.

Inertial navigation is very difficult to do well over long periods of time. The path of a vehicle carrying accelerometers and gyroscopes can be reconstructed from the entire record of every
change
in heading or velocity, provided one knows exactly
when
these changes took place. But here’s the problem: no inertial guidance system has perfect precision. For that matter, no machine has perfect precision. Every measurement of acceleration or heading change contains an error, and these errors will accumulate as inexorably as the interest on delinquent income tax payments. This kind of inaccuracy, called integration drift, will become more and more serious as time goes by. There are two main ways to counteract this kind of error. One solution is to have a means of measuring velocity that does not depend on the inertial guidance system. Another is to allow the guidance system to come to rest. When the machine carrying the system halts, the velocity falls to zero and so does the value for integration drift. Both these error-correcting mechanisms are used, but the second one is obviously useful only on the surface of the planet, where friction and gravity can bring systems to a halt. It is not very good for rocket navigation, where making things stop can be tricky.

Though the gyroscopes and accelerometers in our middle ears look markedly different from those found in missiles and rockets, the principles involved are exactly the same, and the vestibular system found in mammals suffers from the same type of integration drift.

An animal using its vestibular system for navigation is subject to an accumulating error. Every time the animal turns or moves forward, the error for that movement segment is added to the error from all previous movement segments. Although ants also suffer from this cumulative error, the intrinsic error of their estimates of the sizes of turns is smaller than for mammals because the sun compass can yield more accurate estimates of turn size than can the vestibular system.

Given all this, we would expect path integration based on inertial guidance to be less accurate in mammals than it is in ants, and this is an expectation that has been confirmed in every case so far. But path integration can be carried out with other senses.

When furry creatures try to navigate in the lightless confines of a psychology experiment, they can be made to lose their way like human beings stumbling through a dark house during a power failure. It may take longer for them to stumble, but sooner or later, with enough twists and turns, integration drift will take its toll. A brief flash of light, though, like a bolt of lightning seen through a bedroom window, can reset our sense of position and turn integration drift back to zero. The details of how path integration works in darkness, and how brief visual “fixes” can reverse the accumulating errors of integration drift, have been worked out in experiments carried out with hamsters (their nice habit of stuffing food into cheek pouches and carrying it home to store in a larder makes them an excellent species for studying such problems). The main finding from these studies is that, provided there is not a great discrepancy between where a visual fix tells us we are and the location indicated by our inertial guidance system, a brief glimpse will wipe clean the slate, and re-zero integration drift. If the visual fix gives a surprising result, then it might be ignored.
¹⁰

For example, imagine that you’ve arrived at your cottage, late at night, and need to find the main power switch to turn on the
lights. You have a rough idea of which way to go from the front door, and you set off with your hands in front of you, feeling your way through darkness. A car drives past on a nearby road, and the sweep of headlights through the window provides you with a momentary visual fix. If the fix shows that you are walking at a slight angle to the target, you will correct your course. But if the flash of light suggests that you are walking in the completely wrong direction—back toward the door rather than toward the far wall where the power switch is located—you might be prone to disbelieve your eyes, wondering if the poor light has caused some kind of illusion.

In a classic study of animal navigation, Ursula von St. Paul took a group of domestic geese on a country ride in a small covered cart. The ride began at their home and proceeded through a series of complicated switchback turns along narrow lanes through varying types of terrain. For some segments of the ride, von St. Paul covered the cart with a blanket so that the geese were not able to see anything. For other segments, the cart was uncovered so that the geese could see the sights as they rode. At the conclusion of the tour, von St. Paul took the geese out of the wagon and released them. Would they be able to find their way home?

The key finding in this experiment was that the geese picked a homeward route as if the only movements they had made had been those undertaken while the cart was uncovered. While carried around in the cart, the geese would have had very little access to inertial guidance because their vestibular systems would not function well while they traveled passively in such an unnatural conveyance. But the most interesting aspect of this finding was that the geese apparently
were
able to path integrate using the flow of visual motion that they received while the cart was uncovered, and this is a
very different form of path integration from that using the vestibular senses for inertial guidance.
¹¹

Although this experiment showed clearly that path integration works well using vision, surprisingly few studies have aimed at determining how precisely this information can be used, especially in mammals. There are a few very intriguing suggestions, though, that geese are not the only animals capable of using path integration in this way.

In many laboratory studies of spatial navigation, animals are carried from their living quarters to another room that contains a testing apparatus. Many researchers have discovered that animals such as rats and mice are actually able to keep track of their orientation to the world as they are conducted passively through the hallways of a large laboratory. Some experiments require that animals lose all sense of spatial connection with the world outside the walls of the testing room, so that the experimenters can be confident that the behavior of the animals is under the complete control of cues present in the room. It can be extraordinarily difficult to produce that state of spatial detachment in an animal. In research in my laboratory (at the University of Waterloo), animals are sometimes conveyed from one place to another inside a light-free container that is rotated on a turntable en route to the testing room. If this is not done, animals often show signs that they have managed to maintain a consistent sense of direction and distance between the room in which they live and the room in which the experiment takes place, presumably by using a combination of optic flow and inertial guidance to keep track of their paths.