Fifteen thousand years ago, a vast lake sprawled through the valleys of what is now western Montana. Known as Lake Missoula, it was created when a lobe of ice moving south from Canada blocked the Clark Fork river, which drains much of the region. Then, one day, the ice dam broke. Water roared down the canyons at 100 kilometres an hour - 2000 cubic kilometres of it spilling onto the plains of eastern Washington in a few days. There it leapt river channels and scoured new paths across the intervening ridges. When the water receded, it left behind a mystery that geologist J Harlen Bretz was determined to uncover. In doing so, he challenged the foundations of an entire science.

GEOLOGIST J Harlen Bretz found himself in eastern Washington by accident. It was the summer of 1922 and he and a group of students from the University of Chicago were planning to study the glaciers of the North Cascades, a chain of mountains further to the west. They never reached their destination. "For some reason the trip was cancelled and they wound up with two or three weeks to spare," says local historian John Soennichsen, who has researched Bretz's story.

Borrowing a car from a friend, Bretz set out with his students to explore a region called the scablands, where the rich soil of Washington's wheat country is interrupted by great gashes carved into dark, basaltic rock. Bretz was captivated by this strangely scarred landscape. "The scablands", he wrote later, "are wounds only partially healed - great wounds in the epidermis of soil with which Nature protects the underlying rock."

Everyone who knew Bretz describes him as stubborn. He was also meticulous. Summer after summer he returned to the scablands, hiking back and forth across thousands of kilometres as he mapped the tortured land.

What he saw was like nothing else known on Earth. In some places there were lozenge-shaped hills, sharply pointed on their northern ends, like half-eroded islands. Elsewhere, now-dry valleys seemed half-formed, as though erosion had started then abruptly stopped. Other odd features dotted the valley floors, including giant rock basins with all the characteristics of potholes eroded into river bottoms, says Victor Baker, a palaeohydrologist at the University of Arizona's Lunar and Planetary Laboratory. "But potholes are typically a few metres across at most. These are 100 metres across and maybe 50 metres deep."

While everything pointed to an immense flood, Bretz knew such a notion would be anathema to his fellow geologists. In part that was because the quantity of water needed for such a flood would exceed the flow of all the world's modern rivers combined. Worse than that, it ran counter to the principle of "uniformitarianism", which insists that the ancient landscape was shaped by the same slow geological processes that are going on today. To uniformitarians there was one central rule, Baker says: "You don't invoke big, catastrophic processes to explain the past."

In part, uniformitarianism grew out of 19th-century science-and-religion debates over the age of the Earth and the processes that formed it; scientists didn't want to resort to miraculous-sounding cataclysms to explain the planet's history. But Baker thinks scientific snobbery also played a part. As a field science, geology was considered a poor relation to other disciplines, especially those in which theories could be tested in the laboratory. To counter this, geologists needed an underlying principle on which to ground their science. "Uniformitarianism was to make the science respectable compared to physics," Baker says.

Unfortunately, it also blinded its followers to a more important principle, which is to be open to what nature was trying to tell them. In this case, everything pointed to a huge flood that must have had something to do with melting ice-age glaciers not far to the north.

Such floods are not unknown today. In Iceland, they are called jökulhlaups, and are caused by volcanoes erupting beneath glaciers. One, in 1996, disgorged 3 cubic kilometres of water in two days, crumpling Iceland's largest bridge like a used tissue. But no jökulhlaup has ever come close to the scale needed to account for Bretz's scour channels. Besides, Bretz could find no sign of major volcanic activity beneath the Canadian ice. "He was a reluctant catastrophist," says Soennichsen.

Despite his failure to find the source of the flood, Bretz published his findings in a series of ever more provocative articles. Then, in 1927, he was invited to an elite gathering of geologists in Washington DC. At last, he thought, he was being given the chance to prove his case. He was wrong. One by one, the assembled scientists rose to squash his outrageous theory. Dispirited, Bretz boarded a train back to Chicago. After visiting the scablands for a few more years and writing several more papers, he suddenly stopped, and threw himself into the study of limestone caverns and how they formed

By then, however, Bretz's vindication had begun. As Bretz later told the tale, one man in the audience at the 1927 meeting, Joseph Pardee, turned to a friend and confided: "I know where the water came from." Unfortunately, Pardee worked for one of Bretz's biggest critics and kept what he knew to himself until 1940, when he was nearing retirement. Even then, he hid his bombshell in an arcanely titled paper about ripples on the bottom of Lake Missoula.

Pardee had begun mapping the shorelines of Lake Missoula in 1909, and some years later discovered huge ripple marks on the lake bed. They were similar to the ripples ordinary streams leave in mud bars, except that they were enormous: 10 metres high, a kilometre or two long, and spaced at intervals of 70 to 100 metres. The only thing that could possibly have created such features was an enormous current, the sort that would have been generated if the entire lake had drained almost overnight.

Interestingly, Pardee never said anything about where all the water went when it surged out of the lake. Not that he had to: anyone who had ever heard of Bretz knew the answer.

The old guard remained unconvinced. Then, in 1952, Bretz made one last trip to the scablands. This time, he found his own ripple marks - not just in one place, but in 15. While he was at it, Bretz also identified the traces of at least seven floods, caused when the ancient ice dam repeatedly formed and broke. Today, scientists think there may have been dozens of floods over thousands of years.

Scientifically, the case was closed. Bretz's contribution was finally acknowledged in 1979, when the Geological Society of America belatedly awarded him its highest honour. He is said to have told his son that he had only one regret: "All my enemies are dead, so I have no one to gloat over."

Bretz even lived long enough to learn that his floods were not unique. When Baker studied images taken by Mariner 9 as it orbited Mars in the early 1970s, he realised that similar outbursts must have happened there: features of the Martian landscape looked remarkably like those of the scablands.

Bretz would have been less pleased to learn that he is now the darling of "young Earth" creationists, some of whom see his findings as supporting their belief that the world is only a few thousand years old. He would be intrigued, though, to know that geologists now believe that even on Earth, the Missoula floods were not the only ones of their kind. There are signs of other superfloods in Canada, Mongolia, Siberia and Europe.

At the same time, there are questions about whether Lake Missoula was big enough to have created all the flooding seen in the scablands. One hypothesis that, following Bretz's example, originally flew in the face of received wisdom, is that it may have been augmented by water from beneath the ice sheet itself. That theory was pooh-poohed until studies in Antarctica revealed that there are big lakes beneath its ice, and that water can shift rapidly between them.

The idea that subglacial lakes contributed to Bretz's flooding is still a long shot. But as Baker says: "It's a principle of science that if you dismiss something as impossible, you will not learn anything about it."