Kirkby and colleagues have presented new data on the cosmic-ray flux as recorded in the beryllium-10 content of deep ocean sediments. They say that the data suggests a link between the number of cosmic rays arriving on Earth and the glacial cycles. Beryllium-10 is produced when cosmic rays interact with particles in the Earth's atmosphere and then falls to the ground, where it is stored in ice or ocean sediments.

The possible links between cosmic rays and glacial cycles follows on from previous work that linked cosmic rays to climate change. In 1997 Henrik Svensmark and Eigil Friss-Christensen of the Danish Space Research Institute proposed that high fluxes of cosmic rays could lead to more clouds and a cooler climate, and vice versa. The Danish scientists proposed that changes in the strength of the solar wind -- the stream of charged particles that flows from the Sun -- could lead to changes in the cosmic ray flux.

Kirkby and co-workers have now put forward two new mechanisms that could cause the cosmic ray flux to vary. One is an orbital modulation of the geodynamo that would results in changes in the strength and direction of the Earth's magnetic field. Such an effect was recently discovered in long-term measurements of the geomagnetic field and can also, say Kirkby and co-workers, be seen in the beryllium-10 data. Measurements on stalagmites in northern Oman and the Austrian Alps provide further support for this hypothesis.

"The idea suggested is controversial but not crazy," says Peter Thejll of the Danish Meteorological Institute. "I think it is well worth discussing."

The standard insolation model of glacial cycles was first put forward by the Serbian astrophysicist Milutin Milankovitch in 1912. Milankovitch proposed that ice ages were caused by variations in the amount of sunlight hitting the Earth and were linked to a very gradual cyclic change in the shape of the Earth's orbit around the Sun. However, while the insolation model can explain a glacial cycle with a period of the 41 kiloyears (kyr) that is observed in the paleoclimatic data, it predicts a 400 kyr cycle that has not been observed. Moreover, it cannot explain a 100 kyr cycle that is also present.