Andrew Glikson from the Australian National University makes a living by hunting for signs of ancient impact craters, many of which are not readily visible, some of which are buried under oceans that didn't used to be where they are now.

In several recent papers, Glikson maintains that the large craters found so far on Earth only account for about one-fifth of the actual number of major impacts predicted over the past 3.8 billion years. These impact predictions are based on a number of factors, including the vast numbers of craters on the Moon, which don't disappear on the geologically dead and atmospherically challenged satellite.

During these billions of years it is likely that there were at least 10 impacts producing craters more than 125 miles (200 km) in diameter and 30 impacts producing craters more than 60 miles (100 km) in diameter, Glikson figures. And he argues that at least 50 percent of these impacts would have struck locations in ancient oceans where the Earth's crust was much thinner than continental crust and, in particular, some 10 percent would have struck the thin crust adjacent to mid-ocean ridges.

A volcanic eruption is much more likely to occur if an impact occurs at one of these sites and, paradoxically, evidence of the impact is likely to be buried under the eruption and lost forever if that area is later folded, or subducted, into the Earth's crust.

But Jay Melosh, a crater expert from the University of Arizona, doubts this and other links between asteroids and volcanoes. Melosh presented his views last July at a conference on catastrophic events in Vienna.

"There is not a single clear instance of volcanism induced by impacts, either in the near vicinity of an impact or at the antipodes (opposite side) of the planet," Melosh concluded. "The possibility of impact-induced volcanism must therefore be regarded with extreme skepticism."

Researchers who are independently working on various impact mechanisms say Melosh's reasoning is flawed.

Hermann Burchard is a mathematician at from Oklahoma State University. He notes that there are several examples where either a volcanic eruption is speculated to be associated with a known impact or an undiscovered impact is speculated to be associated with a major eruption. Again the problem is that such eruptions tend to obliterate evidence of an impact.

Mark Boslough and his colleagues at Sandia National Laboratories have modeled asteroid impacts. In a 1996 paper, they predicted that the seismic energy from an impact travels through the Earth and is strongly focussed at the antipode to the impact, near the boundary of the crust and the hot, molten mantle.

This, they argue, generates instability in the upper mantle that can produce a "hot spot" under the surface, like the one that continues to form the Hawaiian Islands. They also claim that geologists have no satisfactory explanation for the generation of hot spots by other means, other than that they arise "spontaneously."

Melosh expressed skepticism about the focussing idea after calculating that the seismic energy delivered to the antipode would be insufficient to melt the rocks. Boslough counters that melting is not needed to generate an instability -- the increased temperature of the region may be sufficient to trigger the start of a hot spot.

Dallas Abbott from Columbia University, New York has taken a different approach to the whole question.

Using terrestrial evidence and crater evidence from the Moon, which retains a record of bombardment going back nearly 4 billion years, Abbott has analyzed the timing of large impacts compared with major volcanic eruptions, or mantle plumes.