Ice entombed our planet hundreds of millions of years ago, and complex animals evolved in the greenhouse heat wave that followed.

Imagine the earth hurtling through space like a cosmic snowball for 10 million years or more. Heat escaping from the molten core prevents the oceans from freezing to the bottom, but ice grows a kilometer thick in the 0 degree Celsius cold. All but a tiny fraction of the planet's primitive organisms die. Aside from grinding glaciers and groaning sea ice, the only stir comes from a smattering of volcanoes forcing their hot heads above the frigid surface. Although it seems the planet might never wake from its cryogenic slumber, the volcanoes slowly manufacture an escape from the chill: carbon dioxide.

With the chemical cycles that normally consume carbon dioxide halted by the frost, the gas accumulates to record levels. The heat-trapping capacity of carbon dioxide — a greenhouse gas — warms the planet and begins to melt the ice. The thaw takes only a few hundred years, but a new problem arises in the meantime: a brutal greenhouse effect. Any creatures that survived the icehouse must now endure a hothouse.

As improbable as it may sound, we see clear evidence that this striking climate reversal — the most extreme imaginable on this planet — happened as many as four times between 750 million and 580 million years ago. Scientists long presumed that the earth's climate was never so severe; such intense climate change has been more widely accepted for other planets such as Venus. Hints of a harsh past on the earth began cropping up in the early 1960s, but we and our colleagues have found new evidence in the past eight years that has helped us weave a more explicit tale that is capturing the attention of geologists, biologists and climatologists alike.

Thick layers of ancient rock hold the only clues to the climate of the Neoproterozoic. For decades, many of those clues appeared rife with contradiction. The first paradox was the occurrence of glacial debris near sea level in the tropics. Glaciers near the equator today survive only at 5,000 meters above sea level or higher, and at the worst of the last ice age they reached no lower than 4,000 meters. Mixed in with the glacial debris are unusual deposits of iron-rich rock. These deposits should have been able to form only if the Neoproterozoic oceans and atmosphere contained little or no oxygen, but by that time the atmosphere had already evolved to nearly the same mixture of gases as it has today. To confound matters, rocks known to form in warm water seem to have accumulated just after the glaciers receded. If the earth were ever cold enough to ice over completely, how did it warm up again? In addition, the carbon isotopic signature in the rocks hinted at a prolonged drop in biological productivity. What could have caused this dramatic loss of life?

Each of these long-standing enigmas suddenly makes sense when we look at them as key plot developments in the tale of a "snowball earth." The theory has garnered cautious support in the scientific community since we first introduced the idea in the journal Science a year and a half ago. If we turn out to be right, the tale does more than explain the mysteries of Neoproterozoic climate and challenge long-held assumptions about the limits of global change. These extreme glaciations occurred just before a rapid diversification of multicellular life, culminating in the so-called Cambrian explosion between 575 and 525 million years ago. Ironically, the long periods of isolation and extreme environments on a snowball earth would most likely have spurred on genetic change and could help account for this evolutionary burst.

The search for the surprisingly strong evidence for these climatic events has taken us around the world. Although we are now examining Neoproterozoic rocks in Australia, China, the western U.S. and the Arctic islands of Svalbard, we began our investigations in 1992 along the rocky cliffs of Namibia's Skeleton Coast. In Neoproterozoic times, this region of southwestern Africa was part of a vast, gently subsiding continental shelf located in low southern latitudes. There we see evidence of glaciers in rocks formed from deposits of dirt and debris left behind when the ice melted. Rocks dominated by calcium — and magnesium — carbonate minerals lie just above the glacial debris and harbor the chemical evidence of the hothouse that followed. After hundreds of millions of years of burial, these now exposed rocks tell the story that scientists first began to piece together 35 years ago.

In 1964 W. Brian Harland of the University of Cambridge pointed out that glacial deposits dot Neoproterozoic rock outcrops across virtually every continent. By the early 1960s scientists had begun to accept the idea of plate tectonics, which describes how the planet's thin, rocky skin is broken into giant pieces that move atop a churning mass of hotter rock below. Harland suspected that the continents had clustered together near the equator in the Neoproterozoic, based on the magnetic orientation of tiny mineral grains in the glacial rocks. Before the rocks hardened, these grains aligned themselves with the magnetic field and dipped only slightly relative to horizontal because of their position near the equator. (If they had formed near the poles, their magnetic orientation would be nearly vertical.)

Realizing that the glaciers must have covered the tropics, Harland became the first geologist to suggest that the earth had experienced a great Neoproterozoic ice age. Although some of Harland's contemporaries were skeptical about the reliability of the magnetic data, other scientists have since shown that Harland's hunch was correct. But no one was able to find an explanation for how glaciers could have survived the tropical heat.

At the time Harland was announcing his ideas about Neoproterozoic glaciers, physicists were developing the first mathematical models of the earth's climate. Mikhail Budyko of the Leningrad Geophysical Observatory found a way to explain tropical glaciers using equations that describe the way solar radiation interacts with the earth's surface and atmosphere to control climate. Some geographic surfaces reflect more of the sun's incoming energy than others, a quantifiable characteristic known as albedo. White snow reflects the most solar energy and has a high albedo, darker-colored seawater has a low albedo, and land surfaces have intermediate values that depend on the types and distribution of vegetation.

The more radiation the planet reflects, the cooler the temperature. With their high albedo, snow and ice cool the atmosphere and thus stabilize their own existence. Budyko knew that this phenomenon, called the ice-albedo feedback, helps modern polar ice sheets to grow. But his climate simulations also revealed that this feedback can run out of control. When ice formed at latitudes lower than around 30 degrees north or south of the equator, the planet's albedo began to rise at a faster rate because direct sunlight was striking a larger surface area of ice per degree of latitude. The feedback became so strong in his simulation that surface temperatures plummeted and the entire planet froze over.