ANKLE-BITERS and picnic-trashers they may be, but ants, as Hollywood recently realised, fascinate us. And for 17 years, Deborah Gordon has succumbed to their lure, studying a species living on the US-Mexican border. At the end of it, perhaps surprisingly, Gordon isn't brimming with answers about why ants do what they do. But she does ask excellent questions, and in Ants at Work she also presents remarkable observations of how ants keep in touch with each other.

It's a well-known fact that ants touch antennae and lay chemical trails to communicate. But some forms of ant communication don't fit this explanation. For instance, every morning a small group of ants emerges to "patrol" around the nest. In some of Gordon's experiments, biologists were waiting with vacuum aspirators, and silently removed some of them. Within minutes--sometimes seconds, Gordon found--the other patrollers realised something was amiss, retreated into the nest, and shut it down for the day.

How do they do it? Ants are virtually blind: there is no way they could see the kidnappings. The aspirated ants had no time to release alarm pheromones. Were the other patrollers counting noses? And why? Can ants count? Aspirating dozens of foragers later in the day produced no reaction whatsoever. So how does the colony know that the morning checkup is crucial?

Gordon doesn't have the answers yet, but if anyone can find them, it's probably her. She's been observing about 300 colonies of red harvester ants (Pogonomyrmex barbatus) on a patch of desert on the Arizona-New Mexico border since 1981. The "pogos", as myrmecologists call harvester ants, are active in the summer--as are the professors with their armies of grad students and postdocs. The humans arise at 4.30 am, the ants at dawn. By midday, when soil temperatures reach 52 °C, the ants retreat, leaving the field to the sweating scientists.

The disappearing patrol is only one of the mysteries Gordon is hotly pursuing in Arizona. She has also unearthed some gripping questions about differences between young colonies and older ones. Harvester nests live only as long as their queen. The queen ant squats deep in the nest churning out eggs, and can live for more than 15 years. Gordon found that ants in nests that are between three and four years old are much more aggressive than older colonies in defence of their foraging areas.

This seems to make sense: older nests presumably contain older ants that would have discovered fighting did not pay. Instead, they peaceably go about their seed-gathering. The only problem is that there's no such thing as an older ant: no worker ant lives more than a year. And the queen, despite her august title, has no administrative role. So why do older colonies behave differently? Who or what has "learnt" from the past?

Gordon provides clues about why ants in younger nests fight more. They have, for instance, many helpless young to feed, with few adults foraging for food. She did wonder whether aggressive colonies might be struggling to command the best territory, but found that the seeds pogos collect are scattered randomly, so any square metre of desert is as good as the next.

She left no assumption unturned. With years of data in hand, and help from statisticians, she looked at the costs of fighting to the ants. Although ants have a ferocious image, fighting is very rare. A handful of specialised fighters do almost all of it. So fighting doesn't burden a colony.

"I see little that seems efficient about the ways that ants behave or interact with their neighbours," she says. "Their behaviour clearly works; there are a lot of harvester ant colonies out there, more each year. I deeply admire their harvester-ant-ness, the richness of their responses to a world so alien to me, but I am never struck by their perfection." "Perhaps natural selection is not operating on harvester ant behaviour," she adds. Myrmecologists, it seems, face a formidable obstacle when they seek to apply the tenets of evolutionary biology to ants: they cannot identify the parent ants so no one has yet achieved any good measure of reproductive success.

One summer day after rain, every nest in a broad area releases its "flying ants", the males and fertile females. They mate in the air, then the males die. Although larger, more successful colonies produce more "reproductives", no one knows which nests the founding males and females came from. Gordon has thought of marking queens before their flight, but that would take too long. The females would "miss their flights". So unlike primatologists counting copulations and observing females give birth, myrmecologists work with a tabula rasa when studying parentage. Genetic markers, she adds, may one day enable researchers to track colonies and their offspring.

Even the basics of ant research can be hard, even backbreaking, work. The only way to take a census of a colony, for example, is to dig out a nest. This begins with a rented backhoe and ends, hours later, with picks and shovels. The three stages of digging out a nest, reports Gordon, are "Hope, Existential Despair, and Anger". Wherever possible, Gordon has tried to employ more sophisticated technology. Video cameras have replaced film, but observing ants with a fibreoptic microscope stuck down into the nest proved mostly useless. In practice, she found that she could not make much sense of what she saw--ants, giant in close-up, rushing up and down tunnels.

She has instead opted to create lab colonies to study what cannot be seen in the wild, even though the behaviour of these ants is about as natural as that of elephants in a zoo. This resulted, for example, in an upsurge in undertaker traffic: lab ants, lacking predators, die at home in bed, and more nestmates have to work to remove their corpses.

Gordon writes clearly and with a nice sense of humour about ants and science. Ants at Work is an excellent look at how research is really done, and a self-portrait of a keen, inquiring mind at work.