Mysterious lightning flashes that appear to precede earthquakes could be sparked by movements in the ground below, US scientists say.
Unidentified glowing objects were spotted moments before major quakes in China and Italy recently.
These flickers could be triggered by shifting soil layers which generate huge electrical charge, say scientists.
Using a tub of plain kitchen flour, they discovered an entirely new physical phenomenon.
They announced their findings at the American Physical Society meeting in Denver.
"Our first suspicion was this has got to be a mistake. There must be something stupid we are doing," said Professor Troy Shinbrot, of Rutgers University, New Jersey.
"We took a tupperware container filled with flour, tipped it back and forth until cracks appeared, and it produced 200 volts of charge.
"There isn't a mechanism I know that can explain this. It seems to be new physics. "
Repeat experiments with other granular materials produced the same voltage phenomenon.
If it occurs along geological faultlines, sliding and cracking of soil grains could be generating millions of volts of electrostatic charge.
This in turn could seed lightning in the air above - creating a natural "early-warning system" for impending earthquakes.
Stories of "earthquake lights" have been recorded for 300 years, but were typically dismissed by scientists as hearsay, or fodder for UFO enthusiasts.
However in recent decades - and with the advent of YouTube - sightings of "clear-sky lightning" have been captured, analysed, and confirmed by scientists.
Videos of luminous orbs seen during the Fukushima and L'Aquila earthquakes were widely shared online.
"We want to know - why does this lightning appear sometimes but not others?" said Prof Shinbrot.
"Not every major earthquake is preceded by lightning. And not all clear-sky lightning is followed by earthquakes."
To understand the link, scientists in Turkey have erected towers that measure voltage fields in the air over earthquake-prone regions.
"They've found there do seem to be precursors for some large earthquakes - magnitude 5 or higher. But the voltage signal is not always the same. Sometimes it's high and sometimes it's low.
"Clearly there is a lot yet to be understood."
His first aim is to understand the flour experiment - what is this new, unknown mechanism which generates voltage in the powder cracks?
"This is not what you typically think of as static - it's not like rubber shoes against a nylon carpet. This is two layers of exactly the same material rubbing against each other - and generating voltage.
"How is this happening? Your guess is as good as mine.
"I think the reason that no-one has reported this before is that no-one has thought to look."