An audience awaits the rolling stones

The rocks of Racetrack Playa in California's Death Valley are one of those wonderful and perturbing mysteries of nature. The individual stones are scattered across a vast and dry lake bed, and when nobody is watching, they move. A bizarre sight it is - large boulders sitting matter- of-factly at the end of long trails carved through the desiccated dirt.

Despite the fact that the area regularly attracts tourists, no reliable witness has seen the limestone chunks in motion. And without telltale footprints surrounding the scoured tracks (either animal or human), or indeed markings of any kind, the cause of this phenomenon has puzzled geologists for decades.

An obvious possibility is that the rocks slowly slide downhill. A few simple measurements quickly discount this: the ground is so flat that when it rains, a lake just 5cm deep is sufficient to cover almost all of the 10 sq km area.

The rain, as it turns out, is thought to be an important ingredient. It transforms the arid lake bed into mud so slippery a person can easily run and slide for five or six metres. Perhaps the rocks simply skate along under the push of strong winds.

A researcher tested this idea in the Fifties using an aeroplane propeller to create an artificial breeze. All he managed to do was to shift a 500g pebble 30cm or so - hardly proof of the wind theory, given that far heavier rocks move hundreds of metres. Some of the biggest weigh as much as 320kg - the same as four large men.

Professor John Reid and colleagues from Hampshire College, Massachusetts, have been visiting Racetrack Playa for much of the past decade to try to solve the mystery once and for all. The only way to really see if the wind is responsible is to take measurements. The team pushed and pulled rocks of various weights across artificially wetted sections of the lake bed, noting the precise forces required. A few simple calculations showed that even a relatively small stone of 20kg would require a 700km/hr gale to shift it.

The problem is that the rocks have very jagged surfaces, because the limestone they are made of, dolomite, weathers easily. As a result, even on mud, there is a great deal of friction. Clearly, the wind alone is not the solution.

The vital clue to the full explanation came from examining how the lumps of dolomite move relative to each other. Using a theodolite, the Reid team meticulously measured the orientations of the tracks and showed that for most of their journey, the rocks travel en masse, carving out parallel grooves over large areas. The exception is during the short time before they come to rest when they move more independently of each other. It's as if there is an invisible force holding them together that eventually lets go.

The researchers believe that they have at last found the answer. Writing in the latest issue of the journal Geology, they say the rocks become frozen into a large sheet of ice and the sheet as a whole is pushed along by the wind. The scenario is something like this - rain or melted snow first produces a very shallow lake, which subsequently freezes over. The frozen film need be only centimetres thick and riding on a thin layer of water to act like a low-friction sledge. Realistic wind speeds would be sufficient to propel it and the embedded rocks across the lake bed. Professor Reid's calculations show that a large blanket of ice with 100 smallish stones in it could be shifted by a wind as weak as 14km/hr.

The theory also explains why the stones tend to do their own thing towards the end of the journey. The initial ice sheets, which are hundreds of metres across, eventually break up into smaller pieces.

Such a natural explanation makes you wonder why wandering rocks aren't more common. The answer, Professor Reid says, is that Racetrack Playa's geography is very special. It's higher in elevation than most of the other playas in the region, making ice formation easier. And hills to the south- east provide shade until mid-morning, which means the frosty covering can persist.

But why has no one seen the rocks move? Because a particular sequence of events is needed and this occurs only rarely. There must be enough rain for a shallow lake to form. The temperature has to then drop so its surface turns to ice, and winds must remain light enough not to disturb the freezing process. But then the breeze must suddenly strengthen so that the frosty sledge can move its limestone cargo before melting. Between 1987 and 1994 there were only two major movements, one in the late Eighties and another in late 1992 or early 1993. Of course, another factor that helps to keep the rock's excursions from human eyes is that the cold and windy conditions in which they occur are pretty uncomfortable.

The next move is to try to catch the rocks in the act. Because of the inhospitable climate and the long periods where nothing happens, this will involve setting up automatic instruments.