The evidence, from northern Israel, suggests species such as Homo erectus may have been surprisingly sophisticated in their behaviour.

The find links earlier evidence of controlled fire from Africa with later discoveries in Eurasia, scientists say.

The researchers say that a wildfire is unlikely to be the cause.

Researchers from the Hebrew University in Jerusalem and Bar-Ilan University in Ramat-Gan excavated a waterlogged site at Gesher Benot Ya'aqov.

In 34m-thick ground deposits, they found numerous flint implements belonging to the so-called Acheulean tradition of tool manufacture. Some of these were burnt, while other were not.

The researchers think the clusters of burnt artefacts, which date to between 790,000 and 690,000 years ago, indicate the sites of ancient camp fires, or hearths, made by either Homo erectus or Homo ergaster .

It could have been a primitive form of Homo sapiens , they say, but other researchers consider this improbable.

"I believe fire was a very advantageous technology which empowered these humans," co-author Nira Alperson, of the Hebrew University in Jerusalem, Israel, told BBC News Online.

Some researchers believe the control of fire enabled dramatic changes in human diet, the ability to defend social groups against wild animals and aided social interaction.

'Significant find'

Professor John Gowlett, of the University of Liverpool, UK, said that the find was "very significant".

"Until now we only had two groups of early fire evidence: one in Africa that is more than a million years and one in Europe and Asia that's half a million years," he said.

"So people were rather inclined to say, well, the early group's is probably a natural fire and the later group's is probably a controlled fire.

"The thing about this find is that it lands right in the middle between those two, both chronologically and in area. It's going to make people sit up and think - it blows away any idea that you've got a distinction between the early group and the late group."

There is always the possibility the fires could have been natural. But the authors say a number of lines of evidence make this unlikely.

"They've got wood and other material around and that's not all burnt, so if someone said: 'Maybe these are hotspots caused by some big general fire', that's a good answer," said Professor Gowlett.

The fires also occur in lots of layers at the site, suggesting they are close together in time. Professor Gowlett said the body of evidence suggested wildfires were more widely spaced in geological sequences.

Evidence for controlled use of fire has been proposed for burnt bones at the site of Swartkrans in South Africa at 1.5 million years and for patches of dirt at Chesowanja in Kenya at 1.4 million years.

New data on the bones from Swartkrans using the technique of electron spin resonance appear to confirm those dates are correct.

These results were presented at the 2004 Paleoanthropology Society Annual Meeting in Montreal, Canada, in March.